Changelog: The Tech Behind Gumshoe
CTO Andrew Pile’s inaugural post dives deep into the process of building Gumshoe, Gumball’s new impression tracking tool.
Today, we’re introducing another C-suite quarterly series, Changelog, where you’ll hear directly from CTO Andrew Pile on the engineering processes that make Gumball, Gumball.
In the series launch, Andrew shares more about the creation of Gumshoe, a new impression tracking tool launched in September 2022.
This interview has been edited for brevity and clarity.
Andrew, or Pile as he’s known around the office, started at Headgum back in 2017 when he was brought aboard to create the marketplace that we know today as Gumball. For five years, Andrew has been the sole engineer behind the scenes of Gumball. (“I haven’t been busy at all,” laughed Pile.) Today, he leads an engineering team of four and together, they built one of Gumball’s newest products, Gumshoe.
Before we delve into the process of building Gumshoe, it helps to step back and understand how and where it fits into Gumball’s overarching mission of democratizing host-read ads for advertisers and creators.
Gumball’s goal is to be an end-to-end solution for advertisers to do everything related to their campaigns: find audiences that align with their strategy, easily manage campaign assets in one spot, and most importantly, execute campaigns that convert.
“We wanted Gumball to have a measurement component to it in addition to it being a sales platform in order to increase trust and transparency between our advertising and creative partners. Gumshoe is something that we planned for and wanted to be able to do from the beginning.
While it was something we had planned for, it was a completely different challenge than the Gumball website. Gumshoe is a product with a much different scale — one with potentially millions of users. This was not something the engineering team (me), could tackle immediately. That all changed after our Series A round earlier this year, as the capital allowed us to hire the engineering talent needed to expand the world of Gumball.
In a past life, I worked with Peter MacArthur at Vimeo, where we spent the better part of a decade together building what would become one of the largest sites on the internet. After our Series A, I was able to convince Peter to join Gumball and he led the effort in building Gumshoe. Vimeo was a massive technical operation that served hundreds of millions of users, so I knew we had the experience to architect a solution that could reliably scale in the podcast world.”
Gumshoe’s beginnings
The team initially tried a proof of concept last year with data from pixels, snippets of code that gather information about a user on their device and are used to track ad impressions on dynamic ads. They sought to understand how the various hosting platforms sent pixel data and what kind of information they sent in order to get a sense of the volume of data they’d be dealing with.
“We knew right away that we’d have to approach Gumshoe a bit differently: from capacity planning and how it’s going to have zero impact on end users, to how to process data securely and with privacy in mind. All of that informed our decision-making about how to design it, how to store data, and what systems to use.”
While he understood how to tackle some of the big problems, others were trickier to address. One of the big challenges with Gumshoe and analyzing audience numbers is that there are many best practices around what constitutes a “play”. IAB sets these standards and guidelines, but it’s the job of the people implementing it to make sure that it abides by those rules. To be considered a play, the IAB says at least one minute of the podcast must be downloaded. Still, there is no standard for how to count the number of plays in a given day.
“It turns out there are a lot of differences between the way hosting platforms count plays! Some count a rolling 24-hour window, others count a calendar 24-hour window, and still others have windows of different sizes. We learned a lot about how others operate and I think we’ve come up with the solution that’s the most accurate and consistent across hosting platforms.”
Since all of Headgum’s shows are on Gumball, the team was able to use our own shows and their different hosting platforms as the sandbox to test, learn, and iterate on Gumshoe until it was perfected.
The Gumshoe process
So, how does Gumshoe work exactly? In essence, Gumshoe is looking for what podcast plays count as qualified listens, as determined by the IAB guidelines mentioned above. Let’s break it down.
- A user requests (downloads) an episode of a podcast using the Gumshoe prefix.
- Gumshoe receives their IP address and turns it into an anonymous key to assure user privacy.
- Gumshoe applies IAB guidelines such as removing duplicate downloads and filtering bots and datacenters.
- Gumshoe then checks to see if a user has downloaded at least 60 seconds of the audio file to be considered a qualified listen by Gumshoe.
- Qualified listens are aggregated and displayed in a performance bar in the Gumball UI.
Learnings and thoughts
Podcast delivery analytics can be quite messy, but if the tools are designed correctly, they can add a lot of clarity. Advertisers are hungry for transparency in podcast ad delivery, much like what is available in other self-serve platforms. Gumshoe is the first of its kind to deliver on that need and do it within the same marketplace. For podcasters, a tool like Gumshoe reduces the need to verify data through screengrabs and helps build trust between the creator and the advertiser.
Analytic tools can also require monthly fees. Gumshoe is available to all who use Gumball. If you’re just getting started and trying to track effectiveness, Gumshoe opens the doors and you don’t have to pay an external source.
“It’s crucial that we continue to democratize the business of podcasting. With consolidation increasingly taking over the industry, we’re committed to keeping the playing field fair and accessible.”
Finally, why the name “Gumshoe”?
Gumshoe is called that because it’s a little bit of an investigation. As the team collected data, the process felt like we were investigating something to uncover the truth.
“Gumball is such a fun name and offers a lot of room to play around with when naming new products. I couldn’t pass up the opportunity to grow the 'Gum' universe. Maybe one day we’ll run out of features to name Gum-something, but I think it’ll be a while.”