The online equivalent of “Do you want fries with that?” is fast becoming “Customers who bought this, also bought ………”. But the science around this area, behavioural based merchandising, is a lot deeper than just adding a simple product recommendation based on what someone else has bought. How about recommendations like “Other customers also viewed” or “Other customers went on to purchase” – the whole idea is to make the most relevant offer to the right prospect at the right time.

The first thing is that you need to know the prospect. I mean really know them – the better you know them, the better chance you have of matching an offer to meet their needs. Marketing 101, yes? If all you know about them is that they have put shoes into the shopping cart, then the best you can probably do is recommend other shoes as an upsell or maybe some socks as a cross sell… But what if you knew that before they put shoes in the cart, they also looked at an expensive gold pendant and the last time they hit your site, they were looking at gold earrings and that they came to your site by searching for gold necklaces? Armed with the deeper knowledge of the customer you might conclude that the shoes were a short term impulse buy and that they are really in the market to spend a whole lot more money on something in gold – so the more appropriate recommendation would probably be alternate pendants or maybe earrings (or both).

Another consideration is where to put the recommendation. What a total waste of time and money it is to see sites going to the trouble of putting up recommendations and then positioning them below the fold, so the only way someone sees the recommendation is if they happen to scroll down the page. A recommendation needs to be in your face if it is to have the desired impact.

Part of the positioning decisions are also around the page you use. Different types of recommendations are known to work better on different pages. Here is a table listing some common page types, the preferred recommendation option and an estimate of the results you might expect with a strong online behavioural targeting engine:

Page Recommendations Percentage increase in sales
Home Top Sellers 0.5%
Category Top Sellers 1.3%
Product Listing Top Sellers 9.0%
Individual Product Others Also Viewed 6.0%
Individual Product Others Also Bought 2.0%
Cart Others Also Bought 1.0%
Cart Others Also Viewed 0.2%
Order Confirmation Others Also Bought 0.3%
Order Confirmation Email Others Also Bought 0.1%

And then there are follow up recommendations. Using email to recommend a cross sell or to retarget customers who displayed a predetermined activity on your site. A recent study concluded that targeted emails with the right product recommendation had a conversion rate of 8% compared to 2-3% for broadcast emails… or using the recommendations across channels, say in the call centre or at POS… but that is a whole blog session in its own.

The bottom line is that bringing a scientific approach to the automation of online product recommendations is becoming a complex and specialist process, with smart businesses taking advantage of the seriously strong returns available from getting it right.

“This is my personal blog.  The views expressed here are my own and do not represent those of my employer, Coremetrics.”