Owning growth metrics since 2018. Now as a fractional partner.
I’m Douglas Wang, a growth product manager. Teams call me when a metric has been flat for two quarters and everyone has a theory. I pick one, test it, and write down what the numbers do and don’t cover.
Working with global teamsUS LLC · invoices in USD
01 · TRACK RECORD
Stuck metrics, moved.
IndependentFractional and advisory, 2025 –
IndependentProduct consultant
2025 –Growth mandatesFour teams, one brief each: move a specific metric
TaxActSenior PM, Activation
2024 – 25
FoundSenior PM, Growth
2023 – 24
Thirty MadisonPM, Growth & New Ventures
2022 – 23
MatchPM, Monetisation
2021 – 22Where I learned itThree years scaling at HelloFresh
HelloFreshProduct Manager
2018 – 2102 · EDUCATION & MENTORSHIP
Passing it on.
MentorshipThe part of the job I’d do for free
Pathrise
Thirty Madison
Mentored PMs on the growth team
Independent
Stood up an experimentation function and coached three PMs to run it
EducationWhere the frameworks came from
Reforge · Monetisation & Pricing
Program
Reforge · Consumer Subscription Growth
Program
BBA, Management Information Systems
The University of Texas at Austin
03 · BUILDING WITH AI
Specs and prototypes in hours, not sprints.
AI does the drafting, so the team spends its time on the decision. The clearest example is this site: designed, written and built with Claude. The diary covers what I let the model decide, and what I didn’t.
SHIPPED
ship-team ↗The /ship skill I build with. Nine Claude Code agents that scope, build, review and test a change like an engineering team, then open a draft PR for me to approve. Open source.
The learning loop (/retro) ↗The skill that makes the agents better over time. After a PR lands, /retro reads my review corrections and proposes edits to the agent rules, one at a time, and flags rules that have gone stale. Nothing changes until I approve it. Open source.
BUILT WITH
Claude
ChatGPT
Cursor
Sentry
PostHog
Vercel
Lovable
04 · HOW I WORK
Every engagement is different. The write-up isn’t.
- 01SituationWhat’s actually stuck, in one sentence the team agrees with. Not the symptom the dashboard shows — the thing underneath it.
- 02QuestionThe one question worth answering first, with the prediction written before the data comes back. If I can’t say what would prove me wrong, it isn’t a question yet.
- 03DecisionA written recommendation with what’s kept, what’s cut, and who calls it. One meeting. Then it ships.
- 04Result & limitThe number, and where the claim stops. Projected or realised. Which segment. What didn’t move. That last part is what makes the next question good.
The first two weeks
An evidence read, not a roadmap. Your funnel, your cohorts, your past tests. It ends in a one-page memo: the question I think is worth answering, and what would prove me wrong.
What I insist on
Decisions written down. One owner per test. The result reported with its limit attached. I’ll tell you when I don’t think I’m the right intervention.
05 · OFF THE CLOCK
Coffee and mats.
The same loop as the work, at smaller stakes: a rep, an honest note, adjust the next one.
25-minute fit call
Bring the metric that isn’t moving. No deck.