Crafting Ecommerce Cross-Sells That ConvertReading Time: 12 minutes
Anyone in the world of sales can tell you that cross-selling is important: Why sell one thing, when you can sell two or more? The in-store sales space has been rife with this for ages: staff suggesting monitors or keyboards to go with your new computer, bank employees suggesting a new credit card to go with your checking account, and so on.
Naturally, online stores want in on this action too and quickly toss together recommended products sections for every product page. But online spaces aren’t like brick-and-mortar spaces. You can trust a good employee to determine what offers might be relevant to a customer when they’re dealing with them face to face, even without dedicated training. But you can’t just trust your Shopify theme’s default cross-sell module to make sense of your products and your customer’s goals. All too often, that’s exactly what store owners do.
The key factor in creating an effective cross-sell is relevance. If I’m buying a new gaming controller for my Xbox, recommending a PlayStation game to me is entirely unhelpful and simply won’t sell. I’m sure we’ve all had that experience when buying an item in a store: the sales staff keeps trying to push you to whatever item they are probably incentivized to sell, and it’s clear they aren’t listening to anything you’re saying about what you want. Naturally, you’re not going to buy what they’re peddling.
However, a cross-sell of poor relevance doesn’t just not sell—it can actively harm your company’s reputation and reduce your customers’ trust. If I am regularly being served with cross-sell ads (let’s face it, a cross-sell on your store is an ad) and they aren’t relevant to me, I’ll see that the store doesn’t really care about my needs. Over time, I’ll learn to just ignore that section entirely. I just won’t trust it.
On the flip side, when cross-sells are highly relevant to my needs by serving up the right products at the right time, my feelings about my experience shoot up and the money in my wallet escapes me. Over time, I’ll view that store as the place to go to solve my problems, because they clearly care about me and their products.
Now, chances are, you aren’t a Google or a Facebook. You don’t have 3,000 points of data on every customer and expansive data centers crunching the real-time thoughts of a billion users. But you don’t need to be to have cross-sells that get it right most of the time.
Before we get into how to make highly effective cross-sells, let’s look at the most common examples of bad cross-sells and what makes them bad.
You may also like . . .
This is probably the most common framing for low-effort cross-sells. Following that label is what appears to be a random smattering of items. This one is so common, especially among Shopify stores, that you probably have this or something very similar on your store right now.
The main issue here is simple: Why might I like these products? There’s no helpful information here to help me understand why Sephora thinks these are great products for me. Now, I’m not in Sephora’s target audience, so maybe these are actually the perfect recommendations, but there is little way for a user to really know and trust these when there doesn’t seem to be any logic or meaningful connection between these products and what I’m currently browsing.
Most of these “you may also like” suggestions—regardless of site—are just randomly selected items, featured products for business reasons (encouraging users toward higher-margin items), or selected by “dumb” logic (like best-selling products or items in the same collection). However, the key takeaway here is that your cross-sells need to provide information to validate the relevance of the recommendations. Even if the items end up not being relevant to an individual customer, that customer can recognize how the item could have been relevant and why it was recommended.
Unfortunately, a lot of cross-sell apps on Shopify also use this as the default title for their recommendation widget, even when the recommendations can be designed to be much much smarter. This devalues all of the work you might put into improving the recommendations. So, be sure to change this title on the widget to properly reflect your recommendation logic (as detailed later in this post). If the app doesn’t let you change the title, poke the developers or make the switch to an app that will.
Customers also bought . . .
Often, this is a store’s first step toward having slightly smarter logic for cross-sells. It’s certainly one of the easiest to set up: just look at past orders, have a robot see what items are bought with the current item, and show those items. Often this doesn’t present much better results than the “you may also like” situation.
There are a few issues that hurt this method:
- You need a lot of volume for this to have any chance at all. Tens of thousands of orders would be needed to get decent results for a store with a significant number of products.
- How many times an item needs to be bought together to be relevant can be arbitrary and unclear. Sure, maybe 2% of customers who buy X item also buy Y item, but the other 98% have zero interest in Y item. Is recommending Y item to everyone really a good option?
The items that customers buy together is useful information to help to determine cross-sell relevance, but it’s only a small factor.
Designing a Relevant Cross-Sell
Now you might be saying to yourself, “What the hell? Those are the only cross-sells I have in my store! What do you mean they suck? How am I supposed to fix this?!” but have no fear, there are solutions available to you!
These are the primary types of logic for selecting cross-sell recommendations:
- Similar: An item that has many characteristics matching those of the primary item
- Alternative: An item that serves the same purpose as the primary item
- Supplementary: An item that goes well with, or might be needed with, the primary item
Having at least one of these cross-sell sections is pretty much required to have a high-performing store. Depending on your store and products, having two or even all three of these could be advisable.
Additionally, using different types of cross-sells on different products can be a smart way to make the recommendations even more relevant, as I’ll explain below.
It’s also important to properly label the groups of suggested items based on the type of logic. This helps the viewer understand how these products might interest them.
Next, I’ll go over some examples of these cross-sell types, how to determine which products benefit from each type, and some other details to keep in mind.
Similar items share some of the same characteristics and are fairly easy to recommend. Simply find other products that have a lot of the same characteristics. Essentially, similar items need to be nearly interchangeable in the user’s life. This is a great option to use for simpler, visually based products.
For example, take a rose gold necklace—a flower-shaped pendant on a simple chain. Similar item recommendations include products of the same type (necklaces), same metal (rose gold), same style (chain with pendant), and similarly themed design (floral). Not every user who clicks on this necklace will have all of these details in mind, but it’s a fairly safe bet that the more these characteristics match, the more likely it is that the customer will find the recommended products relevant.
If your filters are set up properly, then filterable options can serve as good basis for determining the important factors in a comparison.
However, not all characteristics are created equal. Not all items that share three tags with the initial product will be equally relevant—not all tags have the same weight, so not all options based on those tags are relevant. Going back to the rose gold necklace: another rose gold necklace with a different style chain and a different pendant design may be more relevant than that exact same design and style of chain and pendant offered in a metal other than gold.
This comes back to understanding your products and what motivates your customers.
Many simple recommendation apps can handle this fairly easily, and this is often the extent of their default capabilities. Other apps will allow custom recommendations per product, which can be used for high-traffic and best selling items or smaller stores. Custom recommendations on all products on large stores may be a huge undertaking.
Alternative items serve the same purpose as the current product being viewed but may accomplish this in a different way. This will often be symptom based (i.e., what problem does the product solve?) and thus work well for more technically based items.
For example, if your store sells an air conditioner, an alternative item could just be a fan or vice versa. For most customers, they solve the same problem (feeling hot) but accomplish it in different ways. This is great for users who have problems that don’t necessarily have the same constraints. It provides users with alternatives if the item they’re currently viewing doesn’t satisfy a constraint they have (not being able to install the ducts for an air conditioner, for example).
Another example is that of Bluetooth and wired speakers, battery powered or not. These all serve the purpose of playing music but do so in ways that may be acceptable to one user but not to another, often with various trade-offs, like price.
You may notice that there can be a lot of overlap between products recommended as similar and those recommended as alternatives. This is absolutely fine. These are just different logical approaches to creating and communicating the list. They don’t need to result in completely different product recommendations.
Alternative Item recommendations will be a bit harder to get using default options in a recommendation app. They will generally need some manual input of recommendations, either by tag or specific to the app in question. Alternatively, you can contract with a developer to build out the recommendation logic for your store.
The difficulty with the alternative logic is that most recommendation logics are really just geared to stick to items that are in the same collection and share lots of tags, but alternatives to the current product could be spread across other collections and have very few tags in common with the result that each recommended item differs from the current item in a different way. This could result in four recommendations, each from a different collection.
Supplementary items, where applicable, are the most crucial to get right! They convert extremely well, with the Baymard Institute finding that nearly 40% of users buying an item also view the product pages of supplementary item recommendations. Furthermore, the added “value” of having these recommendations drastically increases the quality of the user’s experience.
As mentioned, supplementary items are those items that go with the product currently being viewed. That may be simply something that goes well with the primary item (like a bracelet that goes well with a necklace in a set) or items that could be necessary (like a hardware kit to go with curtains).
Without this appropriately placed recommendation, users often have to jump back into the main navigation, navigate to an entirely different section of the site, and then start activating filters to find the appropriate product. The alternative is even worse: they may forget about these extra items entirely.
For items with very strict compatibility needs, the difficulty of finding a supplementary item—and thus the benefit of a good recommendation—is far greater. While a purse that goes well with a dress may be a subjective supplementary option, a phone case is fairly specific to the type of phone you have, and the hardware for installing a specific set of curtains is even more specific. It could be very difficult for users to clearly and accurately drill down to find the appropriate supplementary item on their own even if they know what they are looking for.
The bad news about supplementary Items is that there is basically no way to have this done automatically. It will need to be set up almost entirely manually, either by highly specific tagging or by manually recommending the items in your preferred cross-sell app. But the payoff is worth it!
While having recommendations that are truly relevant is the bulk of the battle when it comes to crafting cross-sells that convert, the details of how these recommendations look and are communicated can make a significant difference.
Add to cart
An excruciatingly large number of cross-sell apps in the Shopify App Store have their widgets styled to include an “Add to Cart” button by default, sometimes not even giving the option to remove it.
This is, in the vast majority of cases, a very bad tactic. You, as a store owner, have put a lot of time and effort into good photos and compelling descriptions. Why would you put in that time if a tiny thumbnail and title is all a user needs to decide on your product? The answer is that a tiny thumbnail isn’t enough—Your customers need to actually review the recommended product just as they would any other product. Does it fit their needs? Do they even like it? Expecting a user to make a purchase decision based on the little information the thumbnail and title offer is nonsensical.
Instead, clearly direct users to the relevant product page and don’t give users an alternate path that skips the product page. Even if the user can technically click on the product image to view the product page, the focus will be on the provided button.
A more appropriate “Call to Action” (CTA) button is one that reads “View” and leads to the product page.
The one exception is for recommendations that are highly relevant, supplementary, and simple. These need little consideration. This essentially means items that are almost required for the customer to use the main item and are very clear in their connection.
The hardware kit for installing a curtain is an example. A curtain needs hardware to hang from, and the hardware itself doesn’t need much consideration aside from any metal finish. Many users can easily make the decision whether or not to add the item based on just what’s available in the cross-sell.
Number of recommendations
This is another detail that most cross-sell apps fail to address. They force the same number of recommendations onto every product page, regardless of the quality of those recommendations. This is only hurting the credibility of your cross-sells.
Instead, your cross-sell module should display only as many items as are relevant (up to a certain limit, of course). If you have two items that are highly relevant and then the next two aren’t even remotely related, don’t show those last two. Nobody is going to click on them anyway, and they increase the noise on the page.
This may require a custom cross-sell section, as I have yet to find any apps that support this kind of thing. If you’re a one-person operation and you don’t have any developer skills, you might need to consider this out of reach. Just add it to your growing list of things for which you may contract a developer.
While cross-sells typically just drop each recommendation with an image, title, and price, a high-converting cross-sell will also include a note with relevant details for each recommendation. If you’re eagle-eyed, you’ll have seen these in the example images throughout this article.
Essentially, when appropriate, provide that extra help to connect how the product being shown is relevant to the product the customer is currently viewing. Explain why this item is similar to the current item and what makes that item an alternative to the current item.
This is a major stepping-stone between the static impersonal recommendations of most online stores to the very involved and knowledgeable help a customer can get when visiting a brick-and-mortar store.
Get It Done
This has been quite a long article, and I’m sure your head is rumbling with ideas, so I’m going to help you with as simple a list as I can manage covering the steps you should take now to bring your cross-sells up to snuff:
- Evaluate your current products:
- Are your items simple or technical?
- Do your products have clear connections?
- Evaluate your current cross-sells:
- Does your cross-sell section communicate why it recommends what it does?
- Are your top products’ cross-sell sections showing relevant items?
- What level of control does your current cross-sell app offer?
- Do users use your cross-sell links?
- Do they convert?
- Build out new cross-sells:
- Do you need to use similar, alternative, and/or supplementary logic for cross-sells?
- Map out connections between products.
- Can you use your current recommendation app or do you need a new one to accomplish your goals?
Baymard Institute. (2019). Provide a Cross-Sell Section That Only Contains “Supplementary Products.” https://baymard.com/premium/guidelines/811