ChatGPT Makes Its Way to Data Clean Rooms

Analytics firm AppsFlyer introduced OpenAI’s ChatGPT within its data clean room capabilities, the company announced today. This means marketers can query data and ask business questions directly without the need for technical support, allowing for faster and easier adoption of data clean rooms. Marketers can input and receive queries across several languages, including English, Russian, German, French, Spanish, Dutch, and Portuguese. Top line Instead of querying business questions such as “Show me my top 10 media sources in the EU” into the SQL, marketers can post the same query into ChatGPT for data analysis. Find Your ‘Why’ Behind Data Clean Rooms Ideally, marketers would need to rely on data scientists and engineers to run this analysis for them, according to Edik Mitelman, general manager of privacy cloud at AppsFlyer. And complex queries, sometimes hard for analysts to write, would take weeks. However, thanks to the ChatGPT plugin, such queries can now be completed within minutes. “The need for us was clear,” said Mitelman. “We wanted to put the power at the fingertips of the marketers.” But large language models (LLMs) tend to hallucinate; make up items that are either inaccurate, don’t make sense, or don’t match the information it is fed. To address this, AppsFlyer gives marketers the option to edit the results, which also enables the LLM to learn from its database. Several partners of AppsFlyer, including Hopper and other gaming, ecommerce, food delivery and fin-tech brands, have used this Generative AI plugin. “With AppsFlyer’s Dynamic Query Engine, we were able to filter our DCR output and quickly focus only on the relevant data we need to make the best bottom-line decisions,” said Leif Meyer, user acquisition engineer at Hopper. Between the lines AppsFlyer’s query tool utilizes the capabilities of OpenAI’s ChatGPT plugin, however, it's not public-facing and all querying happens without the data leaving the system. “Nothing that happens within our model is contributing to the learnings of OpenAI,” said Mitelman. Nonetheless, it's noteworthy that LLMs are designed to learn from the questions posed to them and are trained on data to generate responses. And privacy concerns arise if ChatGPT has direct access to the data sitting within clean rooms. “If ChatGPT just knows the database structure, but has no direct access to the data in the clean room, there is no leakage of that data,” said an industry executive who requested to remain anonymous to maintain business relationships. However, a second executive pointed out that even for SQL results to be shared, ChatGPT would require access to that data. "It would need to know what the data was in order to match a natural language question to the data it was interrogating," the executive said. “It's unlikely privacy compliant unless [AppsFlyers and OpenAI] have a very specific contractual clause." In response, Mitelman told Adweek that the ChatGPT used by AppsFlyer has no access to the data within clean rooms. "It just knows how to write SQL that will answer the question asked of it," he added. Bottom line All this said, the growing use of Generative AI chatbots like ChatGPT is inevitable within ad land. Mitelman anticipates that this technology will see increased utilization across various data clean rooms as the industry works towards resolving issues related to interoperability. Elsewhere, Samsung banned the use of ChatGPT internally in May after an internal trade secret was accidentally leaked into the LLM. Other banks like Bank of America, Citi, Deutsche Bank and Goldman Sachs, too have restricted their employees from using ChatGPT over privacy concerns, per TechCrunch.

Data Management

ChatGPT Makes Its Way to Data Clean Rooms

AppsFlyer introduces OpenAI's ChatGPT plug-in for SQL queries

Top line

Trishla Ostwal

About

Subscriptions

Events

Publications

Top line

WORK SMARTER - LEARN, GROW AND BE INSPIRED.

Subscribe today!

Trishla Ostwal