The 80% Rebuild Test
A single counterfactual, applied honestly, separates what a company owns from what it merely built. Most of the value is in the twenty percent — if it is anywhere.
Here is a question worth putting to any software business, including one you already own.
If a competent team with frontier AI tools could recreate 80% of the customer-visible functionality within 12 to 18 months, what exactly is the buyer paying for?
It is deliberately a counterfactual, not a forecast. Nobody has to believe a competitor will actually do it. The exercise is useful because of what it forces into the open: once the reproducible portion of the product is set aside, whatever remains has to justify the entire enterprise value on its own.
Why 80%, and why customer-visible
The threshold is not precise and does not need to be. It is set where it is because 80% is roughly the point at which a buyer would seriously evaluate the alternative — not enough to be a drop-in replacement, enough to start a conversation about price.
“Customer-visible” is the more important half of the phrase. It excludes everything the customer neither sees nor pays for: internal architecture, engineering craft, the elegance of the data model. Those things matter to whoever maintains the system. They do not appear in a competitive comparison, and they are not what is being bought.
What legitimately survives
A short list of answers holds up under pressure. Every one of them is something a competitor cannot obtain by building.
- Distribution — the relationships and channels through which the product is actually sold
- Data that cannot be reconstructed within a competitive timeframe
- Contracts, particularly multi-year ones with real termination cost
- Workflow embedment: the product is the process, not a tool used within it
- Integrations that took years of counterparty cooperation to obtain
- Regulatory approval, certification, or accreditation with a queue in front of it
- Installed base and the service network attached to it
- Brand, where it functions as a licence to be considered at all
- Network effects that are genuinely two-sided
- Physical assets and the operations around them
Notice what is absent. “Superior technology” is not on the list, because that is precisely the thing the counterfactual assumes away. Neither is “our team”, unless the team is contractually bound and genuinely irreplaceable — which, honestly assessed, is rarer than it is claimed.
How the test fails in practice
Three failure modes come up repeatedly, and each is worth naming in the room.
Answering with quality instead of defensibility
A management team asked what a competitor could not rebuild will describe what makes their product good. Those are different questions. Good is not the same as hard to reproduce, and after several years of investment the two feel identical from the inside.
Counting the wrong twenty percent
The surviving 20% must be something customers pay for. A sophisticated permission model that no buyer has ever asked about is not the moat, however difficult it was to build.
Assuming the rebuild has to be complete
A competitor does not need feature parity. They need enough to be evaluated, plus a lower price. Category leaders have been displaced by products that did substantially less, and this is more likely, not less, when building is cheap.
Using it in a transaction
The test earns its place in diligence when it is applied concretely rather than rhetorically. That means naming the reconstruction: which team, which tools, which twelve months, which parts they would get wrong and which they would get right on the first attempt.
The points at which that reconstruction fails are the moat, and they should be describable in a sentence each. If the exercise produces only general statements about accumulated know-how, that is itself the finding.
The value that survives should then be roughly consistent with the price. Where a purchase price implies a moat that the rebuild test cannot locate, the difference is being paid for something — usually the assumption that the last five years continue for another five.
You are not buying the target’s current product. You are buying its future ability to defend cash flow.
The rebuild test asks whether the product can be copied. The related question — whether the category can be absorbed by someone who does not need to copy it at all — is the hyperscaler test.