If these data clean room walls could talk … well, they can, actually.
Especially now that chatting with clean rooms is becoming a trend.
On Wednesday, AppsFlyer launched a function within its data clean room that allows marketers to pose their business questions in natural language rather than having to use SQL for data analysis.
AppsFlyer’s beta partners include travel booking app Hopper and a handful of other apps across the transportation, gaming, fin tech and food delivery categories.
Data discovery
When marketers want to query their clean room data, they’re usually confined to running predefined reports either daily or weekly, and these reports are limited by whatever fields and features are surfaced within the user interface of the reporting tool.
Otherwise, most marketers would need to call on data scientists and business intelligence engineers to run analysis for them, and those folks aren’t on call. They’re busy.
Messing around in a database usually isn’t a self-serve situation for people that don’t know a programming language.
But sometimes marketers just want to explore their data, said Edik Mitelman, GM of Privacy Cloud, AppsFlyer’s platform for privacy-focused measurement, activation and reporting tools.
“They want to discover new things and access data beyond what they could get from a report,” Mitelman said. “That’s why we launched dynamic query capabilities, so that marketers can tinker with their data and get the insights they need.”
You’ve got queries …
AppsFlyer first released its data clean room, which falls under the Privacy Cloud umbrella, a little over a year ago with a basic reporting UI. It has since added expanded analytics features over time in response to customer requests.
But allowing marketers to query the system themselves is a big improvement, Mitelman said.
“You can run any complex model ad hoc on the data, and this will give you results,” he said. “Just ask a business question, and it will answer you.”
AppsFlyer’s dynamic query tool is powered by OpenAI through a ChatGPT plug-in. Marketers log in and can prompt the system in any language that ChatGPT “speaks,” including English, Russian, German, French, Spanish, Dutch or Portuguese.
Marketers can run a range of queries, from the relatively simple (e.g., “show me my top 10 media sources in the EU”) to the very complex (e.g., “show me the number of app installs cross-referenced with how long it took between download and first open”).
AppsFlyer also accounts for potential hallucinations (when large language models sometimes generate false information) by giving people the option to edit an output before using it to run a full analysis.
Rising tide
All querying happens without data leaving the system, which means that, although the AppsFlyer’s model learns, no private information later appears online or is used to train OpenAI’s tools more broadly (which would negate the point of a data clean room, after all).
AppsFlyer’s data clean room is “cloud agnostic,” Mitelman said, so marketers can onboard their data regardless of which cloud warehouse partner they use.
But the data clean room category still has an interoperability problem. It remains “a headache and a pain in the ass,” Mitelman said, for advertisers and publishers to interact with multiple data clean rooms vendors without having to integrate all of them.
The IAB Tech Lab is working on a solution for that, though.
In July, it introduced the first version of specs for data clean room standards, which Mitelman helped develop as a member of the Tech Lab’s Rearc Addressability Working Group.
“I don’t expect the big walled gardens to ever open up, but we as independent clean rooms need to put our egos and our competition aside,” Mitelman said. “It’s not a winner-takes-all market, and if we can create interoperability, then we’ll be able to grow.”
