A century ago, John Wanamaker said, "Half the money I spend on advertising is wasted; the trouble is I don't know which half." Today, online marketers continue to grapple with the same question in their analysis of metrics.
It might seem like a simple answer because in the online world you can track clicks. The problem is that click-counting and click-based analyses are fundamentally flawed. It's not just that clicks don't tell the whole story, they tell the wrong story -- especially when used alone.
Many marketers, thanks to their web analytics depolyments, only attribute site activity to click-based activities, e.g., clicking on an ad, which is a limited way to gauge response.
Industry click-through-rates (CTRs) are notoriously minuscule, hovering at less than 0.1 percent. The vast majority of people who see online ads do not click. Furthermore, those who do click do so disproportionately; some 85 percent of clicks come from only eight percent of people. Numerous industry studies have addressed this issue.
Yet, low CTRs don't mean ads aren't working -- quite the contrary. Customers are responding to the ads, often with purchases not long after seeing the ad, and often without clicking.
In a recent test, IMVU, an avatar-based social network and virtual world where people can buy virtual goods for real money, wanted to find out whether non-paying IMVU users (who already received email marketing and were exposed to ads in the virtual world) would be more likely to become paying customers when exposed to IMVU online advertising in the real world.
In a test control scenario, customers who saw IMVU ads were 10 percent more likely to become paying customers, regardless of whether they clicked on an ad or not. Compared to the control group, the 10 percent increase was incremental, above and beyond the sales boost from existing marketing efforts. The control group had an equal opportunity to be exposed to all other marketing activity. The ony difference between the the groups was the actual ad exposure. The test group saw an IMVU ad whereas the control group saw an unrelaed ad.
In the same way, IMVU tested whether paying customers would actually spend more money if they saw online ads (again, in the real world) encouraging them to do so. IMVU members who saw promotional display ads to purchase virtual goods, on average, spent more than double an IMVU member who was exposed to a control ad, regardless of click activity. Again, this was an incremental lift above and beyond promotional activity via email and the virtual world. For companies like IMVU, being able to sell virtual goods is like printing money.
In a separate campaign, we looked at an e-commerce company that relied heavily on a web analytics package that utilized post-click attribution (activity and revenue at the site driven by an ad click). The advertiser wanted to optimize based on post-click activity only. The client did not track post-view revenue (sales at the site driven by online ads without clicks), so it had not optimized for it.
Two scenarios were reviewed: optimizing based on post-click attribution, and optimizing based on post-view attribution. Incremental revenue from post-view optimization was 10 times higher than optimization from a click-through perspective. When analyzing revenue from a post-click perspective, the best ad unit appeared to be an ad we will, for these purposes, call creative ad "A," and the worst performer, was creative ad "C". But when analyzing from a post-view perspective, the results flipped. "C" was the best and "A" was the worst, leading to very different optimization scenarios.
Playing devil's advocate, one might argue that a post-view analysis attributes too much undue credit to all online advertising. The argument would be that a prospect exposed to the online ad was likely to purchase anyway and the impression probably didn't influence their decision. Yet, time and time again, we find the exact opposite. We conduct analysis of the window of time between an ad impression and a purchase. This data reveals the immediate increase that happens very shortly after a consumer sees the ad, demonstrating the post-view impact of the campaign. In the examples below, half of the conversions occured withing 6 hours of showing an impression, and 70 percent occured within 24 hours of showing an impression. If there was no such thing as a view-through impact, one might expect that the random distributon of converions over time to follow a more linear pattern rather than curve.
The bottom line is that every campaign is different. They all should be optimized based on as much data as possible. Don't rely on click-based analysis alone. You'd be leaving money on the table.