What’s the real impact of advertising? The answer to this question was ever-elusive before we got the internet and all the not-so-private tracking wonders of today. But, you see, even though we measure everything nowadays – it still doesn’t add up sometimes. It’s easy to lose sight of what’s important among the clutter of numbers, and statistics that most interpret wrongly. Unfortunately.
So, what’s the impact of advertising?
We call it the incremental lift. In other words, the effect that wouldn’t be there if you didn’t run ads. The number of sales you wouldn’t get otherwise. The number of actions you wouldn’t inspire the other way. The number of people that wouldn’t even know about you if not for your ads.
Officially – the lift that advertising provides above native demand.
Unfortunately, proprietary ad tech tracking, in the form as it is now, won’t help you there. Not because it’s broken, but because it simply wasn’t designed to measure incremental lift. It was designed to measure the total effect, attribution. Facebook Pixel measures the total effectiveness of your campaigns, Google’s too, and they attribute it towards the specific ad creatives. Attribution measures the total number of actions people did in a period of time after seeing or clicking on your ads.
But what if they would’ve done it anyway, on their own? And, on the other hand, what if they did interact with your brand, but outside your attributed time period?
Beware the so-called “selection effect”
Imagine people waiting in line to order your pizza takeaway. Would you advertise your special discount offer to them? Probably not. You want to attract new customers with it. However, this happens in online advertising if you base your decisions on attribution only.
Over-optimizing your campaigns towards maximum ROI by basing your decisions solely on attribution creates a high chance that you’re giving away discount offers to people already waiting in line to buy from you. That’s how machine learning works, it seeks the easiest route to your desired results.
By helping people who already made up their minds to buy something in your store, you will achieve a great ROI, that’s for sure. And, they won’t change their minds at the last second and they will stay loyal. But, you will miss out on countless opportunities that lie hidden among the broader audiences with seemingly lower ROI. The keyword is “seemingly” since you cannot know for sure if you’re not measuring incremental lift. And this is how you get sucked into the selection effect. Chasing easiest opportunities to advertise to while narrowing down your users’ pool, instead of broadening your scope. You end up losing money on advertising in the long run.
To put it simply, attribution is not enough for decision-making. You need more. You need incrementality.
Without measuring incremental lift you cannot scale-up
If attribution alone cannot be trusted, how do you scale up? In order to be sure where and when to increase your budget, you need to measure incremental lift. You need to know exactly how much are your ads bringing to the table throughout the whole marketing funnel.
How to measure incremental lift?
In short – statistical analysis & testing. I know, easier said than done and not always available.
The hypothesis is that your audience will react to your brand only after being exposed to ads. To test this hypothesis you should split the target audience into a test group, that will be shown ads, and a control group that won’t.
It can be done either manually (a rare case, unless you have a contact list) or by running conversion lift tests and brand lift tests that some ad networks provide.
By concluding these tests you can find out how many conversions were driven by your ads over a period of time. The results won’t be limited by the attribution window and won’t be polluted by organic traffic and native demand.
But what if you cannot run these tests?
In that case, you can estimate it by previous experience or changes in the total number of conversions/outcomes after switching on or off certain campaigns. I know, it’s not ideal – but it will be better than relying on attribution only. At least, you’ll be able to say “I think that Facebook/Google Ads drove an approximately 15% incremental increase in sales”. Google Analytics is your friend here.
What happens when you shut down the seemingly lower ROI channel that had the incremental effect in favor of the high ROI channel that had no incremental effect, but a strong selection effect?
This graph is worth 1000 words.
Stop relying on attribution only.
Stay in the loop with new blog posts & updates by subscribing to my newsletter.