Role of user data in a DSP Platform - Part II

I had mentioned in my previous post about the importance of user data. Now I will share how one can leverage data more effectively in delivering a campaign via a DSP.

During my playful days that I spent with rich user data, I was able to achieve some really outstanding results during campaign optimizations. We undertook data integration into our DSP platform very seriously and made some good investments around data collection, organization and activation. As a result we saw really good traction and adoption of our platform and it proudly sat over a huge segment data of unique cookies in India region alone. That was some really rich user data that we used for audience targeting and retargeting purposes and delivered significant value for our customers.

During the exercise I saw some really interesting insights which I wanted to share here:-

1.   It takes some time, in some cases 2-3days to a week to 10 days, for user data to start showing desired results. One has to leave a campaign running for some days and do just enough investment for the DSP optimization algorithm to learn and then kick in. A DSP platform or any ad server for that matter run on algorithms and in order for the machine learning predictive algorithms to learn how to optimize themselves, they must achieve a point of statistical relevancy. The campaign has to serve some thousands of impressions over various sources of inventory, and register a couple of hundred clicks or conversions, for the DSP platform to confidently start predicting correct impression value.

2.   Campaigns running on audiences with small number of unique users generally struggle on exchanges for delivery. Small audiences are difficult to scale in terms of volume and the predictive algorithms couldn’t get enough data points to predict accurately on those audiences.

3.     Retargeting works really well for advertisers but only when it is used effectively with frequency capping. Absence of frequency capping leads to over exposure of retargeted campaigns to users thus irritating them and also making them virtually blind to those ads after a point.

4.     Dynamic creative optimization equipped with smart user data collection works really well for Ecommerce players. Customized offers and messaging to users based on their previous browsing patterns really helps in pushing them down the purchase funnel.

5.  Also audience targeting works really well on RTB or an exchange which has much greater reach in terms of unique users than what a managed network can provide. Hence audience targeting can do great wonders for a DSP platform.

I’ll just want to end this conversation by saying that what the online Ad industry is seeing or doing with user data currently is just like scratching the surface. There is hell lot more that can be done with user data. The next big challenge for most companies out there dealing with data is to use predictive analytics and statistical modeling techniques on data, to get more meaningful insights like determining look-alikes, duplicates, calculate user affinity, build algorithmic audiences,  decrease noise in audiences. revenue attribution to right audiences etc.

Role of user data in a DSP Platform - Part I

Key benefit of any good Demand Side Platform (DSP) is to buy media programmatically and efficiently based on its algorithm’s ability to understand data and other business criteria. The most important aspect of the DSP is to utilize user data to accurately predict the value of a user for a given impression.

Hence for any DSP platform it is of paramount importance to invest in their bidding and predictive logic to win impressions that are most valuable for their customers and help deliver the end business goals. It’s much more than clicks and conversions that we are talking about here. A good DSP has to give much more weightage to the audiences from which those clicks and conversions are coming from because it is this audience and their positive mind share, which the advertisers and brands are really after.

So it all begins with a data strategy. How much audience data do we have in our region, total number of unique users, total number of segments to which these user belongs, audience profile of the users, trends and audiences that really convert for a given campaign, uplift seen in a campaign due to audience data etc.

Those days are long gone when marketers used to run after premium inventory and totally ignore remnant inventory. With DSPs becoming mainstream, agencies and ad networks have now started leveraging user data to make every impression coming even from remnant inventory almost as valuable as an impression coming from a premium inventory. Hence with user data the gap between value of impression coming from a premium inventory versus remnant inventory has drastically reduced. And, in time people will stop differentiating between impressions based on inventory source but will start valuing impressions based on audiences.

In my next post I'll share some practical insights around behavior targeting and retargeting optimizations in a DSP platform.