In almost every business I have worked in, the most effective growth work was making something happen every time it was supposed to. It is also the work that gets the least budget, because its effect shows up slowly. This is the case for treating reliability as the strategy.
The evidence from my own work
| Business | The reliability problem | What it was worth |
|---|---|---|
| ZenCabs | Drivers cancelled four in ten accepted rides | Driver-side cancellations fell from 40% to 6%, and 27% of users became monthly active riders |
| Shaadi Mangalam | A funnel that lost almost every lead before payment | Conversion rose from 1.4% to 4% with pricing, sales and website fixes, and monthly revenue grew 4x |
| Restaurants, through Cairros | Waits for the menu, the order, the food and the bill | Revenue grew 30% with menu and table-turnaround work |
| This site | Free AI models that fail, stall or leak | Fallbacks mean a visitor always gets an answer |
The ZenCabs row is the clearest. In Jammu the alternative to an app is a phone call to a driver you already know, so a rider let down twice goes back to making that call. Bringing driver-side cancellations down with 10+ standard procedures across onboarding and dispatch removed a reason people were leaving, which no discount addresses.
Why it gets underpriced
A discount is easy to measure: it goes out on Monday and bookings rise by Friday. Reliability work shows up as customers who did not leave, which appears weeks later in a retention chart that few people watch. So budgets flow to what shows up fastest, and the business ends up paying to acquire customers it keeps losing to its own failures.
The two also build opposite expectations, which is the channel my Push and Absorb framework flags as the most expensive. A discount teaches customers to wait for the next one. Reliability teaches them to stop checking alternatives, and a competitor then has to match it before a discount can even get them considered.
How to treat reliability as the strategy
- Define a failure in the customer's terms: a cancelled ride, an unanswered enquiry, a machine shown as available that was not.
- Measure the failure rate every week, split by cause.
- Give each cause one owner and a standard procedure.
- Review it in the same weekly meeting as growth, so it competes for attention on equal terms.
- Tell customers. A published promise that you keep is a marketing asset a discount cannot buy.
The weekly review I describe in the weekly review that took driver-side cancellations from 40% to 6% is where most of this happened at ZenCabs.
Where it matters most
Reliability pays most where customers have a familiar fallback, where a single failure is costly, or where the purchase depends on trust. That describes most services outside India's largest cities, which is part of my thesis on building beyond the metros. It also describes AI products, where one visible failure can undo a month of good answers, which I learned building the AI clone on this site.
