We’ve long been big header cheerleaders at AdMonsters, but we also have never shied away from one of its biggest setbacks: potential latency issues. Indeed, fear of latency has kept some premium publishers away from header integrations and continues to make others wary about embracing header tech too closely.
At the same time, maybe we haven’t gone enough into the specifics of how header integrations add to latency—and how to measure the effect. PubMatic’s breakout on header wrappers last week at the Nashville PubForum brought to light that there’s a great deal of confusion around where latency rears its ugly head and how to measure and mitigate it.
PubMatic Architect Abhinav Sinha notes that evaluating latency concerns is not just as simple as monitoring ad call timeouts. In the interview below, he offers some actionable advice for monitoring and minimizing latency—not just when it comes to your header business.
ABHINAV SINHA: Ad calls time out because of network latency, right? Not necessarily.
Similarly and contrary to popular belief, there’s actually a cap on how many outgoing network calls can happen at one time. The maximum is seventeen parallel connections and even fewer per domain, which decreases further for older browsers and mobile platforms. In other words, SSP partners that make multiple calls are actually making page performance worse and increasing overhead.
The good news is that most of that added latency is within publisher control. Execution overhead is a matter of page, content and implementation issues that vary from page to page and publisher to publisher. These issues impact all your bidders equally, but they make comparison a bit tricky.
On measuring individual latency components, we observed page execution overhead contributing to majority of the latency:
Device/browser level latency data will be key in determining optimization parameters like timeout, partners and eCPM. Looking at the sample data you might look at changing timeout for mobile web inventory vs desktop inventory.
AS: Here are a few steps you can take:
- Load inline iframe-based widgets and video players using dynamic and asynchronous iframes.
- Reduce network congestion by delaying content load which is below a viewable area, e.g., content which is loaded only when a person scrolls down.
- You can choose to use standard optimization tools to analyze pages on mobile as well as web platforms.
- Analyze advanced run time performance and execution with developer tools.
- Compare apples to apples – with all your header bidder partners, be sure you know what latency data you are looking at: network latency, overhead latency or a combination.