For the record, I have 5 hours of sleep in me now, so I won’t be completely incoherent, just the usual incoherence. Still got the shaking going on though, which is making me crazy.
One of the things that constantly comes up in web analytics/web marketing these days is where to spend money, what is the optimal mix, and what things can we cut because of poor performance. A couple of years ago the easy answer was to kill everything that had a low ROI or low E/R (expense to revenue). The problem with that approach is you would always knock off upper marketing funnel activities like display advertising (aka banners) and non-brand search terms, not to mention traditional stuff like TV and newspapers (which are difficult to track anyways). By cutting upper funnel tactics you’d squeeze your funnel and only invest in lower funnel activities like affiliate and brand-term paid search. Your E/R would look impressive but you are actually adversely affecting your business as you are not reaching a big enough audience to grow the amount of visitors at the top of the funnel.
When online marketing budgets get cut, invariably the first target for elimination is Banner ads. Why? Because generally speaking people don’t click on them. There are obviously exceptions, and companies are getting increasingly clever to devise banners/gadgets in ways to increase visitor interaction (which is a rant for another day). But in general, people don’t click on them and come to your site and buy. If you use a strict definition of click to conversion, banners are always amongst the worst tactics you can throw out there. So how do you prove that banners are worth investing in? Enter the lovely measurement called ‘Post-Impression’ and its other aliases such as ‘View-Through’.
First, a little bit of an introduction into what post-impression means. Essentially, it is the idea of tracking a visitor who has been presented an ad (but didn’t click) and measuring to see if they ever show up at your website. DoubleClick/Google, Atlas, and other ad-serving vendors have this ability. You can track whether or not those visitors that were presented an ad on another site came to your site and did things like purchase, optin, downloaded info, or any number of success events you have concocted for your site.
Example…say we put up some banner ads on CNET and you ventured over to CNET and one of our ads was presented to you in the right hand corner. Basically you have a cookie now that the ad was served to you. If you came to our site in the next 60 days, we can see that you were presented that banner ad. If you buy something, we can attribute that purchase to having served you that ad.
Sounds cool aye? You can now run out and say we can prove banners work because we have revenue attribution! Well, not so fast there. On to my problems with Post Impression…
Cautionary Tales of Post Impression
The biggest hurdle in post-impression is you actually don’t know if a visitor saw your ad. It is the billboard effect. Just because you have a big billboard on I-40 saying ‘Eat at Joe’s’ you don’t know if anyone saw it and actually put that information into their brain that you exist.
So no, we can’t prove that a visitor saw the ad, which is why I wouldn’t subscribe to taking a full-attribution of revenue when talking about Post-Impression revenue. There is also another reason you don’t want to do that and goes back to something I’ve written about many times here. There is an interaction effect of all campaigns amongst each other. If you took 100% of post-impression revenue blindly, you’d actually be double counting because often these same visitors come to your site through other means like Affiliate and Paid Search. We’ve actually been using a calculation of attributing 15% of post-impression revenue as true revenue, and that is actually just a guesstimate. A better way, which I plan on implementing, is using 100% of all the Post-Impression that shows up as ‘None’ under your campaigns in Omniture, as that would alleviate the double-counting problem.
How to use Post-impression responsibly
I stress this to folks all the time, so much so that people inside of Lenovo probably tune me out. In any case, I do propose 2 major uses of Post-Impression:
- With the Omniture/Double Genesis integration you can see the influence of banner ads on other tactics you are doing. For example, with the Genesis integration you can determine what percentage of visitors that come to your site via Paid Search had been previously exposed to your banners and ending up buying something. In this case, banners are acting as an ‘assist’ to help other online campaigns. Essentially, you are subsidizing other programs by doing banner advertising. The idea being if you stopped doing banner ads, your Paid Search will definitely be affected in a negative fashion. The key point here is you can no longer look at tactics in a vacuum by themselves, they all work synergistically together to create sales. In fact, some of the tools we use for web analytics need to embrace this ideal even further to help marketers understand how all of their tactics work together and how to optimize the mix effectively.
- The second major thing I’d suggest using Post-Impression for is to show revelancy of where you are placing ads. Because Post-Impression shows you that someone was exposed to an advertisement and bought, you now effectively know where all your customers are hanging out on the web. This knowledge has some interesting benefits to allow you to maximize your marketing efficiency. Addtionally, post-impression tracking gives you the insight into poor performing sites so that you can test different creatives to see if you can generate more traction (ie sales) for your brand to visitors of sites that traditionally didn’t buy your product.
As mentioned, one of the biggest problems with Post-Impression is the lack of true knowledge as to whether or not someone actually saw the ad. Hard to prove that they glanced at something on a page unless we start doing some retinal-tracking (which I assume might violate some privacy agreements).
So what is a better way? Gadgets and interactive banners are the best way I can think of. Marketers are increasingly embedding more information into banners that visitors can interact with. By interacting with the banners, you now know for a fact they saw the ad because they are engaging with it.
The biggest problem I see so far is not that we can’t track it, but that there isn’t something in place for DoubleClick and Omniture to pass that information. Instead of metrics like ‘view-throughs’ and ‘post-impression’ I think the next logical step is ‘post-engagement’. But right now that doesn’t exist within Omniture nor DoubleClick (as far as I know). It needs to be there as more and more savy marketers are going down this path in order to provide rich content and truly measure banner effectiveness.