The 4 Ps of Optimizing Product Recommendations

It’s 2020. Do you know what your cross-sell attachment rate is?

According to the e-tailing group, 59% of retailers don’t measure the performance of cross-sells and upsells on product detail pages.

It’s reasonable to assume that an even higher percentage don’t test or optimize the merchandising strategies that (should) influence how product recommendations are determined. If you’re paying hundreds or thousands of dollars a month for a product recommendation tool, you owe it to your business (and customers) to regularly tune it for relevance and effectiveness, or it could actually be lowering your conversion rate.

Cross-sells vs upsells vs product recommendations

Technically, not all product recommendations cross-sells. Most ecommerce sites display product recommendations – either alternatives for the product being viewed, or related products based on what’s trending, what’s new, or other merchandising logic.

A true upsell attempts to guide the customer towards the same product, but at a higher price point. This may be a newer or deluxe model, a larger size or quantity, or the item as part of a bundle of products.

Crutchfield’s “check out the next model” suggestion is a rare example of a true upsell, suggesting a higher-end model from the same line, with “fuller sound and deeper bass.”

A cross-sell is complementary item to the product being viewed. The intent is to surface related products a customer would be interested in buying in addition to the product being viewed to increase items per order and average order value.

The classic cross-sell is labeled “customers who bought X also bought” or “frequently bought together.”

“Shop this look” is another cross-sell option.

Product recommendations are often alternatives from the same category or brand. “You may also like” or “customers who viewed X also viewed” are common headers.

Whether your product page includes cross-sells, upsells or simply recommended products, it’s important you implement them based on a solid merchandising strategy.

4 Ps of optimizing product recommendations

The verbiage you use to label each product page merchandising zone is an opportunity to persuade. Rather than generic headings like “recommended,” “you may like” or “more like this,” be specific as to why you’re recommending products.

“Customers who like this item also like” indicates an affinity backed-up by social proof. This is more persuasive than “you might also like,” as it implies the recommendations are based on data.

Alternatively, “our picks for you” implies a degree of personalization. Consider testing this label against “customers who like” site-wide, and track click-through and attachment rates.

Be careful with “customers who viewed X ultimately bought.” This may help a buyer shorten their research process and make a faster decision based on social proof, but consider the caveat: it implies people-like-you preferred products other than what you’re looking at, and likewise complicate a buying decision!

The second caveat is unless you’re Amazon, you likely don’t have enough reliable data to correlate views of individual products with purchases of others with statistical significance. Conversion and sell-through rates are, on average, around 2-3% — you need an astronomical amount of data to make reliable associations between individual items viewed and ultimate purchases!

If you employ this merchandising tactic site-wide, you’ll ultimately see random suggestions based on recency rather than frequency.

Even a site as trafficked as Macy’s presents irrelevant suggestions. For the Rihanna perfume product page (cross-sells shown below), only one “ultimately purchased” item is from the same product category! The Davidoff item is a men’s fragrance, and the baby shorts on the far right couldn’t be less related to perfume.

If you do use the "customers who bought X ultimately bought" tactic, ensure you apply a merchandising rule that excludes products from other categories, so you only present realistic and relevant alternatives.

“Customers who bought X also bought” is also less sticky. It’s a cross-sell strategy, rather than a "buy-this-instead" approach, so products can span categories as long as they are related in some way. Exclude product categories that make no sense as a complementary product, like books with shoes.

You may also choose to label alternative recommendations “top sellers in [category].” This is a straightforward approach that doesn’t rely on an individual product’s relation to sales.

Population

Populating merchandising zones well is all about strategy.

Many retailers are content to buy popular (and powerful) recommendation engines like Bloomreach, Baynote, RichRelevance and Magnetic – and then “set and forget” them. Unless these powerful tools are tuned for relevance, recommendations can be random and low-converting. Don’t neglect merchandising rules like boost and bury, pin and include/exclude.

Include/exclude allows you to constrain the types of products that appear on product pages of a given category, brand or product type – and it’s your best tool to influence relevance.

For example, a very common mistake is to not exclude products from the opposite gender’s product category. Not that these are never bought or considered together, but you’re shooting for maximum relevance.

Spot the rogue Rogaine at the end? ??

Pro tip: Make sure you audit your experience to ensure your merchandising zones show products that add value to your customer’s visit and buying journey. Start with your top 20 trafficked product pages or highest revenue categories.

Auditing your merchandising throughout your site may be time consuming, but it’s the best way to ensure nearly-related (yet way off) products don’t sneak in. For example, GPS units may be relevant to motorcycles, but not tricycles!

Advanced merchandisers can layer in contextual rules into product recommendations, like past purchases, items previously viewed or most recent navigation path – the latter being the best indicator of present purchase intent.

If a visitor navigates to a tie product page via “Gifts Under $25,” it’s most useful to the visitor to see alternative gifts in the same price range. If he navigates to the tie page through “Ties,” show more ties from the category, or items frequently purchased with ties, regardless of price-point.

Ready for ninja-level merchandising context? If a product was filtered by size, populate only items with that size available.

Presentation

Where you place your product recommendations, like any design or conversion element, impacts click and attachment rates, as can relative size, associated information (like prices and star ratings) and the number of products shown.

Recommendations are commonly presented as sidebars (below) or horizontally. In the West, where we’re conditioned to read and scan content from left-to-right, the columnar approach is harder to process — not to mention that tiny thumbnail images make it even harder! The larger the product type, the more difficult it is to see all squishy.

Modcloth’s inline recommendations appear above the fold and above additional content like product details and reviews.

Inline placement works especially well when combined with a javascript or lightbox “mini-cart” that appears and disappears after an item is added to cart. When the mini-cart disappears, the customer remains on the product page with a set of “check these out next” suggestions at eye-level.

Another user-friendly design approach is embedding product recommendations with configurable attributes and quantities within a product page. This is especially useful when more than one add-on is commonly purchased with a given item.

Similarly, embedding a recommendation zone where add-ons can be “checked off” and added to cart from a list can increase attachment rates dramatically by removing open-new-page friction (especially on mobile devices).

Another tactic is to apply QuickLook to product recommendations, so the product can be viewed in more detail and added to cart seamlessly.

The number of suggestions you present is another conversion factor – we all know about the paradox of choice! A/B testing can reveal if “less is more” – just make sure you’re solving for click through and attachment rates for suggested products, not simply site-wide conversion rate.

An alternative and less cluttered way to present multiple merchandising zones is to use tabs or carousels with clear < and > markers.

How you present cross-sells and product recommendations doesn’t have to be uniform across your site. Home Depot treats high-ticket product pages differently from their accessories’ pages. For example, alongside best-sellers, you can “search” for “more products with these features.”

Some accessories and lower-ticket items are offered Amazon-bundle style.

Not every page has to show product recommendations, show the same amount, or show the same “type” of suggestions.

Post-cart

Don’t miss the opportunity to cross-sell after an item has been added to cart. Harry and David take a creative approach…

After adding an item to cart, an interstitial lightbox asks if you’d like to add “something special” to the order.

If you opt-in, you can add your add-on directly from a lightbox.

Walmart merchandises its mini-cart with “customers also bought these products.” This is a great way to keep the customer navigating your site with guided selling, rather than a generic “continue shopping” link.

Macy’s merchandises its mini-cart with best-selling items of the same category. Remember, introducing new options after someone has committed to adding an item to cart can be a risky strategy if it delays or derails the purchase decision.

In the example below, a better strategy would be to populate recommendations with cross-sells like pillows and sheets (in this case) instead of alternative comforter sets.

Crutchfield uses a full-page interstitial to present accessories to products like high-ticket electronics.

Customers can navigate sub-categories of compatible items and add to cart with a single click.

Pulling it all together

Whether you use an automated tool or select product recommendations manually, the first rule is never set it and forget it. Establish a merchandising strategy for each product category, and tune your merchandising rules accordingly.

If your tools permit, A/B test your merchandising hypotheses and measure click through and attachment rates for each experiment.

Finally, conduct a thorough audit of your product page merchandising. Pick a handful of top products from every category, and check out what appears in your product recommendation slots. Do these items make sense? Do they add value? Do they distract from the purchase decision by introducing FUD? If you use manual associations, do they need to be updated?

Poor product recommendations happen to the best of sites, just don’t let them happen to yours!

Originally published on Ecommerce Illustrated as the Ecommerce Guide to Merchandising Product Pages

??Check out Ecommerce Illustrated for in-depth ecommerce optimization guides from Home Page ?? through Checkout ??