In the world of online advertising, targeting is broadly divided into two main domains of data you can target by - adhoc (in-session) data, and over-session data. Adhoc data is data that is collected during the current browsing session of the user, while he or she is interacting with our own website or with banners that we put on other publishers' properties. What's interesting about this data is that it is subject to few privacy issues, as it does not require cookies on the user's browser. And yet, it holds a lot of actionable targeting data.
Over-session data is data which is collected over separate browsing sessions of the user. It is collected using browser cookies (either one or multiple cookies). This data is much more comprehensive than adhoc data, and when both types are combined and analyzed for discovery purposes, they can form a very comprehensive picture of the user's profile, intents and interests, and sometimes even, his current state of mind.
As this data relies very heavily on browser cookies, however, it has three main problems associated with it. The first is the privacy issue which has been discussed at length all over the industry. The second is its reliability, as users sometimes block third-party cookies or remove them from time to time (the average life expectancy of a cookie on a PC is about three weeks). The third problem with over-session data is that it requires slightly more advanced web browsers, and here is the crux of the matter on mobile - most mobile browsers currently do not use or carry cookies.
The mobile device landscape
It is very important to understand that the mobile device landscape as we know it in the US and Western Europe is very different from the global mobile device landscape. In Israel for example, which leads the world in smartphone adoption, 70 per cent of new mobile phones bought are smartphones, whereas the global stats are that only 30 per cent of new phones bought are smartphones; the rest are feature phones, albeit that the proportion of smartphones compared to feature phones is rising rapidly. These feature phones run primitive browsers and do not store cookies, hence the importance of adhoc targeting data.
In light of what I wrote in the first paragraph, it’s tempting to come to the conclusion that there is little data to work with on mobile campaigns, but this would be the wrong conclusion, because there is far more relevant and valuable data in mobile adhoc data than there is in web adhoc data.
Data such as browser type, phone type (smartphone/feature phone), phone manufacturer, phone model, phone operating system, display size, carrier, language, wi-fi/cellular connection is available on mobile, and can be highly predictive for which campaign will run well for a particular user. These characteristics of the session can tell us much more about the user than the session characteristics available to us on the web, hence their increased importance in mobile campaigns.
The data can enable an advertiser to adjust a current campaign in real-time, and use the knowledge learned to inform subsequent campaigns. For example, a brand may have primarily targeted iPhone users because it felt that was the right demographic for its brand, only to find, on analysing the adhoc data, that more clicks are coming from users on feature phones.
So in this data lies the opportunity, but also the obstacle. The large number of parameters available to us is relatively large, but what is more problematic is the large number of values within each parameter, and on top of that, the values in many of these parameters get updated very frequently.
This creates a huge amount of combinations to test and take into account when deciding where we want to run our campaign, as most of these targeting parameters are available to us on the ad servers and ad networks such as AdMob, InMobi, Adfonic et al. The task of creating a target audience based on these Adhoc parameters and data is left almost solely to the account manager to carry out manually.
The account manager is faced with three major problems when it comes to running and optimizing a campaign on a mobile platform. The first is the availability and accuracy of the targeting data - usually, it is the ad network that reports where the ad was seen, based on the partial, structured data that it has.
The second problem is discovery - as the data is structured, it is not granular enough to run detailed analysis and find out which of the parameters actually had the most impact on the effectiveness of the campaign. And even if the data is granular enough to allow such analysis, a human being would not be able to analyze the amount of permutations involved.
The third problem is the false-positive problem. i.e. where did we not run the campaign that would have yielded better results, because we were not able to accurately analyze the data at our disposal.
Adhoc data on mobile campaigns carries a lot of targeting power. It actually enables better targeting than is possible on the web; it is available and very accessible. This puts mobile campaigns today in a position to deliver better ROI and better performance than web campaigns. However, it is accompanied by a data availability and accuracy problem (as it is currently only being reported by the ad server itself) along with a data discovery problem in a very dynamic environment.
In order for an advertiser to exploit the full potential of a mobile ad campaign, they need to address these issues by taking control over the campaign's data and analyze it accordingly, employing a platform that a platform that can manage and analyze mobile campaign data, in order to target as accurately as possible, while bidding accurately and efficiently.
Such platforms are out there; mobile advertisers could do much worse than investigate and deploy them.
Gilad Hellerman is CTO at DMG-DSNR Media Group