How Can We Move To High Predictability Ad Serving?

nanda-kishoreData-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Nanda Kishore, chief technology officer at ShareThis.

When you’re an Olympic sprinter, the only thing that matters is speed. But when you’re an online advertiser, you need more than just speed to beat the competition.

For the past couple of years, online advertisers have focused on low latency as a goal. The basic idea is relatively simple: The faster you get your ad in front of an interested consumer, the better. While the idea may seem simple, simultaneously serving ads in real time and in a relevant context has proven to be one of the more difficult challenges facing modern brands.

When I talk to marketers about achieving speed and context in the same moment, I like to say that we’re moving beyond low latency and toward high predictability. But how do we take this from an idea to an actual practice?

The answer is twofold. The first step lies in developing organizational and technical expertise in processing vast data sets that contain online user behaviors. In analyzing these data sets, we build models and predictions and identify emerging trends. The ability to then apply the resulting insights to advertising campaigns is key to the relevancy component.

The next step in the process is establishing context. The analysis of data and application of insights must happen in near real time. This insures that brands maximize the opportunity to reach the right consumers at the right time and context.

For example, consider a consumer electronics advertiser interested in reaching potential buyers of its newly launched product. Its target customer skews toward middle-aged males, with average household income; all of this information is slow or static data. This specification may be viewed as a general guideline, but not a guarantee, of likely purchasers.

During the campaign, a subset of this audience pool engages in online activity of interest, such as sharing product reviews with friends online, presumably to solicit advice and input, or browsing relevant product informational content of competing products. Further, real-time feedback data on the campaign indicates strong correlation with sports content as a positive context for ad response.

These signals represent fast or dynamic data, and when applied to predictive models, it may score this audience subset as most likely to purchase. The ability to process, compute and leverage these signals and resulting insights in near real time can yield higher than average ROI for the advertiser. It is critical to reach the audience most likely to engage at or near the “moment of truth.”

Understanding Online Inventory

It’s important to remember that while it’s fairly easy to predict the traffic and audience of an entire website, it’s not so easy to understand the life cycle of a specific URL, article or trend. This activity fluctuates on an hourly basis, and brands need to be agile enough to reach consumers when they are engaging with this trending content.

The nature of online inventory and audience behavior is more volatile and less predictable than ever, so to maintain relevancy in messaging and timing, advertisers need to first understand these granularities. Developing predictive models for campaigns has become key for brands to navigate this volatility in order to successfully reach their target consumers. Although we may have landed on a solution, like most things, it is easier said than done.

Data Enables Development Of Predictive Models

When online advertising began, marketers simply assumed it would be a faster, cheaper and more efficient version of traditional advertising. As a result of their limited understanding, they tended to rely on static data such as age, gender, income and geographic location.

Today, we are much more interested in using dynamic data gleaned from active behaviors such as browsing and content sharing. Combining that static data with fast changing behaviors will bring advertisers closer to predicting the precise moments that consumers are ready to buy in, or will react positively to their advertisement.

Smart marketers will take the time and effort necessary to find the right context for the ads, but just because the industry understands the need for context, it doesn’t mean we’ve cracked the code on how to achieve it.

It would be nice if the complexities of online advertising could be reduced to a simple metric, but that’s not likely to happen. Winning this race will require a combination of dynamic data, predictive analytics and great marketing insight. Speed alone won’t bring home the gold.

Follow Nanda Kishore (@nkishore), ShareThis (@ShareThis) and AdExchanger (@adexchanger) on Twitter.

1 Comment

  1. This can be achieved with a higher probability with a Beacon Technology Ad network. Tie this up with user identified preferences, location based certainty, hi viewing rates with push notifications and you have a potential winner. Enhancing this with more user date and you have a real winner.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>