As the marketing funnel continues to get more complex, last click attribution is becoming a less reliable metric. I talked to Xevio co-founder and CEO Nadim Kuttab to get his thoughts on the value of softer metrics when measuring performance, as well as their limitations, and the future of performance measurement as a whole. Be sure to also check out our previous two installments, in which Nadim looks at driving direct response sales on the open web, and building high-converting e-commerce funnels with native content.
Why is measuring engagement and intent crucial for e-commerce performance, even when the primary focus is direct sales?
Because in the e-commerce space, it’s always important to look at several events in the buyer journey. Most affiliates start with last click tracking, because that’s the easiest way of tracking what the user buys in one session. It’s the dream of every marketer: There’s no better case for a marketer to say they’re providing value than if they’re spending a dollar, then making $2 back within 10 minutes. It’s a great case! The truth is, though, that it’s becoming harder to rely on last click in marketing.
People like to take their time in buying — they have options, they want to research it or look at TikTok reviews. There are so many things people do nowadays before buying. I know that when I had more time when I was younger, if I bought a laptop, that was a huge investment, so I’d spend hours reading about processors and graphics cards and battery life, going down these rabbit holes to inform myself before making a purchase.
Most people are no different — maybe they don’t spend 20 hours on choosing what shoes to buy, but they’ll at least inform themselves on what alternatives there are, whether there’s a better priced alternative from a competing brand elsewhere, etc. Maybe you’ll lose them in that process, or maybe they might come back later and buy via another channel — either way, if you’re only looking at last click, another traffic source will be taking the credit for that sale.
First click is also not an ideal way of looking at purchases, though. Both first click and last click have their place, but I recommend a mix of both, which is essentially distribution modeling. Using a mix of first and last click will give you a good idea of where the users are getting the first contact with your brand, and where they’re buying.
Often, though, it’s going to be a mix of a lot of different things. We’ve seen with our native ads on Taboola that if you do it correctly, there should be about 2-4x as many first click conversions as last click, because Taboola is an educational platform, it’s content-based — people are coming, they’re reading, they’re jumping away, buying it elsewhere. If you measure this correctly, it gives Taboola a much better standing for budgets and for scale, because you’re not just looking at last click, which will only take into account 20-30% of all conversions Taboola actually helps generate — you’re looking at the whole pie. That’s why it’s important to use one attribution tool across your whole business.
How can metrics like “Time on Site” and “Session Depth” (as offered by Taboola’s performance platform, Realize) provide valuable insights into e-commerce buyer intent and campaign effectiveness?
The issue that a lot of brands have is that there are a lot of different elements to test on Taboola. You have thousands of sites, you have dozens of ad formats, you have potentially dozens of landing pages and product pages behind those ads. So, if you really wanted to collect enough data to make decisions about everything on a purchase level, you would need to spend thousands a day. Some brands don’t want to do that, so they need to make educated decisions quickly, without spending a ton of money, which is where these top of funnel conversion events come in.
We use Voluum as a tracker, which tracks landing page CTR, and we always push that data back to Realize. We always have add-to-cart in our e-comm campaigns with Taboola, too. Then, if we collect enough data for one type of event, that’s the event that we use. Obviously, the goal is to get enough purchases to optimize based exclusively on purchase data, but in lieu of that data, you need to look at pre-purchase events, which you then optimize until you have a campaign that’s stable enough to generate those purchases. Think of it more as a means to an end, like stepping stones towards having a campaign that’s scaling and spending enough to optimize exclusively on purchases.
Realize's advanced AI capabilities enable precise targeting, engagement optimization, and budget simulation, maximizing advertiser ROI.
How do you identify and track “micro-conversions” or engagement signals (like product views, add-to-carts, or wishlist additions) that indicate an e-commerce user is progressing through the purchase funnel?
The same way we track every other event — either with server to server, or we just fire that event back to Taboola. (If you’re using a tracker in between, that’s the same thing, just with an additional step.) At the end of the day, you want the data in Taboola so that you can look at it and understand, okay, this is an add-to-cart, this is a purchase, this is what I should be working towards.
We’re moving away from micromanagement of media buyers in campaigns. When I started media buying, every hour I’d go in and, like, bid up 5% or 3% on a site. It was a massive amount of work! It gave us a huge edge because we were doing that and others weren’t, but the future of media buying on Taboola is very much automated. The biggest lever will be the creative.
What strategies can e-commerce advertisers employ to leverage engagement data (such as creating engaged audience groups for retargeting) to improve overall campaign ROI?
You need to put the Taboola pixel on your page — ideally on the checkout page or the product page — to retarget. Period. There’s no exception for this, it’s just a no-brainer. It’s very easy, and it’s free money! It can be a very, very easy ad where you just say, “Hey, here’s a 25% discount on the product. Buy it here.” Our recommendation is to pixel everything, because why not?
Taboola is also rolling out smarter features for building audiences to target, based on first party publisher data. That can be used to create pockets of audiences to target. Again, without data, Taboola can do nothing, so just feed it everything! We recommend feeding as much data as possible into the campaigns, because nowadays, the campaigns are very much on autopilot when it comes to media buying.
So it’s essentially a classic case of, the more you put into it, the more you get out of it?
Exactly.
How do you balance optimizing for immediate direct sales with building long-term engagement and brand affinity for e-commerce on the open web?
Xevio is a performance agency, so we’re measured on a dollar in, a dollar out. There are other agencies that have other targets, but that’s what most of our clients will look at and say, “We gave you a dollar, you made $3. Well done.” Or, indeed, “We gave you a dollar, you made $1.20, not so good — we need to improve that.” Obviously we consider lifetime value, which we try to increase by creating bundles or upsells or downsells that people can buy while they’re in the purchasing process. But, the true long-term brand value is not something that you and I can properly assess by any measure. I don’t think it’s our job as performance marketers to measure the long-term impact of brand equity — I think it falls squarely on the brand owners. As a performance marketer, if it’s not our brand, the best we can do is educate the brand on the impact that we can have.
What are the limitations of relying solely on engagement metrics for e-commerce performance measurement, and how do you combine those metrics with direct sales data?
Engagement metrics are a means to an end, so relying on them isn’t smart. It’s more of a necessity — I don’t think anybody with enough purchase or sales data would rely solely on soft metrics. Engagement metrics are really good at the beginning, when you have to get a campaign moving. If you don’t want to use engagement metrics, you need to spend more money to buy data.
Think of it like getting a plane into the air: You need to have engagement metrics to be able to see some data without burning money. Now, you could also just strap a rocket to it and it’ll go up, but it’ll be incredibly inefficient — you’ll buy a lot of data very quickly and therefore be able to make a lot of decisions, and there are brands that do that: They just set, like, a $5k daily budget, spend it for a month and say, “Hey, we’ve learned something!” But, most of those brands are enterprise or very large growth brands. Engagement metrics are needed by, I’m guessing, the majority of the people that are going to read this, because they don’t want to burn $100 grand on learning.
Really, you have a choice: You can be inefficient with your money, but efficient with time, or you can be efficient with your money, but it’ll just take longer. You can save time or money, you can’t do both. Essentially, engagement metrics let us make decisions in a shorter period of time, with less budget. Are they going to be 100% accurate? Hell, no. But, at least they give us an indicator of what works, and if we see a trend between engagement metrics and sales, we can use that ratio to guesstimate what sales data we’ll see. Will it be accurate? Again, no! But, it’ll give us an idea of what to optimize towards.
What’s your vision for the future of e-commerce performance measurement on the open web, moving beyond simplistic last click models to a more holistic view of customer value?
I think this is a great question for everybody that does online marketing, because every single advertiser on every channel will be forced to use a more holistic method, since people’s purchasing behavior is changing. It’s taking longer to get a sale, people are buying over more channels, people are taking their time when buying. Even with spontaneous purchases, they might look at it on their mobile phone, but buy it on their tablet or desktop. How do you track that? It’s extremely difficult.
There are ways to do it, but it requires unified attribution methods that are applied by brands universally, and that’s what we’re seeing. The very big enterprise brands are building these models themselves — good for them! It might cost a ton of money, but it’s good for them because they can use that to make better decisions on their $100 million or $200 million budgets. It’s absolutely worth it if you’re that big.
For everybody else, we’ll just use off-the-shelf solutions that’ll cost a couple hundred or a couple thousand bucks a month, and that will help us make those informed decisions as best as possible. I don’t think there’s a perfect way to do this, but if you’re applying one method to everything, you won’t be over-reporting revenue or attribution. I think that’s the future. We’re moving into an area where I think, in five years from now, there won’t be a successful e-commerce brand without a proper attribution model. First and last click do not cut it, even now!
