Most ecommerce sites give some sort of cross-selling or up-selling recommendations to their site visitor. The most common places to see these recommendations are on the product page, a time when the buyer is closest to adding an item to their shopping cart.

MyBuys.com and RichRelevance.com are two examples of companies that have created a unique recommendation engine that can be implemented with any ecommerce site. MyBuys.com conducted a study of 1,345 online customers and 24 merchants in an effort to understand behaviors, trends, etc. Most consumers expected merchants to offer recommendations on the site, but only 2% felt that the recommendations had any relevance to their likes, dislikes, or search.

MyBuys.com feels that personalized recommendations based on customer shopping behaviors are a lot more effective. In an effort to gain a better understanding about their personalization software and really decide whether or not this is intrusive and taking advantage of the buyer, I conducted an interview with Jeremy Pollock, the Director of Product marketing at MyBuys.com:

Briefly explain cross-selling and up-selling to those who are unfamiliar to the concept and how it works online.

MyBuys seeks to understand the user’s preferences and builds recommendations through monitoring user behaviors online: browsing, searches, purchases, etc. Combining that information with merchant info such as merchandise catalog, sales, and inventory changes over time the recommendation engine outputs a number of algorithms that seek the best recommendations for the consumer as they are interacting.

How do you determine the specific inclinations that a customer may have before making recommendations?

The software is quite sophisticated in that it outputs a number of algorithms creating a portfolio for a single buyer over time. Each algorithm is prescribed a specific weight depending on the price important to the consumers, brand, etc. The portfolio of algorithms looks at the consumer’s buying and browsing trends, pricing inventory, etc. and the software then fine tunes what works best for each consumer.

What are your thoughts of cross-selling within the shopping cart?

On the cart page in general we have seen good performance. We would not like to disrupt the check-out process so we believe we have to test and see the results. Generally we have found that recommendations made inside the cart improves basket size, brand stickiness, and conversion rate – and do not lead to higher abandonment rates contrary to general belief.

Where is the best area through the buying process to promote cross-selling suggestions?

On the product page:

1. By recommending alternatives to currently viewed products up-sells.
2. Product detail page alternatives – complimentary items combined in one area on the page, or when they are broken out in different places throughout the page which is more successful

What type of results do ecommerce companies experience with cross sells?

We have 3 high level metrics:

1. Increasing avg. order value 45%
2. Increasing conversion rate 90% – repeat and new customers –
3. Increasing revenue from 5 – 20% average 10%

What negative experiences have you had with the implementation of cross-sells on a site?

We are forced to experiment with different placements of the recommendations throughout a site, some more successful than others – for example: above the fold placement or below the fold.

Another issue that we constantly face is ecommerce companies that set business rules for a product’s recommendations: For example some of our golf site clients have relationships with golf manufacturers such as Nike, Callaway, Taylormade, etc. So within their product pages they only want recommendations from those specific brands. So Nike drivers only recommend Nike brand family which actually can impact effectiveness of cross-sells, but the business relationship is a lot more important in this case.

The challenge that we always face is more of a tuning the system from our client’s perspective and continue to increase revenue to the client yet adhering to their business rules. Not having that business rule in place would have generated more revenue but a better strategic relationship with brand, again, is more important.

How intrusive do you feel the process of personalized cross selling is to buyers?

Looking at all our data through the e-mail and web channel – generally people are okay with this personalized service. Our survey of an e-tail group from consumers and merchants on personalized product recommendations proved that is not a negative experience. It’s not scary compared to behavioral monitoring websites where you visit a site and upon browsing different sites suddenly a banner ad pops up relating to the previous search.

On the contrary, consumers like the interaction with the site and they expect the e-tailer to merchandise to them. Users have embraced personalization.

Are there any new features on different browsers in terms of blocking cookies that impacts the cross sells?

We are always on top of regulatory browser activity related to privacy and cookies. At this point of time, there are modes that allow people to turn off and browse anonymously such as chrome and IE 8, which is a challenge to any vendors in this space, and we will continue to monitor this. We will also continue to look at data related whether users are inclined to turning off cookies or not – we think that most consumers won’t be entering into this anonymity on the web because of the inconveniences this may pose in terms of re-entering info, cookie driven pop-ups they want to see, etc.

What makes your company unique compared to other competitors?

The true differentiators:

1. We are multi-channeled offering this service across different mediums and channels: the web, e-mail, call centers, and RSS. Yet we deliver recommendations differently to the different types of consumers that make sense to each channel
2. Multiple algorithm approach – not tying it to a single algorithm that makes sense to one vertical and client but having 20 different algorithms that makes sense to various customers it allows to optimize much more effectively different data, community data, etc. to shift focus based on the results.
3. Experimentation framework – it’s not about A/B testing against ecommerce and vendor recommendations but being able to run experiments within their own environment – what titles work well improve the conversion rate – and more experimentation allows us to maximize the results and learn more about consumers as a set of best practices.

Thank you to the MyBuys team for this exclusive interview. Next blog I will dig deeper into personalized recommendations vs. general recommendations. Which do you prefer?