Comparing programmatic ad buying with stock trading is a dangerously inaccurate analogy
If you’re in the ad tech business, you’ve heard analogies between high-frequency stock trading and programmatic ad buying. This is natural since a lot of ex-traders wind up in ad tech—our CEO Adam and myself, for example—and there are indeed similarities, but they’re superficial. Using the analogy comparing ad tech and stock trading at Thanksgiving to scoot past an inquisitive Uncle John on your way to more eggnog isn’t only bad, it’s dangerous since high frequency trading is really nothing like programmatic ad buying. In short, in programmatic, we buy impressions, which are social interactions whilst stock traders are buying assets that have their own intrinsic value.
An impression is not an “asset”
With high-frequency trading, assets (stocks, bonds, etc.) are bought and then sold for a profit (hopefully). This is mostly arbitrage. The underlying thing you bought hasn’t changed.
In programmatic ad buying, we don’t buy assets. We buy the right to show a consumer an advertisement on a mobile phone and the right to have a social interaction with a consumer. We can’t just turn around and sell this right for a profit. Also, we must create value on top of what we bought by delivering a relevant ad to a given consumer. For a more fitting financial analogy, we’re in tiny mergers and acquisitions (but at massive scale)—more on that below.
We buy in an auction, and high-frequency traders buy at fixed prices
With high-frequency trading, you buy at the ask and you sell at the bid. In other words, at any given point in time, there is a price for an asset that is supposed to be the fair market price based on all publicly available information. This price fluctuates, but at any given point in time it’s fixed (i.e. known).
We buy impressions from publishers in an auction. There is no set price in programmatic buying (outside of private placements). This means that to place a bid in the auction, we have to have some notion of the value we can create, which leads to radically different strategies, business relationships, and technologies.
OK, so programmatic and stocks are different. Why does that matter?
For one, the risks advertisers face aren’t adequately addressed in the high-frequency trading model. Since we aren’t buying assets and we’re creating interactions through our ads with potential buyers/consumers, the range of outcomes is huge (and scary). They range from serving a highly relevant ad to “wasting” the impression to serving an ad that is counter-productive (think Toyota serving on a site with inappropriate content). Not only can we completely waste our money, we can even harm our brand. For high-frequency traders, they’ll have an asset that likely hasn’t moved too much in value, and the range of outcomes is smaller.
Building on the above idea, programmatic ad buying isn’t about finding deals for cheap (i.e. arbitrage), it’s about finding opportunities for creating value. In other words, programmatic ad buying should be a very positive idea and not necessarily cost-focused, but value-focused. Focusing on buying cheaply can produce some really bad consequences for advertisers.
Also, the ad you serve matters. This is probably the biggest danger in thinking about the world of programmatic as similar to high-frequency trading. For a high-frequency trader, a dollar is a dollar, whether you’re trading blue chips or the pink sheets. As a marketer, you want to build value for your brand regardless of if you are a direct response buyer or a big brand. To do this, you must be as concerned with the audience you’re reaching and the message you’re sending as you are of the algorithms behind how you bid for impressions. To be clear, I’m not forgetting that marketers need to be profitable. They must. I’m simply pointing to the fact that the ad creative and the audience are the most vital pieces of any ad campaign.
Finally, in high-frequency trading, there isn’t supposed to be asymmetry of information. Asset prices should reflect all publicly available information, so anyone who has asymmetric information is cheating (insider trading, etc.). With programmatic ad buying, asymmetric information is not only acceptable, it’s the only way to run a programmatic business.
To understand this, remember that we are creating social experiences at scale, and by meticulously recording and analyzing all the data you can to “learn” over time, we can better construct social experiences. So the more data we record, the better we get at building these social interactions on the fly. Further, the ability to serve an ad is zero sum—we get to serve it or someone else does. So the better an ad network gets, the better it will continue to get. Ad networks suck up all the data available and use it to build better and better interactions so it can continue to win more and more. As an advertiser, you 100% shouldn’t work with anyone who doesn’t understand that.
So that’s it. High-frequency trading and programmatic ad buying aren’t really similar at all from our perspective, and most importantly, the analogy is bad for advertisers in general.