Reliability
Надёжность
Defense Holds Depth 4 · Mine
low maintenance, minimal downtime
Moats it grows into
- #8 Brand and trust
Trust is the buyer's memory of your uptime; each quiet year adds to it and one loud failure resets it. The reliability record is the brand.
- #5 Switching costs
Switching means trading a known failure rate for an unknown one; the longer your record, the more that unknown costs, and the buyer stays.
Summary
Reliability is defense by the absence of incidents. The product works every time, needs little care, and the buyer stops checking. Over years that turns into trust that a rival's launch cannot buy, and into a switching cost: leaving means taking on the risk of the unknown again. Both moats are dug slowly and only with a clean record.
How it works
The mechanism is risk transfer. Every purchase carries a risk that the thing fails; a reliable product takes that risk off the buyer, and the buyer pays for it in loyalty, in price, and in not shopping around. The edge compounds through memory: one failure is remembered longer than a hundred successes, so a record with no failures is worth more than the sum of its parts. In practice reliability is engineering discipline that is expensive to keep and invisible when it works: redundancy, testing, slow release trains, saying no to features that add failure modes.
The second half is the switching cost. A buyer who has run your product for five years without an outage has to price in a rival's unknown failure rate; the rival cannot show a record it has not had time to build. This is why reliability favours the incumbent and why it stays a feature for years before it becomes a moat.
Why it holds — or only takes
Holds, twice. The trust half is #8: a reputation for not failing lives in the buyer's head and is renewed with every uneventful year, depth 4 because a rival cannot compress ten years of uptime into a launch. The lock half is #5: the cost of leaving includes the risk of the unknown; that moat is shallower, depth 2, and under attack from AI that lowers the cost of migration itself. Neither appears until the record is long enough to be a record; before that, reliability is a feature that a rival matches with the same discipline.
Do you have it
- Customers renew without a review call; nobody on their side owns the 'should we switch' question.
- Your status page shows years of green and procurement teams quote you as the safe choice.
- Support tickets are about usage, not outages; the incident channel is quiet for months.
- Buyers pay a premium over a functionally equal rival and say 'we know it works'.
How to dig it
- Publish uptime and incident write-ups; a record nobody can see builds no trust.
- Cut features that add failure modes; each new path is a new way to break the record.
- Slow the release train for the core; ship risky work behind flags to a small share.
- Run the boring engineering: backups proven by restore, failover rehearsed, on-call staffed.
- Offer an SLA with real credits once the record supports it; the promise is the product.
How it erodes
- One public outage at the wrong moment; the record resets and the premium goes with it.
- Reliability becomes table stakes as the platform underneath, cloud or hardware, makes everyone reliable.
- AI-assisted migration cuts the switching cost; the buyer can leave for a rival in a week.
- Growth adds failure modes faster than discipline removes them; the product becomes the incident.
- A rival buys the record: an incumbent with a longer history enters your segment.
Solo, without capital
A solo builder can be reliable this quarter: small surface, few dependencies, tested backups, an honest status page. What you cannot have is a long record; the trust moat needs years of it, and #8 is only partial for a solo builder. Start the clock now: the incident log you publish today is the asset a rival cannot back-date. The switching-cost half is yours in full, since #5 is solo yes: a customer who has run you a year without a scare will not move lightly.
What the matrix says of its moats
- #8 Brand and trust Solo ~ partly Capital ◐ medium AI ↑ AI-resistant Rent ~ partly
- #5 Switching costs Solo ✓ solo-reachable Capital ○ low AI ↓ under attack Rent ✗ not for sale
In nature, in business
- Nature
- —
- Business
- Zippo lighters, Toyota
Cases
- Toyota — Toyota's production line stops for every defect, and decades near the top of reliability surveys made the make the default answer to which car will not break.
- Zippo — Zippo repairs any lighter free for life; the guarantee is affordable only because the product rarely fails, and the guarantee is the proof.
- SQLite — SQLite ships with a test suite many times the size of the library and sits inside both major phone platforms, because nobody has to think about it.
Combines with
- Simplicity — Fewer parts, fewer failure modes; simplicity is the cheapest way to keep the incident log empty.
- Endurance — A record is only as long as the time you have been running; endurance is what gives reliability its years.
- Transparency — A published status page and honest post-mortems turn a private record into trust a buyer can check.
- Craftsmanship — The care that goes into each unit is the care that keeps it from failing in the field.
Fights with
- Iteration — Every release is a chance to break the record; a short cycle and a clean incident log pull against each other.
- First-mover — Shipping first means shipping untested; the record that reliability needs starts only once the rush is over.