A moat you can't A/B test.
Software defends with metrics you can read weekly. Deep tech defends with physics, certification, and accumulated proof — moats that move on a different clock.
By Owen E. H. MeyerJuly 2, 20264 min read
Software diligence has instruments for measuring a moat. Cohort retention shows whether users stay, network density shows whether the product gets stronger as it grows, and churn puts a number on switching costs. Run the experiment, read the dashboard, size the defensibility. Point those same instruments at a deep-tech company and they return nothing — no cohorts, no churn, often no revenue for years. The tempting conclusion is that there's no moat to measure.
The defensibility is real; it just lives somewhere the dashboard can't see. It's a manufacturing process that took five years to make work, which a rival will need roughly the same five years to replicate. It's a certification, a safety case, or a flight heritage that a competitor cannot skip no matter how well funded. It's patent coverage on the one narrow step that makes the physics economic, and an operating dataset — cycles run, miles flown, batches produced — that no amount of money can synthesize retroactively. None of it can be A/B tested. All of it takes years to build, and years to breach.
Illegible isn't weak
Because these moats don't render in a metrics dashboard, capital trained on software tends to underrate them — and to overrate the legible kind, which can vanish fast. A feature gets copied in a quarter; a network tips when a better one arrives; a switching cost erodes with one clean migration tool. The deep-tech moat fails differently: slowly, visibly, and with warning, because the challenger's progress is as physical and public as the incumbent's was.
A qualified production line doesn't churn.
Evaluating this kind of defensibility means swapping metric-denominated questions for time-denominated ones. What would a competitor with unlimited money still need years to do? What does the regulator, or the customer's qualification process, actually require before a rival can be considered? What does the accumulated operating data allow this company to do that a fresh entrant, starting today, cannot? The answers won't fit in a dashboard. They fit in a diligence memo written by someone who understands the engineering.
“Moat” was a metaphor about time before it became a metric. The water mattered less than the months a siege would cost. Deep tech returns the word to its original meaning: the width of the moat is the number of years it takes to cross, and some of the widest ones surround companies whose dashboards, for now, have nothing to show.