Running four AI products at once at Xcellerant Ventures, the easiest trap to fall into is letting each one adopt its own, locally sensible north-star metric. FastRead.io optimizes for finished chapters, Myautobio for completed life stories, Formulai for engagement with a health profile. All defensible in isolation. All quietly corrosive to the portfolio, because they make every roadmap trade-off a local optimization with no shared language for comparing trade-offs across products.
What "one north star" actually means across different products
It doesn't mean forcing four different products to share one literal metric. "Finished chapters" means nothing for a scheduling tool. It means every product's locally relevant metric has to answer the same underlying question: did a user reach an outcome they'd pay for again. That's the portfolio-level north star (repeatable, paid value delivered), and each product translates it into its own currency.
A shared metric across different products isn't a number everyone reports the same way. It's a question everyone is required to answer the same way, in whatever units their product actually produces.
Why this mattered more than any single feature decision
With a shared underlying question, prioritization conversations across the portfolio stopped being about whose product's roadmap was more impressive this quarter and started being about which product's next bet most increased repeatable, paid value right now, a comparison a weighted scorecard never resolves because it's a strategic judgment, not an arithmetic one. Some quarters that favored FastRead's onboarding rebuild. Some favored Formulai's retention work. The north star didn't make the comparisons easy; it made them comparable at all, which they hadn't been before.
80%+ of new features across the portfolio were adopted by users at launch, a direct result of every feature, across four very different products, having to justify itself against the same underlying question before it got built.
Where the discipline almost broke
The hardest moment wasn't choosing the north star. It was holding it when a product lead, correctly, pointed out that their product's healthiest near-term move looked bad against the shared question. Clintwin/Medsim AI's early value is pedagogical trust, which doesn't convert to "paid value" on the same timeline as a consumer tool. The answer wasn't abandoning the shared question. It was being honest that different products sit at different distances from it, and building that distance explicitly into how we compared them, rather than pretending every product should show the same signal on the same calendar.
The test for whether your portfolio has a real north star
Ask any two product leads in your portfolio to justify their next quarter's priority in one sentence. If the sentences use the same underlying logic, translated into different units, you have a north star. If they use entirely different logics that happen to share a dashboard, you have four products politely coexisting, not a portfolio strategy.