Solving Reach Problems through Strategic Geographic Expansion:
Application of Tobler’s Law to Programmatic Optimization Strategies
Accurate impression forecasts have long been the unicorn of the programmatic universe. Even the most sophisticated forecasting tools yield avails that need to be passed through additional buffers to provide realistic expectations on a proposal. This is especially true when the geographic target is smaller than a DMA.
AdOps teams work with their sales counterparts to establish what those buffers should be based on real execution experience mixed with additional targeting considerations and a dash of hope. This forecasting methodology is accurate most of the time. However, the real-time bidding environment often has other ideas about what is pragmatically available when it comes time to bid on impressions.
One example that comes to mind is a political campaign we ran last election season. The flight was short: just the two weeks leading up to election day. The geo target was also extremely small – just one zip code in suburbs of the Twin Cities. The forecast may have been accurate any other time of year, but it didn’t account for the fact that the two weeks leading up to election day every four years are insanely competitive. Our reach was shot as soon as the campaign started – we weren’t winning the daily impressions we needed to keep on track to fulfill the campaign.
What can we do when our win rate is abysmal at a price point that still yields a positive margin and we only have 2 weeks to right the ship? Let’s assume we are already using all creative ad sizes, and have adjusted the frequency cap to match the campaign goals. Other than bidding higher or compromising the intended user/site targeting, what else can we consider?
The famed geographer Waldo Tobler states in his First Law of Geography that:
“Everything is related to everything else, but near things are more related than distant things.”
Sounds too simple to require its own law, but it is the foundation of geographic research and it applies here. In our example, the target was a single zip code. This was important to the client because only users registered to vote in the one zip code could vote on his candidacy.
Fair enough. But why not hit those voters on their mobile devices when they are waiting for a table a restaurant in the next zip code over? Or when they are doing “research” on their computer at work a few miles away, but in another zip? Why not hit their friends with whom they converse with regularly?
By using a secondary geographic target, we can still prioritize the serve in the original zip code, but can then bid (at a lower amount, and as a back-fill option) on the adjacent zip codes and extend our reach so we can fulfill the campaign.
Now we’re hitting more of the target audience, and can even lower our clearing CPM, increasing our margin. This all comes at no determent to the campaign! Most forecasting tools don’t account for hyper-seasonality or other short-term competitive upswings. As such, we need to be strategic and flexible when we go live and encounter pacing issues.
Effective communication with the client is obviously critical in these cases, and talking about contingency plans such as this before going live will always make the conversations smoother if/when it comes time to make a change. Geo-targeting flexibility is a must when targeting anything smaller than a DMA, especially when layering on additional targeting strategies.
Written By: Adam Shaffner, VP of Operations