Geotargeting: Not As Cool As You Think

Sometime in the middle ages, a young lad approached me and asked*: “If we already know where our products are popular, or if we want to test-market in a certain city, what’s the point of wasting money and lives by transporting our ads over mountains and across rivers to foreign lands? Why not only display the ads where the customers are?”

Well, in that context, the kid had a point. After all, if a location on your map says “Here There Be Tygers,” it’s a safe bet that the people who live there aren’t going to be interested in ads for your local mead. So we scaled back on the mules and ships, and our ROI went way up.

It’s too bad the kid died in 1351 (typhoid), or I’d be able to tell him that his ideas might have been sound in Chaucer’s day, but they just don’t apply to online media. Here’s why.

Ad inventory on a website is fluid. A print publication can more or less guarantee a certain number of eyeballs, because it has subscribers, and presumably, if you’re paying to receive a magazine once a month, you’re going to read it once in a while. But a website…who knows? Just because you got a million unique pageviews last month doesn’t mean you’re going to get them again this month. Maybe people won’t be as interested in your content this month. Maybe they won’t like the new background color, and stop coming. Inventory prediction is not an exact science.

That being the case, online publishers like to sell their inventory in broad strokes. Placements that they can guarantee without worrying too much. The homepage: it’s practically guaranteed to get views. If we’re talking about a big magazine site, then they might sell you the ‘news’ section. Or the ‘style’ section. Whatever. (Yes, we’re talking about contextual targeting only here, for purposes of the model; bear with me.)

Well, geotargeted inventory is just about the hardest type of inventory to guarantee. Try going to just about any successfully monetized site, and tell them that you want them to only show your ads to people in, say, Texas. Let’s assume they agree to do this (which is a big assumption). How many impressions are they going to guarantee you?

Let me put it this way: not enough to make it worth your while. Most large sites can break down their traffic by region, but traffic for specific regions is so hard to predict that it’s just not worth doing. The publisher doesn’t want to end up owing you a makegood, so they’ll under-promise. What will you get: one thousand impressions? Two thousand? What are you going to do when the end of the month comes around, and you’ve only spent 1/100th of your budget for that month? Sure, there are worse mistakes to make, but trust me, it’s embarrassing.

Of course, there are certain sites that will enthusiastically tell you that they get read by millions of Texans each month. chron, for example. That’s right; it’s a Houston-focused site.

That is the only practical way to geotarget. I think of it as ‘manual geotargeting.’ You pick sites that are already getting the majority of their traffic from the region you’re trying to target, and you just buy space on those sites.

It’s also reasonable to buy on sites that are not region-specific themselves, but that have large region-specific segments, such as CitySearch.

Or, if you’re aiming at New York and California, just buy on large national sites and don’t bother targeting. You’ll save yourself and the publisher heartache, and you’ll end up hitting your target just as effectively.

* Translated from Middle English.