Trading at the Speed of Light
In the world of high frequency trading, speed is everything and everything that affects speed matters. Firms that engage in high-frequency trading use optimized software and bleeding-edge hardware designed and built from the ground up — not only for the fastest speeds but also for 24/7 up time and the lowest possible latency. (The lag time between giving a command on one end of the network and having it executed on the other.)
Latency is especially bad for computerized trading. Since the fastest participant to respond almost always has the upper hand in a trade, shaving off a just few extra microseconds can mean the difference between making money and losing it.
Madison Avenue Looks to Wall Street
Both programmatic advertising and real-time bidding face similar technological issues around speed, latency and 24/7 availability. That’s where programmatic ad tech can take some clues from Wall Street.
Since programmatic RTB involves an auction, we usually think of price, not timing, as the determining factor. The ad impression simply goes to the highest bidder, right?
Not always. A number of factors can influence who participates (and therefore wins) a real-time ad auction — and who shows up in time to bid.
Let’s be clear, speed and latency requirements for programmatic RTB are nowhere near that of financial markets. RTB auctions have timeouts around one-tenth of a second, while high-frequency trading windows are orders of magnitude shorter. But just because transaction times are longer doesn’t mean it’s OK to be slow.
Speed plays a critical role in real-time bidding for several reasons.
DSPs need to compute thousands of variables to make a bid decision. A programmatic ad auction calculates a lot in each transaction — from demographics to location to service being advertised.
Faster DSPs can factor in more data points during that brief window of opportunity. This allows software algorithms to make a more accurate determination of where an ad should be placed, as well as how much that placement is worth to an advertiser.
As faster DSPs have access to more data, they’re able to factor in a greater range of variables — and that means more granular targeting to ensure the most engaged audience. Latency bidders are becoming increasingly valuable to advertisers who require precise targeting.
Latecomers Lose Their Seat
Exchanges often “white list” only the faster and error-free responders in auctions, while filtering out the slower participants unlikely to bid. This filtering can be based on machine learning (in the case of Google) or done manually. Either way, previous sluggish bidding activity can hamper future bidding opportunities.
Even within the fraction-of-a-second window of an RTB auction, the fastest DSPs will often have time to submit multiple bids. Faster platforms have time to optimize their price against their competitor’s bids without overspending, while sluggish platforms might be lucky to get a single bid in before an auction times out.
Given this, advertisers with slower DSPs are more likely to be bidding blindly and end up overpaying when they do win.
Save Time, Money, and the Planet
Along with processing more data-points in a split-second auction window, DSP software architected from the ground up for performance can also handle more overall queries per second than a platform running less optimized code. The keys here are scalability and availability.
Advertisers need a platform that can scale to meet demand and won’t get bogged down during peak usage.
Speed isn’t only beneficial for the buyer. In some situations, such as single-bidder auctions or RTB Direct relationships, a faster DSP can even help pages load more efficiently by taking less time upfront to complete the bidding.
Of course, more efficient software is greener, too. Fewer servers translate to lower energy use for computers and cooling systems. That means less CO2 being generated and a smaller environmental footprint. That’s not just good for advertisers — it’s good for everyone.
Eventually, slower ad networks will need to be completely re-architected or replaced if those companies want to stay competitive against faster rivals. In the meantime, advertisers looking for the highest ROI should realize the advantages of working with a high-speed, low-latency DSP, as well as the disadvantage of not doing so.