How should companies measure customer acquisition cost given the difficulty of attributing conversions to specific channels?

Harry Stebbings · published Feb 15, 2023

Ambiguity in conversion attribution leads teams to misidentify what caused an uptick and then take the wrong actions

He cites spending more on Facebook ads on a Friday evening that happened to be a World Cup evening, then repeating the spend the next Friday with different results

34:41 20Growth: Top Five Growth Lessons Scaling Stitchfix to IPO, How to Master the Art of Paid Marketing, Why CAC/LTV is a BS Metric & How To Use Payback Period as an Alternative to CAC/LTV with Mike Duboe, Partner @ Greylock

Harry Stebbings · published Apr 5, 2023 · hedged

Attribution is almost impossible because conversion takes on average about seven brand touch points, making it unknowable which touch caused the sale

With seven touch points you can't tell whether the Facebook ad at touch four or the Instagram picture at touch seven drove the conversion

Scope: cites remembered 'scientific data'

24:34 20VC: The Memo: Scaling to $600M Revenues with No Venture Funding, The Most In Detail Breakdown of Consumer Subscription Unit Economics & Why D2C and Consumer Subscription is Not a VC Backable Model with Mike Salguero, Founder @ ButcherBox

Mike Duboe · published Feb 15, 2023

Last-click attribution and expensive black-box attribution models draw wrong conclusions about channel performance; incrementality testing via holdouts should be the north star

Last click hands 100% of credit to one channel — usually search because it's the highest-intent — and impact is rarely that direct; only holdout tests that incur real cost reveal true incremental CPA, which you can then apply as multipliers back into your attribution model

Scope: holdouts incur real cost

39:08 20Growth: Top Five Growth Lessons Scaling Stitchfix to IPO, How to Master the Art of Paid Marketing, Why CAC/LTV is a BS Metric & How To Use Payback Period as an Alternative to CAC/LTV with Mike Duboe, Partner @ Greylock

Matteo Franceschetti · published Sep 12, 2026

You cannot trust platform-reported CAC — Meta's reported CAC is roughly 20% below the true CAC — so brands need their own models plus regular geo incrementality tests, turning channels off in matched states to measure lost revenue

Platforms over-attribute conversions to themselves; only turning a channel off in one of two similar geos reveals its true incremental value

Scope: tests roughly every six months

8:20 20VC: 7 Predictions for How AI Changes the World: Labour, Engineering, Social Media, GrokBots Buying Cybercabs and more with Matteo Franceschetti, Co-Founder @ Eight Sleep

Mike Salguero · published Apr 5, 2023

Founders should calculate CPA by dividing total monthly marketing spend by number of signups rather than building per-channel attribution

A truly functioning attribution model is extremely hard to build because customers bounce between Instagram, search and direct before converting, so the effort of measuring it day in and day out isn't worth it

23:55 Blended total spend over signups beats granular per channel attribution

Mike Duboe · published Feb 15, 2023

The question of the right organic/paid mix is itself flawed, because attributing a single acquisition source to a given user is usually a fallacy

Incrementality, not single-source attribution, is the right frame

36:37 The organic paid mix question is flawed since single source attribution is usually a fallacy favor incrementality

Matteo Franceschetti · published Sep 12, 2026

Incrementality testing can't be done from day one, will involve multiple mistakes, and must be repeated continuously as you add more channels

You will screw it up multiple times, and adding channels changes each channel's incremental contribution

10:20 Incrementality testing requires continuous iterative refinement not a one time setup