PredictiveIntent

Cshop integrates behavioural personalisation technology from PredictiveIntent

James Doman — Fri, 03 Feb 2012 08:43:06 +0000

Cshop, a leading end-to-end ecommerce solution powering brands including Hobbycraft and Wolsey, has chosen to pre-integrate PersonalMerchant, the advanced behavioural personalisation and merchandising technology from PredictiveIntent.

The partnership enables online retailers using the Cshop platform to quickly begin increasing sales and conversions by up to 35% by optimising their sites into dynamic selling environments, based on their visitor’s behaviour.

As well as powering Amazon-style product recommendations, PredictiveIntent’s Suggestion Choreography technology understands a visitor’s unique context and allows retailers to utilise over 160 smart algorithms to target visitors with product recommendations, personalised banners and other content at each stage of the purchase journey over multiple digital touchpoint including promotional emails, mobile sites and social networks.

“We found the integration to be straight forward and can see the benefits this will bring our clients on the Cshop platform”
Garry Billson, Technical Director, Cshop

“Ecommerce personalisation is fast becoming a must-have technology for retailers of all sizes. We’re pleased to be partnering with Cshop to allow their online retail client base easier access to the latest behavioural technology to drive increased revenue and happier customer journeys.”
Neil Hamilton, CEO & Co-Founder, PredictiveIntent

Cshop retailers wanting to find out more about increasing conversions with PersonalMerchant can contact their account manager, or visit http://www.predictiveintent.com for more information.

ENDS

About PredictiveIntent

PredictiveIntent’s advanced behavioural personalisation technology allows retailers to have full and transparent control over their onsite personalisation strategy, helping retailers to grow revenues and conversions by placing the right products in front of the right people at the right time.

For more information, contact James Doman, marketing manager, at james.doman@predictiveintent.com or +44 (0) 1202 832030.

About Cshop

Cshop is a comprehensive Ecommerce platform to fit all multi-channel businesses. Packed with features to help increase sales and deliver services, Cshop is backed 24X7 by a helpful and knowledgeable development team.

Cshop has been the choice of some of the UK’s most successful retail on line businesses for almost 12 years, and our team are ready to assist with all your on line needs.

Our thoughts on recommendations and personalisation.

James Doman — Wed, 01 Feb 2012 13:57:27 +0000

There’s been a lot of talk in the ecommerce and technology press about the differences between recommendations and personalisation, so we thought we would weigh in to the debate with our thoughts.

Personalisation can’t fully exist without recommendations.

We see personalisation as using visitor behaviours, preferences and other information to suggest content – that content might be products, banners, or text. The type of behaviours we use range from buys and browses to remove-from-cart’s – they’re all valuable nuggets of information that help our algorithms calculate relevant content.

Understanding visitor preferences and other information is the first step to personalisation. For example, if a visitor looks at more red products than blue products they might be more likely to buy a red product. Knowing a visitor’s size means you can best merchandise your site around products to suit them. Even knowing their gender from a sign-up form is enough to begin adapting the experience.

Personalisation as a filter

Our framework consists of over 160 algorithms that look at the site, visitor, content, session, search terms used, or attributes.

By running these algorithms, and then filtering by our understanding of the visitor’s preferences and intent, we can best match content with visitors.

For example, a store might have the Widget t-shirt, with both blue and red versions available as child products. It might turn out that the blue version is the best selling item on the store. A visitor with a preference for red arrives on store. Standard personalisation might just ignore the red version of the best selling blue Widget top, and suggest other blue products similar to what the visitor has bought/looked at before, the visitor can find themselves in a Locked Loop.

By running our understanding over other algorithms (rather than just finding similar products), and with our Replacements technology, we see the best selling blue Widget t-shirt, know there is a red version, and will suggest it – presenting a much more relevant product suggestion than a old-season red t-shirt.

Just as you find a diamond in the rough, the jewel in the crown; personalisation is best served by the filtering of the crowd.

How To Avoid (But Still Benefit From) ‘No Search Results Found’ Pages

James Doman — Tue, 24 Jan 2012 11:39:43 +0000

As an avid online buyer and researcher, there’s nothing that disrupts my journey more than the following phrase.

No Search Results Found.

Especially when I know the site does sell what I’m looking for. Ok, I might spell something wrong, or use the wrong word for something, but asking visitors to try again can disrupt their journey, and even annoy them.

Why does it appear, and how can it be avoided?

There are three main reasons why a specific search might not surface any results.

Fuzzyness

You like potato and I like potahto,
You like tomato and I like tomahto,
Potato, potahto, tomato, tomahto!
Let’s call the whole thing off!
Ella Fitzgerald and Louis Armstrong

Mispells, typo mistakes, plural variations and sound-a-like phonetics all present extra challenges for ecommerce search engines. The very nature of online shopping (offering a wide range of products to a wide (possibly global) audience) means spelling mistakes are inevitable.

Rather than educate each one to spell properly, try using “fuzzy” search algorithms – these are readily available through various ecommerce platform plugins, APIs or third party outsourced services (including our PersonalSearch ecommerce search).

Fuzzyness uses algorithms to look at the characters used and swaps them until it can find results. For example a search for “Calvin Cline” would find “Calvin Klein” results by removing, swapping and adding characters.

Synonyms

I use a laptop, but you might call it a notebook. Users searching for “ipads” might just mean “tablet computer”. Synonyms rear their ugly heads in in all industries, from clothing to electronics, beauty to office equipment.

Synonyms are different words that have the same meaning – if you’re not catering for these in your ecommerce search then you’re probably missing out.

They’re easy enough to implement – most ecommerce platforms can handle these, they can be managed through third-party search tools or they can be set as a specific attribute.

You just don’t sell it!

Not everyone is Amazon. Sadly, you can’t offer every product on the planet, as much as you’d like to.

How can it be an opportunity?

“No results” pages are great for personalisation, using one important algorithm to find relevant products.

Similar searched and bought

This type of algorithm looks at the search term used by the visitor, and looks to see if any other visitors used the same term, then looks at the products they eventually bought.

Here’s an example.

A visitor searches for “notebook” on an electronics store, and the store doesn’t have any synonyms set up. The search returns no results.
Other people have also searched for “notebook”, have seen no results and then used category navigation to find and eventually purchase a laptop.
In this case, the most popular laptops bought by others who searched for the same term would be displayed on the “no results” page.

To maximise the chance of predicting the most effective product for the visitor, personalisation filters can ensure the returned products are relevant to the individual. Using past behaviour, we can mine visitor preferences and show, for example, a visitor who has shown interest in “Sony” products, a Sony laptop as part of their “no search results” page.

Consciously test and optimise to achieve your goals

If you could, you would want to sell each customer a more expensive version of the product they are buying. Business rule support lets you achieve higher revenues, increase profit or shift old stock by skewing suggestions so that, whilst remaining relevant to the visitor, they have your best interests at heart.

Summary

Use a search tool that uses “fuzzyness” and supports product synonyms
If you can’t find any results, show “people who searched this eventually bought” product suggestions

If you’d like to find out more about how personalisation and search technology can work together, get in touch today! Contact us, LiveChat with the blue box in the bottom right-hand corner or tweet us @TweetIntent.

Ecommerce Product Recommendation Success – Location Matters!

James Doman — Thu, 12 Jan 2012 10:14:42 +0000

Following from our previous post, where we found that up-sells perform 20 times better than cross-sells on product pages, we’ve put together a graphic showing the most successful locations for product recommendations.

Here’s the graphic:

The graph shows the percentage of revenue driven by different areas that are dynamically merchandised by our PersonalMerchant ecommerce recommendation engine technology.

PersonalMerchant helps retailers optimise their store into dynamic selling environment. For example, PersonalMerchant can adapt product recommendations to be as relevant as possible for each individual visitor. Not only does this improve visitor journeys by reducing the need to browse and search, it helps retailers increase revenue and conversions by putting the right products in front of the right visitor at the right time.

As you can see, category pages drive the most amount of revenue – an average of 24.2%. Retailers using this space are consciously achieving business results by skewing product recommendations by rules – for example, show “worst-sellers” to shift old stock, or show “most expensive” suggestions to increase revenue.

The second best-performing location is the homepage. Dynamically driven PersonalMerchant blocks on the homepage can use profiled information to present returning visitors (who are most likely to arrive direct to the homepage) with the most relevant products based on their history. Our unique, patent-pending Suggestion Choreography technology allows different algorithms and “logic” to adapt based on the indivdual visitor’s unique context – for example, once a returning visitor begins their journey our smart algorithms will change from using “profiled” behaviour to “real-time” behaviour, allowing for personalisation based on intent.

Retailers wanting to cross-sell products, in order to increase average order values and average basket sizes, can see an average of 4.6% of revenue driven by checkout page suggestion blocks. These areas, also driven by PersonalMerchant, allow retailers to promote other products, such as accessories, using a mixture of “people who bought this also bought” algorithms (filtered by business rules) and predictive, profile-based suggestions.It’s important to test checkout page suggestions, as they can have adverse affects on conversion rates.

Product page suggestions, as covered in our last blog post, are best used to up-sell similar but more expensive products. Up-sells should be smarter than just “same category, but more expensive” suggestions – taking into account a visitors profile and preferences can make that all important step from a £30 product to a £50 product – increasing your revenues and profit.

Worst-performing, out of the locations used for this research, was Order Confirmation pages. Typically, these are pages presented to visitors after they have checked out and completed an order. A simple reason for underperformance may be that customers have completed their task, and require nothing else. Testing this location with different types of algorithms will increase the performance, especially if discounted or impulse “no-brainer” products are suggested.

What do you think?

How do you think product suggestions in different locations can help a retailer, and help a visitor along their journey? Tweet us @TweetIntent and we’ll post the best responses here.

Product Pages: Should you cross-sell or up-sell?

James Doman — Fri, 06 Jan 2012 14:46:02 +0000

Having recently launched a new reporting tool, we’ve been digging around some of the statistics and numbers that we can now extract from our technology.

This statistic stood out.

Up-sells perform around 20 times better than cross-sells on a product page.

An up-sell is where suggested products are similar to the product in view, but more expensive – a great way for retailers to increase their revenue.

With our PersonalMerchant ecommerce personalisation technology, our up-sells are decided not just through price and category based rules, but by using individual visitor behavioural data to ensure we are up-selling the most relevant products. And it looks like it’s working!

It demonstrates that product page suggestions are not that effective when using the simple “people who bought this also bought” algorithms commonly used for cross-sell suggestions.

By better understanding the visitor and their unique preferences, behavioural up-sell product suggestions leads to more clicks on upsells (10 times more clicks than cross-sells), and, most importantly, result in up to 20 times more sales and revenue.

For retailers now thinking ‘how can I effectively suggest accessories if I can’t cross-sell’, think again.

How can I cross-sell more effectively?

Checkout page cross-sell suggestions receive, on average, the second highest amount of clicks after homepage product suggestions. Using “people who bought this also bought” algorithms, combined with maximum price rules and filtered by personal preferences, retailers can realise an average of 3% of revenue from effectively placed product cross-sell opportunities on the checkout page.

We’ll be releasing more stats and numbers over the next few weeks – make sure you’re signed up to receive our GoodIntent newsletter for a monthly roundup of personalisation and ecommerce industry news and views! Sign up with your email in the footer below.

What do you think?

The above is the aggregate data from all of our PersonalMerchant clients – have you experienced something different? Are cross-sells more important than up-selling?

Tweet us @TweetIntent with your views and we’ll add them to the post!

Suggestion Choreography Explained.

James Doman — Thu, 05 Jan 2012 13:29:24 +0000

Suggestion Choreography® is like our secret sauce: it’s possibly the most important part of our framework – so please forgive us if this blog post doesn’t satisfy your appetite for technical details!

In a nutshell, Suggestion Choreography® is our unique, patent-pending visitor context understanding technology. It uses what we know about the visitor and their situation, and decides what pre-set algorithms and filters (called ‘suggestion strategies’) will be used.

If you were a known visitor, we could do some very clever things with your profiled behaviour like changing the homepage and menu links to suit your needs. For totally new visitors, who have no profiled behaviour assigned to them, adapting menu links to show “Sandals” above “Trainers” might disrupt their journey.

Why do I need Suggestion Choreography®?

Not all visitors are the same. Let’s think about visitors to real-life bricks-and-mortar stores: some will be loyal, returning visitors, others will be totally visitors from out of town and others might just pop in because they’ve been told you sell what they’re looking for.

If you could, you would want to provide different types of visitors with different store layouts: loyal visitors would like the store set out in an obvious way (where you can earn extra money by psychologically placing more expensive items at eye level), whilst new visitors would like well signposted stores (where you can upsell impulse purchases).

Giving visitors different suggestions based on a visitor situations means that suggestions can be more relevant to the visitor at the time.

What does Suggestion Choreography do?

Suggestion Choreography®, in ecommerce, helps our recommendation engine and personalisation technology decide which algorithms and filters to use. An algorithm is what mathematics and behaviours we use to pick products, and a filter is used to squeeze down a massive list into just a few products – filters can be personal (e.g. ‘visitor’s favourite colour’) or business rule based (e.g. ‘only products under £5′).

How does it work?

When a visitor arrives onsite, we begin working out who they are and where they’ve come from. This can include things like how they have arrived, and if they are known.

Based on this information, the visitor session is directed to use a set of ‘suggestion strategies’ for that context. Here are some example:

A previously known visitor arriving after searching for “Adidas Trainers” on Google would see product recommendations based on the search term ‘adidas trainers’, as well as known information as filters such as shoe size, preferred colour, in stock products only, etc.
A new visitor, previously unknown, arrives straight onto a product page. Unsatisfied with the product, he clicks onto a category link. Using the details of the product, we can guess what other products he is interested in and use category page suggestions to put the right products in front of him.

How can I get it?

Suggestion Choreography® is at the heart of our technology, so whether you’re interested in our ecommerce recommendation engine technology PersonalMerchant or our IntentPredictionServer framework you’re sure to get industry leading technology at Software-as-a-Service cost.

Get in touch today to find out more!

Sell more of the products you want to sell, with category page recommendations.

James Doman — Thu, 22 Dec 2011 10:18:08 +0000

We always suggest that retailers place personalised suggestions and product recommendations at the top of category pages, as not only does it benefit the visitor by showing them relevant products quicker, it allows retailers to consciously achieve their own business goals.

Here’s an example.

A returning known visitor arrives on the “Sweatshirts and Hoodies” category page.
The first product in the top row shows the last item in the category browsed by the visitor
The other three products are the most popular sweatshirts and hoodies that are related to the visitor’s profiled colour and brand preferences, but over £40 to increase revenue

There are two main ways that category page suggestions can be integrated into your site.

Take over the first row of products

This approach involves using personalisation technology to return the first row of products, and styling these to match the normal product listings.

If looking to user this approach, there are two important factors to think about. Firstly, make sure that as users filter further into the category, the product suggestions also adapt. If not, it can disrupt the visitor journey and irritate the user. Secondly, ensure that if users select another sorting scheme (e.g. price, name, etc) the top row is taken over by the platform, as there are all sorts of complications that can arise if not.

This integrated approach means that consumers can skim over the first row as normal, without presenting any disruptions or new concepts for visitors to psychologically process.

Seperate suggestions with a block

Seperating product recommendations from standard category listings allow retailers to make these suggestions stand out more and subtly encourage visitors to consider these products.

This approach has many different uses for each different retailer. For example, one retailer might prefer to call the block Bestsellers, but actually promote “worst sellers”; another retailer might prefer to name the block “We think you might like” and show the most relevant suggestions, regardless of any price or filter.

What algorithms should be used?

There are many different algorithms that can drive the “logic” behind category page product recommendations.

Firstly, it depends on the visitors situation and context – if a visitor is new and unknown, the system won’t have any behavioural information about them and therefore should fallback to crowd-wisdom suggestion. If a visitor is returning, and therefore has a profile built up, more personalised suggestions can be shown.

Secondly, it depends how your visitors shop, and your store is set up. For example, if you have a limited product range, (e.g. you sell computers and laptops and not much more) then profiled behaviour can be taken into account more prominantly, as visitors will likely be shopping for the same or similar reasons.

If you are a department store with many different product categories, then visitors could be shopping for themselves, for family or for friends. In this case, real-time algorithms such as clickstream (people who have had a visit like you bought/browsed) or attribute shadowing (you’re looking at products made from leather in a size 12) can understand the visitor’s intent, and recommend the appropriate relevant products.

Anything else I should know?

Currently, only sites that are deeply integrated with personalisation technology can benefit from category page recommendations. This deep integration is needed because unless the personalisation technology can understand the products already shown in the natural listings, it may show duplicate products that lead to a disrupted visitor experience.

With our deeply integrated Magento extension and deep integration into other playforms, category page suggestions can appear in both PHP and Javascript mode. Javascript-only technologies might find it hard to understand which products from the listings to exclude from the suggestions.

How can I find out more?

We’re glad you asked!

You could take a look at our Technology Guide for more information on the different types of technology used, or alternatively you can:

Call us on +44 (0) 1202 83 20 30 or +1 (877) 298-9380

Email us at hello@predictiveintent.com

Live Chat with us – click the blue box on the bottom-right corner if we’re online.

Behavioral targeting in Magento with PersonalMerchant

James Doman — Tue, 13 Dec 2011 11:08:44 +0000

Retailers and agencies are flocking to Magento, the world’s fastest growing ecommerce platform, because of it’s great development community and the power and flexibility offered in both enterprise and free community editions.

By offering an accessible API, the Magento platform makes it easy for third party developers to create new and exciting features to be offered in a simple, packaged extension, such as our PersonalMerchant extension.

What is PersonalMerchant?

PersonalMerchant is quite simply, the most advanced personalisation, search, navigation and behavioral targeting solution for Magento. With over two years inhouse develpment by our Magento experts, any Magento store can accurately and effectively target visitors based on their implicit and explicit information.

What can be personalised?

As Magento is so flexible, our extension can be dropped in to template files, blocks and CMS pages with ease to allow the personalisation and adaption of:

product recommendations
search results
banners
category listings
transactional and promotional emails
blog pages and blog content

Even the look and feel of pages can be adapted based on different behavioural signals. For example:

If a visitor is female, show them female-orientated banners, products and show more customer reviews on product pages
If a visitor is male, show male-orientated banners, products and show more product details

All of this can be done with our Magento behavioural targeting extension.

It’s not just product recommendations.

Search results can also be personalised with our Magento extension. Magento stores can outsource the search decision making to our search technology, hosted in the cloud. Out of the box, our search technology catches misspells, plurals and phonetic spellings, but can use behavioural data to skew results by the individual visitor preferences – for example, if a woman is searching for jeans, don’t show men’s jeans. If a visitor is interested in pink t-shirts, put those to the top. It’s simple things like this that greatly improve conversion rates and reduce buying distances.

Find out more!

If you’d like to find out how PersonalMerchant for Magento can help you increase conversion rates, contact us today.

If you could adapt your site based on gender, would you?

James Doman — Mon, 12 Dec 2011 13:05:03 +0000

Having read this article on Practical Ecommerce, we wondered how personalisation could be used to adapt an ecommerce site based on a visitor’s gender. According to Empthatica:

Men and women use product pages differently. Women quickly scan product pages, then skip to the next, whilst men carry out thorough research, viewing all product details and pitures.
Females tend to search by brand, whilst males search by product type or category.
More men browse and research products every few days, compared to women.

Marcia at Practical Ecommerce also weighs in with some of her recommendations. She suggests:

For men, present detailed product descriptions, feature comparisons and customer reviews
Women should be offered live online chat, discussion forums, high-quality imagery, customer reviews, and videos.
Provide recently viewed items for women, as well as easy switching between related categories.

She also suggests targeting women are more sensitive to relevant promotions and sales.

This is all possible.

Everything above is totally possible – if a visitor has implicitly or explicity declared their gender, then you’ve collected a valuable behavioural signifier.

The question is, though: would you? Or instead should you strive to provide all features in the hope that something works?

However, this could just add to the general noise and clutter of a product page – restricting features based on a visitor’s gender would adapt the experience in the way that best suits them.

What do you think? Tweet us @TweetIntent with your thoughts.

Poll: do you “get” personalisation?

James Doman — Fri, 09 Dec 2011 11:01:27 +0000

Get in touch today:

Email: hello@predictiveintent.com

Call us: +44 (0) 1202 832030 or +1 (877) 298-9380

Tweet: @TweetIntent

In the Press: Astley Clarke

James Doman — Wed, 07 Dec 2011 11:02:46 +0000

One of our great clients, Astley Clarke, were kind enough to mention our technology in an article for the Sunday Telegraph’s “Seven” magazine, on how retailers are using technology to entice visitors to spend more whilst in a troubled economic climate.

Here’s our mention:

Bec Clarke, founder and managing director of the online luxury jewellery site AstleyClarke.com, knows exactly how to entice me into buying, thanks to a new program, PredictiveIntent, installed on her website.

“If you look at pearl earrings for three months, then next time you click on the site you’ll see pearl earrings. If you’ve been checking out amethyst bracelets, then we’ll show you amethyst bracelets,” she says. So that explains why I keep seeing diamond necklaces. Well, a girl can dream. But from my previous spending patterns, which an online site can easily monitor, Clarke also knows that in reality my biggest spree with her so far has been some £150 earrings.

“The high-spending customers will be sent emails showing jewellery that’s from £1,000 to £10,000, but there’s no point sending them to someone like you, it would scare you off,” she says. “Similarly, we know you respond to emails offering discounts, which is why you’re more likely to receive them than a customer who’s not influenced by them.”

Thanks Bec!

MeetMagentoUK – the video is here!

James Doman — Tue, 29 Nov 2011 10:54:18 +0000

At MeetMagentoUK last month, our CEO & co-founder Neil Hamilton presented to the 400 strong audience on how to put Magento at the heart of your personalisation strategy.

You can watch the presentation below.

Find out more

To find out more about how PersonalMerchant could help you increase conversions with your Magento store, get in touch with us today.

Send an enquiry by clicking the orange button above
Call us on +44 (0) 1202 832030 or +1 (877) 298-9380
Drop us an email at hello@predictiveintent.com
Chat with us – look for the blue box in the bottom right-hand corner

We need your input!

James Doman — Mon, 28 Nov 2011 11:45:32 +0000

We need your help!

Are you wasting hours manually linking related products? Do you despair at hours of staff time spent choosing promoted products for your homepage? Are your up-sell suggestions costing the business more than they bring in?

We’re carrying out an online survey into the cost for a retailer to manually merchandise their online store, and we need your input!

In Christmas spirit, we’re adding $1 to our Kiva loan fund for each response, loaning to third world countries to help communities prosper.

You can answer the survey here: http://predictiveintent.com/mmc

There’s only 5 quick questions, and it takes less than two minutes.

We’d really appreciate your time and help – if you have any questions, drop us an email at hello@predictiveintent.com or tweet us @TweetIntent!

Luxury jeweller Astley Clarke implements personalisation technology from PredictiveIntent

James Doman — Wed, 23 Nov 2011 10:20:28 +0000

Online designer jeweller Astley Clarke has implemented a personalised merchandising and onsite search technology from PredictiveIntent to supercharge the online purchase experience whilst reducing the need for manual merchandising.

Bec Astley Clarke, Managing Director, founded AstleyClarke.com in 2006 and continues to be the inspirational force behind the luxury etailer. The London based brand designs and creates its own-label jewellery collections, offers a high-end bespoke service and also curates exclusive jewellery collections from internationally recognised designers. Moreover, AstleyClarke.com was awarded the title of Best Luxury Brand Online by British luxury goods industry group Walpole in 2009 and is committed in its vision to celebrate design and craftsmanship in fine jewellery which in turn delivers “a new golden-age of luxury shopping”.

Astley Clarke homepage

The online store, managed by IT and Web Manager Dan Coleman, uses PredictiveIntent’s PersonalMerchant for Magento extension to connect Astley Clarke’s Magento Enterprise ecommerce platform to their cloud-hosted personalisation framework. Total implementation, including the configuration of unique logic created around Astley Clarke’s product catalogue and visitor behaviour, took only three weeks.

“Serving relevant recommendations to customers are a particular challenge when dealing with a product as personal as jewellery and the number of integration points required to deliver personalisation effectively means it is typically time consuming and costly project to implement. By combining comprehensive logic capabilities, search enhancements and a deployment that is as simple as installing and configuring a Magento extension, PredictiveIntent provides an ideal personalisation platform for Astley Clarke.”
Dan Coleman, IT and Web Manager, Astley Clarke

Astley Clarke is also using PersonalSearch, PredictiveIntent’s new onsite search technology, which uses enterprise level search-based algorithms that look at the search terms used and then skews the results by what is known about the visitor and their preferences.

“We’re really excited to be working with such a respected, high-end brand that is committed to pushing the boundaries of online technology to provide the best possible shopping experiences.”
Neil Hamilton, CEO & Co-Founder, PredictiveIntent

ENDS

Media links:

Astley Clarke homepage: http://www.predictiveintent.com/wp-content/uploads/2011/11/Astley-Clarke-screenshot.png

Neil Hamilton, PredictiveIntent: http://www.predictiveintent.com/wp-content/uploads/2011/09/2.jpg

Editor’s Notes:

PredictiveIntent

PredictiveIntent provide advanced behavioural personalisation technology for online retailers and content providers around the world through its core IntentPredictionServer framework. The company has established an international partner network to provide consultation and delivery of its SaaS and SaaP services and is demonstrating fast growth, especially within the expanding ecommerce market.

For more information about PredictiveIntent, visit the website at http://www.predictiveintent.com or contact James Doman, Marketing Manager, at james.doman@predictiveintent.com

Astley Clarke
Founded in 2006 by Bec Astley Clarke, AsltelyClarke.com sells contemporary and fine designer jewellery from over 30 of the most talented British and International jewellery designers including Monica Vinader, Carolina Bucci and Katie Hillier, in addition to own-label jewellery collections.

The brand has a keen VIP following amongst women at the very top of their professional careers in politics, culture and academia – with a celebrity following that includes, amongst others: Emma Watson, January Jones, Nicole Kidman, Claudia Schiffer, Elle Macpherson, Amanda Seyfried, Olivia Palmero and Liv Tyler. Astley Clarke is regularly featured in top media titles including; Vogue, Tatler, Harpers Bazaar, Elle, Marie Claire, Sunday Times Style, Financial Times, The Telegraph and Grazia.

So, you want to build a recommendation engine?

James Doman — Tue, 22 Nov 2011 11:49:21 +0000

We often get a lot of enquiries from people at companies who are not sure whether to build their own recommendation engine, plug in a lightweight recommendations solution, or dedicate some time to implementing “personalisation” properly. Our advice usually consists of three main points:

Focus on your goals – will spending too much time building a recommendation engine take your development cycle off track?
The importance of technology – thowing a few lines of javascript code on a side and manually uploading datafeeds might be sufficient for the time being, but it will restrict you from innovating with recommendations.
Don’t underestimate performance – can you support a 99.95% uptime with multiple redundancy systems, 60 millisecond response times, peak loads of >100 transactions per second, and more?

There is a wide, and varying range of technology available on the market from low-end, low-rent solutions to mega-solutions that require hundreds of billable consultation hours and large development teams.

Even around the middle market, there are many different variations available falling into the two main camps: Recommendations and True Personalisation. Sadly, there are many examples of companies blurring the boundaries between these two quite different groupings.

Recommendations

Recommendations are based on the “wisdom of the crowd”. They are not personalised at all.

Crowd based recommendations - is everyone in the crowd the same?

“Recommendation” technology uses variants of collaborative filtering to understand the relationships between “users” and “items” – “people who liked this, also liked” or “people like you bought”.

Companies promoting “recommendation” technology will often implement with a javascript solution using “zero integration” methods by pasting two lines of javascript code in your site’s header, whilst needing a product datafeed manually uploaded or automated by cron jobs. They drop third-party cookies and are updated whenever users browse and buy items, and return “recommendations” by painting a javascript area or iframe on your site.

They will have limited control on the algorithms behind the recommendations. For example some will have only one algorithm available for each page, such as “people who bought this also bought” on product pages, “best sellers” on the homepage and “popular products” for category pages.

And for most, that’s all. No chance to review the implementation, no ability to test algorithms – nothing.

Personalisation

“The future of the web is about personalisation… about ‘me’. It’s about weaving the web together in a way that is smart and personalised for the user.”
Tapan Bhat , Yahoo VP

Personalisation is a strategy of looking at the “user”, understanding their context, situation and preferences and predicting relevant content for them, based on implicit and explicit information such as behaviour and demographics.

Personalisation - focus on understanding the visitor

Personalisation isn’t “people like me bought” – it’s “I have declared an interest in a product with multiple different attributes, and other information stored against my profile means something – firstly, what have other people like me bought, but before returning the results, make sure I will be interested in them at some level.”

Personalisation might also be “I have been looking at a number of different products – take the last few products, what attributes do they have in common? What other products in the catalogue are like those? What have other people who have had the same journey ultimately bought? Are those products similar to the ones I’m looking at?”

There aren’t many companies doing this. Firstly, the technology needs to be able to handle the pure amount of data that is available – product data, visitor data, behavioural data. Secondly, there needs to be the flexibility in crunching the data through various algorithms to decide the best one to use – it’s not a case of one size fits all, there are many different algorithms with many different uses that work best in different situations, and a “framework” setup is best for this.

There is a difference between “personalisation” and what we are now having to call “true personalisation”. Some technology companies are using clever marketing to sell their “recommendation” tech as “personalisation” – sadly, their customers are being missold.

That’s not to say that true personalisation can’t tick the boxes for broader “crowd wisdom” uses – it can, but it can go that step further.

Personalisation technology companies, such as ourselves, prefer to deeply integrate with the content platform and instruct the platform the correct content to present for each visitor, in real time. This avoids legislation by using the first party cookie to watch behaviour, and can better understand the content catalogue by using built in attributes.

Of course, completely transparent configuration, A/B split testing and full reporting means we can work together to improve the logic that drives personalisation, testing different filters or algorithms to achieve the best results.

Finishing up…

Hopefully, this post has given you an idea into the complexity of building and maintaining a recommendation engine, and how you can be a step ahead by looking at “true personalisation” technology whilst not getting caught up by basic “crowd-wisdom” solutions.

What do you think? Let us know, tweet us @TweetIntent or comment below.

Ocado’s “Visual DNA” approach

James Doman — Mon, 21 Nov 2011 13:42:08 +0000

Having signed up for Ocado for the first time, I was surprised to see Ocado using a “Visual DNA” (VDNA) process to attempt to provide a more relevant experience for online shoppers. To better understand the user, the process asks users several questions about:

Who you might be buying for
Dietary requirements
What you usually buy

It’s a very interesting approach that uses explicit information given by the user to adapt the experience – or at least, attempt to.

Data collection

The first stage looks at who the user might be buying for and how many people are in the family (including pets) – this type of data could be very useful in recommending different sizes (a family with three children will buy larger packs of pre-packed food than a family with a baby), different offers (we’d expect buy-one-get-one-free offers to be more attractive to large families), and different sub-categories in general (chocolate bars for lunch boxes rather than packs of biscuits for the biscuit tin).

Ocado VDNA - stage one

The second stage looks at dietary information. Looking at just the title, I would expect to see perhaps allergy/intolerence options – in this case, “dietary info” refers to how much organic, meat/poultry and fish you buy, as well as the type of milk you buy. I suppose this information is useful in some ways (e.g. vegetarians wouldn’t be shown meat/poultry), but this “dietary info” could be expanded into areas like food intolerences, dislikes, dieting (e.g. not dieting, dieting to lose weight, dieting to gain muscle), and so on. Simply looking at the amount of fish bought is a little bit disappointing.

Ocado VDNA - stage two

Finally, the third stage looks at what you usually buy. This stage is quite uninspiring – you click on pictures of food that you would usually buy – and that’s all. They don’t seem to change or differ, and are seemingly random. Of course, there could be some very clever maths and assumptions behind the chosen foods, but I doubt it.

Ocado VDNA - stage three

This could be much improved if there was some way of “seeding” your profile with a natural text “shopping list” input.

Before you can see your recommended products, you have to book a delivery slot – in my opinion, this jolts the visitor out of the experience, and back into the real world. Perhaps it would be worth using the information provided in the first stage to suggest some slots – I would expect families with an adult woman, an adult male and two children might prefer an early evening or weekend slot, whilst a family with just adult woman and an adult male might prefer an early morning.

Finally, you can get to your recommended products!

Ocado VDNA - your recommended products

Data usage

Whilst I’m sure the data given is invalauable to Ocado’s marketing, positioning and even financial teams, the data doesn’t seem to be used in an innovative way at all on the site.

After completing the VDNA process, users are shown their recommendations in a pretty standard way. Users can quickly add suggested products to their basket, choosing from a range of different categories such as Fresh, Bakery and Cupboard.

Ocado VDNA - suggested products

Users can get more product recommendations from each shelf, but so far they can only add products to the cart that can be found on the page. For example, I’ve registered two accounts with different family members but can only see kitchen towel in “Home & Kitchen” – at this stage it’s impossible to find some tin foil.

To search for other products, users have to click through to the next page and then search for individual products – it doesn’t seem like users can browse through different categories as they can on Tesco.com or Sainsburys.com. In my opinion, this severely restricts the usability of the site, and probably impacts their profits too (as I would expect browsing to lead to more impulse purchases).

I’m also not sure on how much purchase data is used over different sessions and for different users, even at a crowd level. For example, the recommendations might be manually managed, or they might be “people like you bought”.

Recommendations

Sadly, I don’t think Ocado are being as innovative or helpful as they might like to be. Removing the “browsing” element and restricting the products immediately available causes more work for the user, as does seperating the search from the main experience.

Personalisation could really help smooth over the experience. Here are my recommendations on how Ocado could implement personalisation technology to make the experience better whilst selling more.

Gather more data

The first stage of the process is a really good idea – expanding this to include ages would provide even better data for targeting and marketing. Asking for other dietary information such as intolerences, dislikes or diet goals will help filter out irrelevant products that a user will never buy, whilst helping Ocado better understand visitor dieting behaviour.

The third stage is, to be honest, quite awful. A natural-search “shopping list” input will help Ocado learn more about what visitors care about (assuming importance in order of entry) whilst collecting data to seed the account with basic suggestions.

Don’t upset the status quo

Browsing is a given in online shopping – don’t change it. Using categories and faceted filtering to drill down is second nature to most users, and changing this might upset the equilibrium.

A better approach would be to include the categories and hover-over menu’s on the left hand side, whilst still showing the “shelves” in the main content area. Products on the shelves could be personalised based on crowd-wisdom, overlayed with personalisation algorithms, to reduce the need for visitors to search, whilst still retaining the ability to shop how they usually would.

Gather even more data!

Every click, browse and purchase means something. Record this and crunch it through algorithms to decide it’s importance and place in the grand scheme of things. A click that filters tin foil from a 7.5m roll to a 10m roll is not that significant – however a click that filters out dairy milk to show soya milk is quite important, and can really affect the whole experience from thereon.

Your thoughts?

If you’re an avid Ocado customer, what do you think of their “Personal Shopper” process and how do you think it could be improved?

Tweet us @TweetIntent or comment below!

InfluencedRevenueIndex #5

James Doman — Tue, 15 Nov 2011 10:20:18 +0000

This blog post is part of our monthly series that reports on the performance of our smart technology for online retail clients using our Software-as-a-Service PersonalMerchant personalisation technology.

October was a bumper month for PredictiveIntent, with record growth of signed clients and a rush to get the technology implemented before a late-November lockdown for the Christmas period.

The statistics take into account the aggregate totals of our online retail customers as of the 31st October, and some sites will be subjected to a noticable two-week lag between switching on and full power, allowing for behaviours and relationships to be mined.

It’s important to remember that all retailers have different reasons for implementing personalisation technology:

for some, it’s just to increase conversion rates by using personalisation to put the right products in front of the right people at the right time,
others use personalisation technology to increase turnover – upselling more expensive products at the expense of conversions,
some use the technology to increase the average order size.

With this in mind, let’s take a look at the numbers!

Influence on Sales

PersonalMerchant's influence on sales

The above graph looks at the sales which can be indirectly and directly attributed to PersonalMerchant.

The percentage of direct sales (where a visitor is suggested a product by PersonalMerchant, clicks on it and immediately purchases) stands at 11.79%, having hovered around 11% since August.

Overall influenced revenue (where a visitor is suggested a product by PersonalMerchant and purchases later in the same session, and including direct sales) has increased slightly from 32.87% to 33.2%. This shows that PersonalMerchant is correctly predicting and encouraging around 33% of sales.

Influence on conversions

PersonalMerchant's influence on conversion rates

Having increased conversions consistently since June, October has seen a drop in conversion rates from 1.08% to 0.98%.

We are currently in the process of adapting our statistics to look at a number of new metrics, which will be explained on this blog post over the next few weeks. As part of this, we will move away from looking at the overall “site” conversion rate as the metric is skewed by a number of factors including bot and non-store visits; instead, we will look at the conversion rate of sessions from new and returning profiled visitors, and from sessions started by search-term referrer, to more accurately present the statistics in an insightful way.

Influence on buy distance

PersonalMerchant's influence on buy distance

As with conversion rate, the aggregate buy distance has increased slightly, from 18.5 to 19.2. This coincides with the increase in influenced revenue – as retailers use personalisation technology to increase their average order sizes, visitors are going to click further by at least one click.

Predictions

November is shaping up to be another bumper month with more clients going live every day – we are predicting another slight drop in conversion rates, but buy distances and sales metrics will regain or increase.

Going into December, as retailers lock down their sites for Christmas, we expect to see higher conversion rates and sales metrics, and lower buy distances as our customer ops team revisit new client setups to futher tune the powerful logic behind the personalisation.

Case Study: 93% increase in conversions

The one statistic you need to see to realise why predictive personalised search is so important.

James Doman — Mon, 14 Nov 2011 10:20:16 +0000

Predictive, personalised search (such as our PersonalSearch technology) uses a mixture of enterprise level search technology and personalisation filters to predict a user’s search intent based on the information that is known about them. With increasing amounts of content creation and replication, search results should not only present the “right” content but also the most “relevant” content.

What happens when you search?

And, if there ever was a statistic to back this up, take a look at the keynote below at 40:08 for just a few seconds.

Did you listen to it?

This is what he said.

99.8% of users search using only a single word.

Yes – that’s right. 99.8% of users use only one word when searching. 0.1% use two words, and 0.1% use boolean expressions.

An example

An online retailer sells men’s and women’s clothes. They stock over 500 different t-shirts in both male and female, from various brands, in different colours and different styles. But the most popular search term is just “t-shirts”.

What do you present - the most popular? The cheapest? A mixture of both female and male clothing? Does the product with “t-shirt” in both the product name and description (therefore being a better percentage match) go before the product with description that calls it a “tee”?

Confusing search results on ASOS - male or female?

Or, with PersonalSearch, you could surface the most relevant, based on the individual visitors preferences and contexts.

“Hello, you’ve shown interest in grey, Animal sweatshirts from the men’s department – would you be interested in this grey, Animal men’s t-shirt?”

“Good afternoon madam, I see it’s snowing in Aberdeen – previous visitors from your area in the last two days purchased long-sleeve t-shirts, here are our two most popular in both grey and black.”

As part of our IntentPredictionServer framework, we allow for any sort of individual visitor inputs, such as location and weather lookup, explicit information such as gender and age, implicit information such as colour and brand preferences, and business rule support (“surface products with at least a 50% profit margin over others.”)

All of this information can be used to supercharge search results in order to increase the search term conversion rate, reduce click to buy distances and drive revenue through relevance.

Find out more about PersonalSearch

To find out more about PersonalSearch and how it can help you supercharge your onsite search, get in touch today:

Call +44 (0) 1202 832030 or +1 (877) 298-9380

Email hello@predictiveintent.com

LiveChat by clicking on the blue box on the botton right corner (if we’re online)

5 Steps to getting Personalisation Right

James Doman — Wed, 02 Nov 2011 12:55:35 +0000

The following was previously featured in .Net Magazine in September 2011.

Personalisation is becoming a must-have technology for e-commerce and content-driven sites, putting the right content in front of the right people at the right time. Properly configured, it transforms static sites into dynamic, personalised environments, driving revenue and engagement. Follow these five simple steps to maximise the benefits of personalisation.

1: Understand your shoppers

Personalisation is based on understanding what your visitors want – and this can vary by industry. Some customers shop by brand, by category or by product features – this should shape your strategy. Transparent configuration and continued testing will help you achieve this.

2: Use your data

You’ve probably spent a lot of effort adding information to your products – it makes sense to use it, and giving a personalisation tool more data to play with means it works harder to better target users.

3: Be selfish

Advanced filtering tools means you can skew results to ensure recommendations with the best result for your business are presented: try skewing by highest margin to increase profit, or “worst sellers” to clear old stock.

4: It’s the little things that count

There are more than enough reasons for visitors to abandon their purchase – don’t disrupt visitor journeys by suggesting unavailable products. Ensure your personalisation tool never recommends out of stock products.

5: Take it beyond the site

Personalisation can adapt every digital touchpoint your customers engage with – create personalised transaction/promotional emails, suggest relevant products on Facebook or in mobile apps, and more. The possibilities are endless!

What Google’s SSL Search means for personalisation.

James Doman — Wed, 19 Oct 2011 10:56:08 +0000

Google have announced they will be protecting the privacy of their signed-in users and as a result of this, “websites… won’t receive information about each individual query”.

At first look, this appears to be a change that affects some of our technology, mostly to do with our entrance analysis algorithms that use search term referrer keywords to present relevant products on the first page load. But then we looked through the Google blog post and read some interesting articles, to come up with the following points on where we, and personalisation technology in general, stands.

Firstly, this is only for signed-in users.

Google users who aren’t signed in to any Google services will still have their referring search terms passed over. We’re trying to get some numbers on how many Google searches are made by signed-in users and will update in due course.

PPC ads will still have terms passed over.

Businesses using Google Adwords to drive traffic to their site will still have the referring search terms passed over, as it enables businesses to measure the effectiveness of their advertising. This means, that for PPC advert links, our personalisation technology will still be able to dynamically adapt the site based on the users incoming search term.

Your SEO strategy will be based around the longtail.

Your SEO strategy will probably be based around the longtail and categories – trying to get your product pages and top-level category pages higher up the natural search results. These will be linking direct to the appropriate page, and therefore when a user arrives, they will declare their browse and our technology will begin to adapt based on the product information – such as upselling relevant similar items or listing best sellers in a category.

What does this mean for retailers?

In a nutshell, retailers will be deprived of a large amount of valuable data. But with solid SEO, PPC and personalisation strategies, retailers will still be able to maximise the chances of a conversion by presenting the right products and the right information at the right time.

A badly designed pictoral representation

As you can tell, we’re not designers – but the above represents how referreral terms will be passed depending on a user’s signed in status, and where they are likely to land.

More information?

If we can help you with any information, either on the Google announcement or if you’d like more information on how personalisation can help your site, please get in touch:

Email: hello@predictiveintent.com

Call: +44 (0) 1202 832030 or +1 (877) 298-9380

Tweet: @TweetIntent

LiveChat: if we’re online, click the blue box on the bottom-right corner to talk to us.

Why you should introduce personalisation technology on your mobile site.

James Doman — Wed, 19 Oct 2011 09:50:32 +0000

Latest figures out from the British Retail Consortium state that mobile searches account for 10% of all internet shopping-related searches.

“…Online retailing is still expanding quickly compared with selling through stores and searching from mobile devices is showing the most dramatic increase. Retailers are engaging with, and encouraging, this shift in shopping behaviour by providing more and easier ways to search and shop via smartphones and tablets.”
Stephen Robertson, Director General, British Retail Consortium

By providing easier ways to search and shop via smartphones and tablets, retailers are beginning to pay attention to accessbility issues that arise with small screens, touch-based control and the fleeting nature of mobile research. However, we aren’t aware of any retailers using personalisation on their mobile sites, but there are many benefits to be had.

Here are a few examples where personalisation can help your mobile strategy:

Entrance Analysis

Mobile searches account for 10% of retail-related searches, and retailers can benefit from entrance-analysis personalisation to put the right products in front of the right people based on their search terms. Using a mixture of search and crowd-wisdom algorithms, personalisation technology can deduce the reason for a visitor’s arrival and suggest relevant products based on the search term.

This massively helps increase conversions, as it reduces the need for visitors to search on their arrival.

Profiling

If a retailer has a loyal customer base, for example a fashion retailer such as Asos, then advanced profiling allows the site to adapt based on what is known about a visitor. This can be based on basic preferences, such as gender, to more specific attributes such as colour, brand and size preferences. Using this information, the whole mobile site can be skewed around the individual visitor, supercharging product discovery and encouraging conversions.

Connected Behaviour

It’s possible to join up the behaviour from a visitor’s computer-based behaviour and their mobile behaviour – this allows a seemless experience on all devices, and provides data that can be analysed by advanced algorithms to work out the most effective suggestions to be returned.

For example, a known visitor who has been browsing Laptops on a mobile site will perhaps be in buying mode when visiting the site on his computer – using this connected behaviour, personalisation technology can predict the product and accessories he is most likely to purchase.

Behavioural PersonalSearch

If your visitors do need to use your onsite search, it’s likely that faceted filtering will be less intiutive on a mobile screen than on a desktop or tablet – this is where behavioural search can make a difference to buy distance. Behavioural search, such as our PersonalSearch feature, uses the same behavioural data collected and used in product suggestions to find and order search listings by the most relevant to the individual visitor. Again, this can work with simple preferences such as gender to more advanced interests such as colour, brand and size.

The opportunities for behavioural search are massive – by taking into account a visitor’s connected data, behavioural profile and entrance analysis, personalisation systems can predict the most effective products to display to the individual visitor that has the best opportunity for the retailer, whether that’s to simply encourage a conversion or to skew the product journey to encourage a conversion with the best outcome for the retailer (products with the highest margin, etc).

Let’s talk

Your business can benefit from personalisation technology to maximise the ROI of your site, mobile site or app, SEO and PPC campaigns from the first day – get in touch today to find out more.

Email: hello@predictiveintent.com

Call: +44 (0) 1202 832030 or +1 (877) 298-9380

LiveChat: if we’re online, you’ll see a blue box in the bottom-right corner – click to chat!

InfluencedRevenueIndex #4

James Doman — Tue, 11 Oct 2011 11:45:51 +0000

This is the 4th issue of our InfluencedRevenueIndex series, where every month we’re reporting the success metrics for our personalisation for ecommerce technology, PersonalMerchant.

To recap August’s metrics, we began looking at the performance of PersonalMerchant based on the number of items sold, and saw an average of overall influence of 32.3%.

It’s important to remember that the InfluencedRevenueIndex looks at all of our online retail customers, for the month 1-30th September. New customers went live with PersonalMerchant at differing times throughout the month, and we have noticed a two-week period between go-live and noticable improvements on revenue.

September has seen a slight increase in the influence on sales, growing by 5%. We’ve also seen a reduction in the clicks-to-buy distance, showing that PersonalMerchant is helping to put the right products in front of the right people, faster.

Influence on sales

(Click graph to enlarge)

We can see that for the fourth month, there has been an increase of influenced sales generated by PersonalMerchant, however directly influenced sales have fallen. This can be attributed to the number of new clients who went live in September, where not enough behavioural data has been collected to supercharge relationship-driven algorithms in use.

Direct influence on sales: 11.84% of sales were directly generated by PersonalMerchant. We call this a “shotgun” purchase, where the visitor is suggested a product, clicks on it, adds to basket and checks out immediately.

Indirect influence on sales: Indirect influence on sales for September was 21.03%. These are sales where we have suggested the item in the same session as it is purchased.

Total influence: An average of 32.87% of sales were influenced or generated through PersonalMerchant.

Influence on buy distance

(Click graph to enlarge)

PersonalMerchant also tracks the number of clicks it takes a visitor to complete a purchase. A lower buy distance indicates the visitor found what they were looking for quickly, and part of PersonalMerchant’s remit is to try to reduce the buy distance by suggesting the right products in front of the right visitor, quicker. Our click-to-buy distance includes all clicks to pages viewed before the transaction in that session – this includes blog pages, homepage, category and product page views, as well as checkout process clicks (where we tracking them).

We can see the beginings of a steady trend of lower click to buy distances, having dropped from cicks to buy distance fell from 23 in July to 18.5 in August. This also fits with a reduction in the average number of unique page views per visitor, which fell from 1.97 in July to 1.94 in September.

Interested in finding out how we can improve your revenue and lower click distance with personalisation?

Get in touch today:

Email: sales@predictiveintent.com

Call – UK: +44 (0) 1202 83 20 30

Call – US: +1 (877) 2980-9380

Live chat: If we’re online, you’ll see a blue box on the bottom right corner of your screen – click and talk to us.

Relevant personalization influences 80% of Nuwear.com sales

James Doman — Thu, 06 Oct 2011 08:19:33 +0000

Nuwear.com, the Richmond, VA men’s underwear and swimwear online retailer, is seeing 25% of sales generated by industry-leading personalization technology from PredictiveIntent after only three months of operation.

Boston ecommerce agency One Pica introduced Nuwear.com, to PredictiveIntent’s advanced PersonalMerchant personalization solution as part of a site redesign, tasked with increasing sales and conversion rates whilst reducing the buying distance.

“We evaluated many different personalization providers, and we were surprised at how comprehensive PredictiveIntent’s offering is, compared to some of the other providers here in the US. The ability to set up unique logic that drives personalization for individual visitors based on their interests and situation is really powerful. Additionally, the metrics are impressive and we would thoroughly recommend PersonalMerchant to any retailer.”
Greg Segall, President, One Pica

Together, Nuwear and PredictiveIntent have created a unique personalization strategy that understands their visitors, customers and products – all made possible by PredictiveIntent’s transparent configuration dashboard which allows granular targeting based on individual visitor interests.

PersonalMerchant creates rich profiles of every visitor to Nuwear.com and analyses how each visitor found the store, what their interests are and their predicted purchase intent. This behavioural data is then fed back into patent-pending Suggestion Choreography® technology that understands each individual visitor’s context, and concurrently selects the best algorithm and filters to accurately predict what the visitor is intending to buy. These profiles can then be used to personalize transactional and promotional emails and any other digital consumer touchpoint.

PersonalMerchant’s performance can be clearly observed through an online reporting dashboard to prove the effectiveness of the real-time technology. Currently, 80% of sales are influenced by correctly predicted by PersonalMerchant, which plugs straight into Nuwear’s Magento Enterprise ecommerce platform with a mature, deeply integrated extension that supports various complex features including multistore, multicurrency, configurable and composite products.

“We are really impressed with the results we’ve seen from PredictiveIntent. With their integrated Magento extension and their ability to configure the logic behind the personalisation for individual visitors, PredictiveIntent’s technology was a great fit for us and has exceeded our expectations.”
Edward Upton, owner, Nuwear

25% of Nuwear’s sales are directly attributable to PersonalMerchant through “shotgun” purchases – where a visitor is suggested a product, clicks for more information and purchases immediately – representing 21% of their total revenue.

“Not only do these statistics show that our technology is correctly predicting purchases, it shows we are actively driving sales and generating revenue for Nuwear through smart personalization logic. It cements personalization’s place as an exciting technology that retailers can see a great return on investment in a short time.”
Neil Hamilton, CEO & Co-Founder, PredictiveIntent

Retailers looking to understand how they can increase their revenue and improve product discovery can download the case study from http://www.predictiveintent.com/nuwear-cs.

ENDS

About PredictiveIntent

For more information about PredictiveIntent, visit the website at http://www.predictiveintent.com or contact James Doman, Marketing Manager, at james.doman@predictiveintent.com

Facebook takes 16% of US online time. How?

James Doman — Mon, 03 Oct 2011 11:24:57 +0000

You may have seen this article and accompanying graph over the internet lately – The Facebook Chart That Freaks Google Out.

The graph shows where US consumers spend their online time, and it’s not surprising that Facebook is rocketing whilst mature internet destinations are becoming less sticky.

Internally, we’ve been discussing why we think Facebook is powering ahead, and we came up with four primary reasons, listed below. But first, let’s look at why Facebook want you to spend more time on site.

The Why

Advertising. To those “social media gurus” who believe Facebook is all about the “engagement”, the “interaction”, the “social lifeline” - sorry, but it’s not. Facebook make money through advertising – cost per impression and cost per click.

Cost per impression – the more page views you make, the more impressions you add to Facebook’s ad impressions, which makes them money quicker.

Cost per click – the more ad views you’re presented, the highier likelihood that you’ll find one interesting enough to click on, and make Facebook money for that click.

So, by understanding why Facebook want to increase “engagement” (read: impressions of adverts), we can then understand why they are doing the things they are doing to make their platform sticky.

Reason 1: Semi-Closed Platform

Facebook is a semi-closed platform – users live in there. Communicating with friends, playing games, watching videos, looking at photos, creating events, connecting with brands, showing interests in specific shows or songs, asking questions, shopping – you can do almost anything within the confines of the 3-column page with the blue header.

And Facebook know about it – which leads us on to reason two.

(P.S – having a semi-closed platform also means they can put adverts everywhere – on your profile, on pages, on app pages, etc.)

Reason 2: Integrated Platform

Everything you do on Facebook is recorded and related back to datapoints – that might be your profile, a friend, a brand, a page, an external property (a Liked website, for example), an integrated property (a Youtube video), a photo, etc.

Facebook can then understand the context of your social “transaction” – tagging a friend with you in a photo means you will be quite close friends; mentioning a brand in a status update with a negative word means you probably won’t want to see them in your timeline.

With Facebook’s recent “Like”, “Watch”, “Read” and “Listen” button extensions, they can now better understand your “transaction” with other properties in the appropriate context. Integrating this with other services, such as Netflix and Spotify, also gives Facebook more data to play with, such as your previous history, genres you might be interested in, actors you like, etc.

The “Like” button has got Facebook so far, but the three new buttons really help Facebook take the next big step into understanding your life.

(P.S. – having an integrated platform also means they can put more relevant adverts in front of you – if they know you like Johnny Depp, you might click on an advert for the latest Pirates of the Caribbean film.)

Reason 3: Explicit, fact-based personalisation

You’re a man: aged 28, living in London, you note Music as one of your interests, and you “Like” Oasis. One of your female friends posts a Youtube video to a new song by a pop girl group – it probably won’t appear. One of your male friends posts a Youtube video for Liam Gallagher’s new solo single – it’s a Top Story.

This isn’t coincidence. Facebook is continually using everything it knows, factually, about you, other users and items, to skew your Facebook experience to encourage you to “engage” more.

(P.S – this is why Facebook is so good for advertisers – they can target users by age, location, interests and more.)

Reason 4: Implicit, behaviour-based personalisation

This is where Facebook can get clever. With every social “transaction” you make, you also give away lots of other information that means something. Here are a few examples:

Tagging yourself and a friend in the same photo means you might be quite close friends – they can promote that friend’s status updates more.
Liking a Facebook Page because your friend likes it means they might be an Influencer – they might promote that friend’s status updates more.
Continually looking at and leaving comments on a particularly interesting photo means the photo is good, and will be promoted to encourage other friends to click on it
Making positive status updates about a brand might mean Facebook will promote their adverts higher up the page.

There are many examples that look at different behaviours, from whether you scrolled to the time you spent on the page, and more. All of the data is available to be collated, crunched and understood by complex algorithms; it can be manually analysed to provide new user insights which can then be sold or published.

Concluding

This blog post covers the top 4 reasons we think can be attributable to Facebook’s massive 16% share of US consumer’s online time, why we think Facebook are determined to build a sticky platform, and takes a look at how they have used personalisation to increase “engagement”.

Personalisation is now no longer just for the big players – our IntentPredictionServer helps content publishers, aggregators, and any other platform, services or systems to improve engagement and pageviews by implementing personalisation technology that utilises onsite implicit behaviour with explicit information to dynamically adapt user experiences.

Find out more about the IntentPredictionServer.

PredictiveIntent launch supercharged PersonalSearch technology

James Doman — Wed, 28 Sep 2011 13:34:38 +0000

Online retailers looking to maximise revenue from the rich behavioural data their visitors create can now benefit from PersonalSearch, a new supercharged behavioural onsite search solution from leading personalisation technology company PredictiveIntent.

Used in conjunction with PersonalMerchant, the company’s advanced ecommerce personalisation platform, PersonalSearch uses a range of individual visitor behaviours, crowd wisdom and a deep understanding of product catalogues to decide the most relevant search results for each visitor.

Fully transparent configuration means that retailers and agencies can use the core framework, algorithms and unlimited product/visitor filters to create an onsite search strategy that fits their visitors, customers and business – for example, a retailer might want to increase profit and therefore return relevant search results skewed by profit margin.

By mining individual behaviours, search results can be ordered by interests and preferences that visitors declare in their session, such as brand, colour, gender and more – adapting standard onsite search into a truly personalised experience.

Retailers can create unique onsite search strategies to improve conversions using different algorithms for all or part of their search results, including:

product attribute matching – using the product catalogue details to match results
fuzzy string matching – to catch miss-spells, sound-alikes and single/plural variations, and
behaviour-driven result matching – using visitor preferences (for example brand and colours) to surface relevant results

Retailers can use all existing product information and don’t need to upload product catalogue datafeeds because of deep integration to ecommerce platforms, which allows PredictiveIntent’s technology to have a real-time view of all product details, including availability, updates and additions. Built in split testing allows the continual measurement of highly-configured PersonalSearch technology against basic and simple search logic, allowing for future refinement and optimisation.

“We’re fast becoming a disruptive force in the ecommerce personalisation market because of our technically superior platform, deeply integrated approach and more efficient pricing model. PersonalSearch is a natural extension to our core framework and company vision, helping visitors easily find the products they want to buy quicker, whilst helping online retailers improve their conversion rates and revenue.”
Neil Hamilton, CEO & Co-Founder

ENDS

About PredictiveIntent

PredictiveIntent is a privately held UK company, providing behavioural personalisation technology for online retailers and content publishers/aggregators around the world. The company has established an international partner network to provide consultation and delivery of its SaaS and SaaP services and is demonstrating fast growth, especially within the expanding ecommerce market.

For more information about PredictiveIntent, visit the website at http://www.predictiveintent.com or get in with with James Doman, Marketing Manager, at james.doman@predictiveintent.com

We’re presenting at Meet Magento UK!

James Doman — Thu, 22 Sep 2011 08:40:49 +0000

We will be presenting at Meet Magento UK on 26th October 2011 at 11:20am.

We’ll be giving away an iPad! Give your business card to one of the team and we’ll enter you into the draw!

“Putting Magento at the heart of your e-commerce personalisation strategy”

Personalisation is now a must-have technology for any online retailer. Once connected, personalisation technology can drive up to 40% of sales and influence up to 80% of revenue by putting the right products and content in front of the right visitor at the right time.

Retailers can now implement advanced personalisation technology into their Magento store in less than a day, and have full control over the real-time logic that crunches together individual visitor behaviour, the “wisdom of the crowd” and all available product information to decide the most relevant products for each visitor. It’s also possible to consciously achieve business goals, for example to promote higher margin products to drive profit. The rich, behavioural data gleaned from your onsite visitors can also be used to personalise other consumer touchpoints such as transactional and promotional emails, Facebook apps and pages, forums and more.

Join Neil Hamilton, CEO & Co-Founder of PredictiveIntent, as he demonstrates in real-time how personalisation can adapt a site around each visitors unique context, interests and intent, and the statistics to prove the substantial return on investment from smart technology.

Find out more and register your place now: http://www.meet-magento.co.uk

Neil Hamilton is the CEO & Co-Founder of PredictiveIntent, a true behavioural personalisation and recommendation engine technology platform. With over 15 year’s experience working with disruptive technologies, Neil is committed to helping retailers increase their revenue with relevance using the most advanced personalisation solution available.

Content Creation and Replication

James Doman — Tue, 20 Sep 2011 08:45:33 +0000

I found this amazing statistic whilst researching for a magazine article.

We [mankind] created and shared 23 exabytes of data in 2002 – in 2010, we were sharing 23 exabytes of data every seven days. That’s 1196 exabytes of data in 2010.

Two more stats to put it into perspective:

In 2006, we created 161 exabytes of data – “three million times the information contained in all books ever written” – source.
In May 2009, it was estimated that “the internet” was only 500 exabytes big – source.

Click to enlarge

With the massive amount of content creation, how can you ensure users and visitors can find the content they are interested in?

New case study coming soon – personalisation for content driven sites.

InfluencedRevenueIndex #3

James Doman — Wed, 07 Sep 2011 13:22:50 +0000

Welcome to issue 3 of our InfluencedRevenueIndex series! Every month we’re reporting the success metrics for our personalisation for ecommerce technology, PersonalMerchant.

August was a bumper month with many new customers brought online – with our advanced cold-start algorithms we can power personalised recommendations from the get-go, increasing revenues from implementation.

In July’s IRI post, we mentioned we would change the main metric we use to report our influence. This is because a larger number of our clients are using multi-store/multi-currency implementations and using A/B split testing more often, which makes the reporting the impact on our technology on every currency quite difficult.

From now on, we will look at the influence of our technology on the number of items sold, which is measured in exactly the same method as revenue:

Influenced sales – where items were suggested at some point along the visitor’s journey, which were then later bought in the same session
Direct sales – where items were suggested, immediately clicked on and purchased straight away

August is also the third mass-collection of data, so we have three rolling months of data to compare. Let’s begin with this month’s numbers!

Influence on sales

(Click graph to enlarge)

As you can see, the above stacked graph shows three values over three months: total revenue as 100%, the percentage of indirect influence, and direct sales as a percentage of indirect influence. This statistic doesn’t include the effects of email personalisation.

Total influence on sales: our total influence on sales for August is 31.29%. Total influence has consistently grown over the three months.

Direct influence on sales: Our direct influence on sales in August stands at 12.14%, having grown from 8.8% in June. We can see a jump in direct influence from 8% in June to 11% in July, however there was only a 1% jump between July and August – however, more of our retailer clients are using our technology to increase average order values and basket sizes through effective cross-selling, reducing the amount of “shotgun” purchases, which pushes the sale to an indirect influenced sale.

Indirect influence on sales: Indirect influence on sales for August was 19.15%, an increase of 3.2% from June. Again, we can see there was a larger jump between June and July than between July and August. It’s worth remembering here that these numbers are looking at from all of our online retail customers, and that retailers are using the technology in different ways – for example an online music store are using PersonalMerchant to suggest relevant content, rather than products.

Influence on buy distance

(Click graph to enlarge)

Clicks to buy distance fell from 23 to 19.2 in August, whilst June and July’s figures were pretty consistent with July being 0.03 higher, however in July we began working with a retailer who has a large blog and content section, which can explain the slight increase in average click distance to purchases.

CLick to buy distance includes all clicks to pages viewed before the transaction in that session – this includes blog pages (if we are tracking them), all homepage, category and produt page views, as well as checkout process clicks (where we are tracking them).

Interested in finding out how we can improve your revenue and lower click distance with personalisation?

Get in touch today:

Email: sales@predictiveintent.com

Call – UK: +44 (0) 1202 83 20 30

Call – US: +1 (877) 2980-9380

Live chat: If we’re online, you’ll see a blue box on the bottom right corner of your screen – click and talk to us.

111 days until Christmas 2011!

James Doman — Mon, 05 Sep 2011 11:46:27 +0000

It’s only 111 days until Christmas 2011 – can your personalisation/recommendations technology provider help you make the most of the valuable behavioural data available?

We’re starting to work with our online retail customers to implement real time attribute shadowing and clickstream logic as part of their personalisation strategies. Whilst already both key algorithms available to use, implementing them in time for the holiday season will be useful for sites where a large percentage of sales are expected to be gift-buying, where a visitor (who may be known and profiled) is presented products and personalised content based on the interests and behaviours they are declaring in their session.

For example:

a known visitor arrives on site through the homepage

immediately, they can be presented with products that match their profile – taking into account past purchases, browses and declared interests

they then jump out of profile and start to look at buying gifts

our patent-pending Suggestion Choreography technology understands the visitor contexts and begins to adapt suggestion logic across the site

smart algorithms based on current browsing come into play

these can include a form of “people who browsed like you bought/looked at these products” algorithms, but filtered by in-session implicitly declared attributes (such as colours, sizes, etc)

This is the power of personalisation, as opposed to simple recommendation technology. By looking at the in-session journey, we differentiate from the “buyer” and the “consumer” of the purchased product. Adapting product suggestions, content and page banners to help meet the buyers expectations and end-goal whilst ensuring a really great gift for the consumer and a jump in AOV, conversion rates and revenue for the retailer.

Let us help you maximise the benefits of personalisation over the holiday season - get in touch today

We’re looking for a name for our new demo store!

James Doman — Thu, 01 Sep 2011 10:43:10 +0000

In the next few weeks we’ll be launching our new demo store and we want you to name it!

The store will “stock” streetwear-style clothing for both men and women, with brands including Roxy, Hurley, Etnes, Volcom, Quiksilver, Converse and DCs.

Inspiration

Enter

Tweet us @TweetIntent with the hashtag #NameOurDemo, and if your suggestion is chosen we’ll send you a £25 Amazon voucher!