Conversion – orders divided by visits – is one of the most common metrics in retail. Easy to calculate and simple to understand, it’s a measure of “efficiency,” like batting average in baseball, that tells a retailer how well their stores are, well, converting traffic into sales. And like batting average, while it is a decent indicator of one aspect of performance, it has several shortcomings. It should remain a key component of any retail dashboard, but modern retailers need to look beyond conversion to better understand and optimize their businesses.
The old standby
Conversion has been tracked by retailers since long before ecommerce rose to prominence. As a metric for traditional brick and mortar retail, it had the appeal of being easy to calculate: just count the number of people who came into the store and the number of sales made in a given time period and compute the ratio. Advanced analytics this was not. Conversion doesn’t require that you know who your visitors and purchasers are, what they bought, or why. But it does provide a good yardstick against which individual stores can compare their performance and the impact of various business decisions over time. And given its success as a key metric in brick and mortar retail, it’s not surprising that it became one of the go-to metrics in digital commerce.
Conversion: looking where the light is
In a way, retailers’ heavy reliance on conversion is like the old story of the drunk looking for his lost car keys under the streetlight. When asked why he’s looking there, and not further down the alley where he might just as well have lost them, he replies that he’s looking there because that’s where the light is. Conversion is one of the easiest metrics to track – the data is readily available from even the most rudimentary web analytics implementations – and everyone knows what it means. Compared to more sophisticated metrics, which might necessitate investment in new instrumentation, require more effort to calculate on a regular basis, and more education of the consumers of the data, conversion is “where the light is” when it comes to retail metrics.
On the surface, conversion helps tell a story of improvement or degradation in the business, whether we are comparing before-and-after, year-over-year, or test and control populations via an A/B test. And this is true, as far as it goes.
More visits are good, right?
But the thing that makes conversion easy to calculate is also the source of its biggest problem: because it is denominated by visits, an increase in visits, without a corresponding and proportionate increase in orders, results in a reduction in conversion. In fact, this is entirely the point of the metric – if the visit count rises but transactions don’t, then the business is performing worse. There is some truth to this, of course, but it creates some perverse dynamics. For example, from the conversion point of view, we would rather have a customer that ultimately makes one purchase visit our site (or physical store) just once rather than five times. But we know that the five-visit customer is likely more engaged, more likely to shop again in the future, and perhaps even less likely to return her items given how much consideration she put into the purchase. If we rely primarily on conversion, we risk making business decisions, both explicitly and implicitly, to drive the former behavior.
Conversion treats all visits equally and does nothing to categorize them. A visit that does not result in a purchase is scored as a failure, an “out” in the baseball analogy. However, from the customer’s point of view, making a purchase is not always the intent of all visits. We can place visits on a spectrum of intent: on one end, we have the committed shopper who comes specifically planning to buy something and will only be stopped by a broken experience or a lack of inventory. Near to the committed shopper on this spectrum is someone who is likely to transact if, for example, an item they want that is available at a discounted price. Continuing across the spectrum, we have customers who aren’t explicitly here to buy something but could be convinced. To this point on the spectrum, we’ve seen shoppers who will pull the trigger if the product, price, inventory, and experience are right. But then we get to the visitors who have no intent to purchase today. In some cases, this might be because they are doing initial research for a future purchase, but they may also be visiting to gather some specific information (for example, store hours or location) or perform some task other than purchase (maybe to get order status, or even apply for a summer job). Online, we even have traffic from bots – visits that will never convert and aren’t even human.
All of these types of non-transactional visits count against conversion. At best, they depress the conversion number consistently over time, but this consistency isn’t guaranteed. Bot traffic, in particular, can surge and subside unpredictably, and other non-converting traffic, such as job seekers visiting a careers page, might be seasonal. With effort, some of this traffic can be identified and excluded from the conversion calculation. This data cleansing should be done, but, in my experience, it rarely is. Even if we clean the data, however, that still leaves visits that are not intended to result in a purchase, but are still “good” for the business: the customer doing research on a specific product before pulling the trigger, exploring the new season’s offerings, learning about a new type of product, or preparing for a store visit by pre-shopping online.
Think of these like walks or sacrifice flies in baseball – even though they are not hits, they are good. In fact, baseball excludes walks and sacrifices from the batting average formula for this very reason.
Mobile changes the mix
It’s also important to recognize that the rise of mobile has fundamentally changed the mix of traffic that a typical retailer sees. What used to be a single focused online shopping session at home on a computer has been replaced by a series of “snacks” across devices – a browse here, an add to cart there, and, with luck, an eventual checkout. Phones bring a double-whammy to conversion – the smaller screen, lack of physical keyboard and mouse, and sometimes spotty connectivity can make transacting difficult, while the always available, addictive nature of the phone drives snacking behavior, spreading those fewer transactions across a greater number of visits. Phones tend to increase the denominator much more than the numerator in the conversion formula.
Meanwhile, when it is an option, people may still prefer to do their “serious” shopping on a traditional computer where they have the benefits of a bigger screen, keyboard, and easy navigation between multiple browser tabs. In this way, as more of the pre-shopping traffic has migrated to the phone, desktop conversion gets a boost, even if total transactions don’t increase. And to further muddy the waters, relative usage rates of mobile vs. desktop vary demographically, with younger and less affluent shoppers more likely to access a retailer by mobile only.
In the early days of ecommerce, a multi-channel shopper was recognized as being particularly valuable to a retailer. Likewise, a shopper today with a more fractured journey, spread across multiple visits and devices, may drive conversion down but is likely to be a more engaged customer. A common online shopping behavior is to precede a purchase with an intensive period of product research and comparison, increasingly spread across several separate visits. In fact, astute retailers encourage this behavior with well-timed triggered emails. This is a very good thing, but the conversion would tell you otherwise.
What to do?
So, given the drawbacks and limitations of conversion as a metric for ecommerce, what are you, a modern digital retailer, to do?
First, don’t eliminate conversion from your suite of key metrics! It is broadly understood and remains a crucial component of any retail dashboard. But make sure your organization sees it as just one of several indicators of the health of your business.
Just like baseball (and other sports) have gone through an analytics revolution, so must retail. To this end, consider implementing richer metrics, such as a variation of conversion that measures transactions over a period of time longer than the visit. This notion of “customer conversion” should span across not only visits, but also devices, and even channels if possible (loyalty programs or other mechanisms to encourage store check-in can help here). A single purchase by an individual might be preceded by multiple visits across a variety of devices, via the web, app, and physical store. Under traditional conversion, each of these visits (except the last) would drag conversion down, but in reality, each of these was a step towards the purchase and should “get credit” in the metric.
Also, explore metrics that measure the “success” or “failure” of a visit not just in terms of transactions, but also in terms of customer intent and progress: did the customer achieve what he or she intended to on this particular visit? Did you, the retailer, make progress in building your long-term relationship with him or her? From my perspective, this is the core idea of measuring and truly understanding engagement – not by simply counting the number of touches or minutes on site.
Finally, dig in. While metrics such as conversion are convenient and give a quick read on the health of the business, there is no substitute for a deep understanding of your underlying sales and traffic patterns. How has the shift to mobile changed the mix? Does the elimination of certain classes on non-transacting traffic in the calculation of your metrics fundamentally alter how you think about your customer? Understanding the dynamics that are driving the top-level indicators is enlightening and will help you look beyond traditional metrics to think about your business in new and innovative ways.