1 min lesson
When to invest vs. accept a directional read
Work through this case: "In Cursor's developer-led motion, why does a last-touch model tend to over-credit a partnership or sales touch and what's the most honest way to correct for it?" Say what you would do and why.
Step 1 of 2
When to invest vs. accept a directional readthe pragmatic call
Better attribution costs engineering time and sometimes deliberately withheld pipeline. Spend it where the decision is expensive and reversible-only-at-cost. For a small experiment, a directional read is enough; for a multi-quarter budget allocation, a holdout earns its cost.
When asked "how much revenue did the program drive?", don't pick a model and quote a number. Say which model you'd use, name its bias out loud, then propose the cheapest holdout that would turn correlation into a causal read. That sequence is exactly the truth-seeking the founder round screens for.
"Last-touch says the program sourced $1.2M, but it can't see the product usage that preceded most of those deals, so treat that as an upper bound. A geo holdout next quarter would tell us the incremental number. Until then I'd report it as directional, not sourced."
The pressure in the room is to give a confident single number. Giving one you can't defend is the fastest way to fail a truth-seeking screen. Quantify the uncertainty instead; "directionally positive, here's the read I trust" beats a precise figure built on a model you'd disown under questioning.