How to get ROI and tracking to play nicely together

Tuesday, 07 June 2011

The concept of the silo has been so overplayed in pitches to improve your marketing that I would understand if you stopped reading at this point. Well done for carrying on! I’ll try to reward your perseverance with a very innovative technique that really does help to bring together two worlds in client organisations.

Tracking is a key pillar of the management of brands. It is the province of research professionals and it covers the key topics of usage and attitudes in some depth, with diagnostics applied to the communications depending on the focus of the survey.

ROI (Return On Investment), meanwhile, has come to mean sales models, and is usually the province of the quant profile in the organisation. It will generally only focus on money or sales as a variable, and the subtleties of the message and media deployment get rolled up into simpler chunks to fit the model: TV spend by week, sales by SKU by week, a minimum of two years hard data!

If you try to model ROI from tracking insights then all sorts of problems crop up. The most basic of them relates to using aggregate data and having small samples sizes. That sounds like a technical cloak to hide behind, but just think of it like this. Most people have made up their mind about a brand, so let’s say 10% of people have something interesting happen to them each week and that you use 2–3 media, each reaching 10% of people each week. Somehow you have to untangle the ROI from the one number the tracking data gives you each week for purchase intent or whichever metric you’re looking at. You are trying to do that with quite small numbers of people exposed to the different media.

So, how would you untangle this in the normal world? Well, you might go for more depth, which means talking to individuals and understanding exactly what they saw or heard and what they think. In other words, looking at each person as an individual and not as part of a herd.

That approach is at the heart of the new contact estimator technique we have developed, which is delivered as a service. We take a regular tracking style survey and make sure it collects a little bit of detail on media habits. Then we take the full media schedule, such as TV ads, radio ads, newspaper ads etc. Using the two bits of data we have we can then work out the likelihood that each person in the tracking saw or heard each of the specific ads. There is a little bit of clever work needed analytically, but essentially, for every person interviewed, we know what they think and we know what they saw or heard.

This means our aggregate ‘one number per week’ is now 100 answers per week, and those answers contain a wide range of situations, such as those who did not see much TV but heard a lot of radio, or those who did the opposite, and all other combinations.

Give that to an ROI modeller and suddenly we can play with tracking data. We can isolate the effects of different media, understand the combinations and work with all the tracked variables. So ROI is not just about sales effects but can also be about checking if we have achieved our key marketing objective of changing perceptions of the brand and which media were effective at doing that.

Like many of Pointlogic’s techniques, this cuts across the disciplines of research (you need to make some small changes to the research design) and modelling (working with the data about what individuals saw and what they think).

We’ve learned a lot already, particularly in the area of synergy between media, which is traditionally very hard to pick apart.

As a case study, we applied this technique to campaign data of an insurance company and gathered some interesting results that can inform future campaign planning.

One of the key outcomes was how each medium contributed towards achieving ad awareness and the optimal frequency for each medium. For example, TV showed the highest potential increase in ad awareness by increasing the number of contacts, as well as the greatest build-up speed (2 TV contacts were enough to deliver about 70% of the potential ad awareness). Radio had the second-highest potential and print ads worked quickly (few print exposures delivered most of the potential).

The analysis showed how the actual TV ads delivered by the campaign (6½ on average) were far greater than the optimal ones (about 2), as diminishing returns for TV had already set in. This means that the insurance company would have been better off allocating part of the TV budget to other media or marketing channels. (Note that the data analysed in this case pertained to one campaign and the results may differ for other campaigns.)

The technique is generating a lot of excitement and we are working directly with clients, and interestingly also with research suppliers, who recognise the opportunity to let ROI and tracking play in the same sandpit without any tears and tantrums!

To find out more contact Tim Foley ( This e-mail address is being protected from spambots. You need JavaScript enabled to view it ).

Figure 1: example of a dashboard delivered as part of the contact estimator service