An hour on signal engineering, incremental measurement, and what LLMs and agentic workflows actually change.
Speakers
Gufeng ZhouCreator of Meta's Robyn
Arthur AngladeMeasurement strategy, Haus
Amit EdelsteinAsks the hard question
Sam NamProduct Lead, PvX Delta
Aniruddha ChattopadhyayLead Engineer, PvX Delta
Lucas LovellInterim CPO, Paddle
Moderated by Chris Jones. Hosted by Noy Rotbart, PhD and Brian Brost, PhD, co-founders of Churney
We will email the joining link to before the session. Add it to your calendar now so it is in the diary.
Six guests who work on different parts of the problem, and the two people who build the signal. Chris Jones moderates.
Creator of Robyn · Executive Advisor, Vectorial Advisory
What Robyn is
Meta's open-source marketing mix modelling package, and the best known open-source MMM in the industry. Gufeng created it.
Fifteen years in measurement, most of it at Meta, where he led the global open-source measurement roadmap and represented Meta at industry events and standards bodies. He left in October 2026 and now advises measurement vendors and large advertisers through his own firm, Vectorial Advisory. On the day he takes the measurement slot with Arthur Anglade: pLTV is built on individual unit economics, mix models and most other measurement are aggregated, and today's ways of joining the two are naive enough to cap what pLTV can do.
Principal Measurement Strategy Specialist, Haus
What Haus is
An incrementality measurement company founded in 2021 by a former Google data scientist. Geo experiments, causal attribution, and a mix model built on top of those experiments rather than beside them, across billions of dollars of ad spend.
A marketing mix modelling specialist who has sat through more model review meetings than most people would volunteer for. His reason for joining Haus is the pattern he kept seeing in them: a marketer asks how far they should trust the model, and the answer is always to go and validate it with an experiment, which almost nobody runs well enough to settle the question.
On the day he and Gufeng take the measurement slot together. Not a finished answer, because the area is too new for one: where combining pLTV with aggregated measurement breaks today, and what that means for a UA lead deciding budgets.
Marketer, building his third startup · The opponent
Twenty years in go to market and growth at scale, most recently at monday.com and before that Taboola, and now building in applied AI. On the day he has no slides. He speaks for the client: after every talk he puts one hard question to the speaker, who gets two minutes to answer.
Product Lead, PvX Delta
What PvX is
A user acquisition financing firm for games and consumer apps, with cohort data from more than 600 companies. Delta is its AI agent for UA: every night it reads campaign, geo and creative data and flags ROAS opportunities with the root cause and a suggested action.
More than a decade in mobile UA and ad tech. He built Netmarble's first data-driven UA stack, including its LTV and churn prediction models, and ran more than $90M of spend across titles including Marvel Future Fight. He then founded Remake, grew it to $10M in annual revenue and sold it to Neptune, part of Krafton. Now he leads Delta. With Aniruddha he gives the builder's account: why UA teams need a copilot, nine months of technical and cost barriers and how they got past them, and why you might use Delta rather than build your own agent.
Lead Engineer, PvX Delta
He leads engineering on Delta. Before PvX he was an AI engineer at Maxim AI, which makes evaluation and observability tooling for teams shipping LLM products, and a data scientist at AB InBev, building customer and product segmentation across its European markets. On the day he takes the engineering side: how you get an agent to read every row of UA data and come back with something a UA manager will act on.
Interim Chief Product Officer, Paddle
What Paddle is
The merchant of record for more than 10,000 digital businesses, handling payments, billing, tax and compliance so software companies do not have to. It sits between the purchase and the advertiser's own data.
A former SaaS founder, he joined Paddle as a product manager and now leads its product and product design teams. His work spans how software gets sold: global payments and billing, app-to-web commerce and emerging business models. On the day: app-to-web, and how the web drives LTV. Who should see a web checkout, how to hand off from an in-app paywall without denting conversion, which payment methods produce higher-value subscribers, whether web subscribers renew differently, and how to measure one user across the whole funnel. Dos and don'ts, then one real example with the numbers.
Co-founder and CEO, Churney
Closes the hour on the objective discrepancy: why the platform is optimising for something your business does not value, and what a UA lead can do about it on Monday morning. Plus a new way to audit your data warehouse and ad accounts, so you know you can use pLTV properly before you start.
Co-founder and CPTO, Churney
Opens with Meta's new pLTV value optimisation: the pitfalls, what is working, and what you need to do differently if you have been doing pLTV some other way.
Why this session
Platforms optimise inside a seven day window. Your revenue arrives over months. So budget goes to cheap early converters every day, and the customers worth having get underbid.
Fixing that is two problems, not one. Designing the signal, and proving it worked. They usually sit in different teams. An hour on both, plus what LLMs and agentic workflows actually change underneath them.
60 minutes. Ten minutes per speaker, then two minutes of cross-examination.
Chris Jones, Churney.
Brian Brost, co-founder and CPTO, Churney.
Gufeng Zhou, creator of Robyn, and Arthur Anglade, Haus.
Lucas Lovell, Interim Chief Product Officer, Paddle.
Sam Nam and Aniruddha Chattopadhyay, PvX Partners.
Post them in the Q&A at any point in the hour. Chris reads the top three.
Noy Rotbart, co-founder and CEO, Churney.
After every talk, Amit Edelstein puts one hard question to the speaker on your behalf. Two minutes to answer.
Heads of growth, UA leads and performance marketers, plus the data and measurement people who get handed the modelling work. Useful whether you are evaluating, mid-build, or live and wondering why the results are not holding.
About this session
Churney is a Copenhagen research company founded by four PhDs. We build predicted long-term value signals and work alongside the ad platforms every day. This is educational, not a pitch. Much of what we cover you can do in house, and we will say which parts.
No. If you want one, ask afterwards.
The platforms have built the bidding. None of them predicts the value for you. They accept a number you send and optimise towards it, which means the model, the signal design and the upkeep are yours.
Good, and this will be useful to them. Most in-house builds do not fail on model quality. They fail on what gets sent, when, and what happens to it afterwards.
Yes, sent to everyone who registers.