So, we’re all pretty happy with how online measurement works. Right?

Ask a digital marketer why your marketing pounds should be spent online and, among the many (good) reasons they will come back with, will be its unrivalled accountability and targetability (not actually a word, but who cares).

They will tell you that you can find your likely customers (users who look like your existing customers), target them with your message and, best of all, see how many of them buy something from you as a result.

All of that’s true and, when put together well, it results in incredibly powerful marketing activity. The problem is that, with all of the data that’s available, it’s become a very easy system to manipulate. Online exchanges and DSPs have become very adept at ‘playing’ it so that everybody always appears to be doing an awesome job.

Thanks to the last-click model, in particular, running display advertising that looks good is actually pretty easy. Simply develop a method of finding users who look (for whatever reason) as if they are likely to convert, serve them an ad (and a cookie) and then take credit for this group’s higher than average conversion rate.

As such, serving an ad in an undesirable location (but to a desirable user) below the fold (where it’s nice and economical) has become a successful way of cheaply dropping a cookie on a user who looks likely to convert anyway.

Thus, taking all the credit for a conversion it had little, if any, role in driving. Think of it as the kid in the school playground smashing the ball into the net from two yards out when it was going in anyway and then running off to celebrate.

No one liked that kid.

The world of online display is increasingly becoming a battle to continue to look efficient by dropping the final cookie. I’ve written about the causality issue when evaluating online marketing here on ExchangeWire before, and also about how we deal with those issues here at Havas Media.

The market is well aware of these problems I’m sure, I’m certainly not the first person to point them out. So, why is this still the way it’s done?

Because the prevailing opinion seems to be that the last-click model provides a good proxy for site-placement and keyword performance, even if the overall ROI figures are inaccurate. So, whilst not perfect, it should still allow for analysis of relative performance.

That’s not true though. The last click model assumes that every user who sees a display ad then converts did so purely because of that ad. This means that, inevitably, display ends up taking credit for conversions that would have happened anyway. So, every piece of display is starting from a base conversion rate, the likelihood of the user converting whether or not we serve them an impression.

Without taking this base into account, analysis is measuring the incidental rather than the incremental effect of the activity. ROI analysis inevitably benefits cheaper placements in this model because, with everything starting from a high base, incremental effect makes up only a portion of the overall ROI.

This base conversion rate effectively gives cheaper placements a head-start when it comes to ROI and CPS and the incremental effect is rarely enough to compensate.

As a result, demonstrating the effect of any innovative and different display activity is extremely difficult. Branded online display still hasn’t really taken off because, no matter how good the creative, in a last-click-wins model it can never compete with dropping cookies on people who already look likely to convert.

Marketing is not supposed to (only) be about preaching to the converted, it should also be about changing people’s minds and introducing them to things (brands/products) they’ve never heard of.

The thing is though, stuff aiming to do that (rich media, brand building-display activity, etc.) never does well at driving the last click. People usually like to do a bit of research, particularly for high-value items, before they convert. So by the time they do eventually buy something, they’ve often been hit by some piece of opportunistic retargeting that steals away the credit.

To honestly evaluate display advertising would mean an inevitable reduction in ROI. Econometrics shows this to be true by continually assigning less credit to display than the last-click model.

In order to be able to make display campaigns be better, we need to accept that they will have to look a bit worse. Who’s brave enough to take that jump?