Illustration of various product recommendation design layouts

When it comes to the use of product recommendation solutions, the common tactic is to set it and forget it. But why? With real customer data we explain how a systematic, continuous approach to deploying recommendations exponentially increases new revenue. And, how eCommerce’s obsession with the ‘upsell’ might be a mistake.

The ‘set it and forget it’ mindset is most likely down to the nature of conventional recommendation solutions. Typically, they only provide a few simple recommendation options. These include widgets such as Frequently Bought Together, Recommended for You, Upsells, and Cross Sells. Once deployed, merchants typically focus their energy elsewhere.

With the arrival of session-based product recommendations, the number of possible product recommendation scenarios available massively expands.

This new method provides the ability to recommend relevant products based on shoppers’ real-time actions and context.

And there are almost infinite permutations of how shoppers behave, together with attributes of that behavior.

For example; the source of traffic, device used, or time and date are important contextual attributes. Combine that with visitor type (new, returning, etc.), prior browsing history, and products viewed. Now, you have the scope to generate several high-probability recommendation scenarios.

The following data story illustrates the importance of continuing to increase the number of product recommendation strategies on your Shopify storefront.

To summarize, product recommendations from HiConversion Recommend powered by Amazon Personalize:

  • Exponentially increase overall revenues despite linear traffic growth.
  • Disproportionally increase percentage of overall revenues attributable to product recommendations.
  • Directly correlate to increase in RPV lift and the increase in revenue.

Want better Shopify product discovery? Start small then quickly scale

We advise our customers to improve their product discovery experience incrementally. No matter what other solutions might promise, no single recommendation will fix your product discovery problems.

When it comes to starting out with HiConversion Recommend, the primary focus is making sure all the systems are working and the initial recommendations are producing positive results – be it engagement, conversion rate or revenue.

The best way of determining if your product recommendations are working is to compare the revenue per visit (RPV) of shoppers who engaged with product recommendations with the RPV of those who did not.

If your product recommendations are influencing visitors effectively, the RPV of those who engaged should be higher. This is because the product recommendation was effective, and more visitors were able to easier find products of interest.

In this example, our customer – a popular US beauty brand, running on Shopify Plus – has shown initial product recommendations were very relevant. As a result, they saw 101% higher RPV compared to sessions where visitors did not engage with recommended products.

Chart explaining increase in revenues from personalized product recommendations

(It is important to note that the initial deployment was paused for approximately ten days due to concern about use of a new product recommendation solution during the Black Friday and Cyber Monday period. You’ll see that reflected in the flatlining of the green line on the right of the chart.)

Recommendations ramp-up

After realizing how strongly product recommendations performed during the initial evaluation period, the decision was made to gradually increase the number of product recommendation strategies across the site.

The most interesting result of the ramp up? We measured a disproportional increase in percentage of overall revenues attributable to product recommendations.

Chart explaining exponential revenue impact of product recommendations

While the number of visitors engaging with recommended products grew in a linear way, the percentage of revenues attributed to those recommendations was growing exponentially.

Month 1 Month 2 Month 3 Month 4
% of revenue 2.40% 11.30% 25.10% 37.90%
% of visits 1.20% 4.90% 7.80% 9.60%

A similar dynamic was observed in the RPV lift, which is providing a direct correlation between increase in RPV lift and the increase in revenue – all attributed to product recommendations:

Chart illustrating revenue per visitor growth vs percentage of visitors

Month 1 Month 2 Month 3 Month 4
RPV lift 101% 149% 295% 475%
% of visits 1.20% 4.90% 7.80% 9.60%

The market perception of product recommendations’ purpose, is primarily to upsell and cross sell shoppers. Largely in an attempt to increase the average order value (AOV).

Our analysts began to question this perception. In turn leading to additional comparisons, revealing some fascinating new discoveries.

Namely, the RPV lift and consequently the increase in overall revenue that can be attributed to product recommendations. This lift is closely correlated to an increase in conversion rate (CR) but not to the average order value (AOV).

Chart illustrating difference between RPV, CR, and AOV

Month 1 Month 2 Month 3 Month 4
RPV lift 101% 149% 295% 475%
CR lift 52% 39% 265% 411%
AOV lift 31% 154% 8% 12%

The observed increase in AOV during the initial months can be more attributed to visitor behavior during the Holiday season, rather than the upsell and cross sell features of the product recommendations.

However, the continued growth in RPV and correlated conversion rate is proving the ability of session-based product recommendation algorithms to unlock previously untapped revenues from 90% of visitors who otherwise would not buy anything.

The merchant takeaway

The typical Shopify Plus site is a highly dynamic environment, where buying preferences of approximately 90% of visitors are unknown. Not to mention the makeup and preferences of those visitors, which are in constant flux.

To create the effective product discovery experience – plus unlock the revenue potential of those visitors – one can’t rely on a small number of product recommendation and merchandizing strategies.

Instead, merchants must be able to cast a wide net of product recommendation strategies. All driven by real-time algorithms, generating on-the-fly product recommendations based on visitors’ actions and buying context.

By adopting such an approach, merchants will consistently grow new revenue from existing web traffic.

Ben Virdee-Chapman

Author: Ben Virdee-Chapman

VP Marketing