The Dogfooding Boundary Rule
Use your own product walkthrough to generate the issue list — never to prioritize it.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 91%
Antin's contrarian correction to the popular advice that PMs should dogfood and do product walkthroughs. Dogfooding is genuinely valuable for one job — surfacing a candidate list of problems, the kind of issues you only need a pulse to notice. It is actively dangerous for the next job — deciding which of those problems matter, how much, and for whom — because that requires context of use, priorities, and constraints you do not have and cannot imagine from your own usage. The rule draws a hard line between the two activities.
Origin
Judd Antin, offered explicitly as a contrarian response to prior guests on Lenny's Podcast who had endorsed PM product-walkthroughs as a primary decision input.
Core principles
- 01You are nothing like the user, in ways that bias you and that you cannot recognise from the inside.
- 02Some problems you only need a pulse to spot; most good PMs have a pulse.
- 03Context of use, priorities, and constraints cannot be imagined from your own usage.
- 04Generation is safe; prioritization is where your intuition betrays you.
- 05Dogfooding is a source of hypotheses, not a source of rankings.
How to run it
- 1
Dogfood and walk through the product deliberately
Everyone on the team should use their own product. Do the walkthrough and write down every potential issue you hit.
- 2
Stop at the list
Treat the output strictly as a list of candidate issues — hypotheses to be tested, not defects ranked by importance.
Pro tip Label the artefact 'candidate issues' rather than 'bugs' or 'priorities' so it cannot silently become a roadmap.
Watch out The moment you attach a severity or a 'for whom' to an item from your own walkthrough, you are back in intuition territory.
- 3
Hand prioritization to evidence
Determining which issues matter, how much, and for which users requires research, data, or both. Be extremely wary of relying on your own opinion, expertise, or intuition here.
Pro tip This is exactly the micro-research slot — pointed, fast (often 48 hours), and directly tied to funnel metrics.
In the wild
No Airbnb PM dogfooding the flow would have flagged the CTA copy as scary — they knew what the button did. Research on real users revealed the fear that the click would initiate a purchase.
→ A seven-character copy change lifted conversion ~1%. The problem was invisible from the inside of the team's own usage.
Airbnb theorized hosts would be more comfortable with instant booking if guests connected via Facebook. Behind a one-way mirror in Paris, hosts turned out to be strongly unwilling to give Facebook any data — far ahead of US hosts on privacy.
→ The team's own intuition about what would reassure hosts was inverted by the actual users; France and Germany became Airbnb's bellwethers for privacy sentiment.
Common mistakes
Treating your walkthrough list as a prioritized backlog
The severity and frequency of an issue depend on context of use, priorities, and constraints you do not share with your users — so your ranking encodes your biases, not theirs.
Concluding you don't need research because you use the product
Anyone can talk to a user or click through a flow; that doesn't constitute insight. One user can be powerful and one user can be idiosyncratic — the skill is knowing which.
Is it for you?
Best for
PMs and founders who run regular product walkthroughs and feed the findings straight into the roadmap
Not ideal for
Developer-tools or internal-tools teams where the builders genuinely are the target user in the same context of use
From the transcript
“is realizing you are nothing like the user you are nothing Li them in ways that will bias the way you think about what's good…”
“you should definitely dog food your own product doing product walkthroughs to identify lists of potential issues is a great thing to do prioritizing that…”
“the thing is anyone can talk to a user that does not constitute research or insights work”
From the episode
The UX research reckoning is here
Judd Antin (Airbnb, Meta)