Matthew Mamet

@msmamet
8 Followers
36 Following
131 Posts
Fractional CPO/CGO | Built $200M revenue engines at TripAdvisor and EverQuote | Now advising B2C & marketplace platforms
My SiteMatthewMamet.com
Matthew Mamet2h ago

I was on a call with a leader who grew revenue per user from $3 to $17 at a large edtech company. Eighteen months.

First thing she did: asked every room, "Do you want me to match the pace or set the pace?" 95% said set it. She took one product and launched five more. Shifted to platform selling.

I have started using that question myself. Match the pace means they want help not disruption. Set the pace means they want change.

Ask before you push. The answer tells you how far you can go.

Show threadMatthew Mamet2h ago
@mariusz Exactly what I'm seeing in the companies I work with. They've removed every obstacle to shipping and now face a different problem: the org optimizes for velocity when the real scarcity is judgment. The product leaders who adapt early are reorganizing around direction and editorial control, not throughput.
Matthew Mamet1d ago

A media company hired a CPO from Big Tech to go AI-first. His all-hands showcased an AI project. Behind the scenes, one person built it in spare time with no resources.

That is not an AI strategy. That is a press release.

Meanwhile every A/B test result lives in PowerPoint decks on SharePoint. The first real AI win is not agents. It is feeding those decks to Claude and producing a two-page report in 24 hours. One person. No budget.

Find the PowerPoint graveyard.

Show threadMatthew Mamet1d ago
The 18 month timeline tells you more about the person making the claim than about the technology. Every executive I advise who has tried to replace a team with AI discovered the same thing within weeks: the tool cannot do the job because nobody in the organization actually documented what the job was. The bottleneck was never the worker. It was the fact that institutional knowledge lived in one person's head and was never written down.
Show threadMatthew Mamet2d ago
The companies I advise are living this in real time. The executives most excited about AI headcount reduction are the same ones who have never mapped what their teams actually do. They see a job title on a spreadsheet and assume a chatbot can replace it. Meanwhile the one person who used Claude to summarize six months of A/B test results in an afternoon gets no recognition because that does not look like a transformation initiative. The psychosis runs in both directions.
Matthew Mamet2d ago

A client calls them "content affiliates." They are SEM lead-gen operations bidding on the company's own keywords.

I saw this at EverQuote. Affiliates bought our keywords, piped garbage leads, collected a CPL. Progressive pulled out because quality tanked.

Two kinds of affiliates. SEM affiliates sell you back your own customers. SEO affiliates like NerdWallet build trust, creating demand that did not exist.

If your program is mostly SEM, you are paying someone to compete with you.

Show threadMatthew Mamet3d ago
@simon The underappreciated audience for this is people who are not developers. I run a fractional advisory practice and use coding agents daily to analyze meeting transcripts, scrape data, and build reports. Six months ago I would have needed an analyst for every one of those tasks.
Show threadMatthew Mamet5d ago
@davidculley The volume is the tell. 19k lines in a single PR means nobody is reviewing anything, regardless of how it was written. The PRs that cause the worst production incidents are always the large ones waved through because the author was senior or the deadline was real. AI makes the volume problem exponential. The answer isn't "review harder," it's shipping smaller so humans can actually understand what they're approving.
Matthew Mamet6d ago

A strategy only exists when it excludes something.

"We serve enterprise and SMB across North America with a best-in-class experience" is not a strategy. It is a market with adjectives. Any competitor could sign that document.

Most product strategy fails for one reason: it is not actually a choice.

matthewmamet.com/blog/product-strategy/

Show threadMatthew Mamet6d ago
@tijn The output looks finished so nobody questions whether it actually works well. At every company I've scaled, the most expensive bugs were the ones that technically functioned. They passed QA, they shipped, and they quietly degraded the experience until a human finally noticed the numbers. AI generated code that "kind of works" is the same trap at a faster clock speed.