In Kose’s first seven weeks we completed 724 rides with 60 drivers and 147 riders. Small numbers. I have looked at them more times than is reasonable.
The thing that surprised me sat outside the numbers. When riders talked about the app at all, almost nobody mentioned the interface. They mentioned whether the car came.
That sounds too obvious to write down. I think it is the whole business.
The market pays a premium you did not price
In an economy where the power is uncertain, the currency moves, the supplier delivers late, and the government portal is down, the ordinary experience of transacting is one of low-grade anxiety. Everybody plans around the possibility that the thing will not happen.
A venture that simply does what it said it would do competes on relief rather than on features.
This is a strange kind of advantage because it does not look like one. Nobody writes a case study about a company whose deliveries arrived. You cannot put it in a pitch deck without sounding like you have nothing else to say.
But watch what customers actually do. They pay more, they wait longer, and they forgive more, for the supplier who is predictable. In a market full of variance, low variance is a product.
Why better-funded competitors keep failing to buy it
Reliability cannot be bought in a round. It accumulates.
It accumulates out of decisions that individually look too small to matter: the driver incentive you set at a level you can actually sustain, the feature you refused to ship because you could not support it, the customer you turned away because you were already at capacity.
Each of those decisions costs you something visible now in exchange for something invisible later. That trade is hard to make when you are being measured on growth, which is why well-capitalised competitors so often lose it. They can outspend you on acquisition. They cannot outspend you on having already done the boring thing for two years.
When we cut the Kose driver incentive from twenty five dollars a week to ten, our supply numbers looked worse for a period. What we were buying was an incentive we could keep paying. A promise you have to withdraw in month four does more damage than a smaller promise you keep for three years.
Reliability is measured at the tail, not the average
Here is the part most founders get wrong, and I got wrong for a long time.
You do not experience a supplier as their average. You experience them as their worst recent instance. One catastrophic failure resets a year of competence.
So the number that matters is your ninetieth percentile rather than your average delivery time. What happens on your bad days, because your customer’s memory is built almost entirely out of bad days.
At Cafe Oldrock we reconcile every night. Not weekly, not when something looks wrong. Every night, whether or not anything happened. The value is not in the average night, when the numbers agree and the exercise seems pointless. The value is in the fourth Thursday, when they do not, and we know within hours instead of within a month.
Most of what a reliability discipline buys you is a shorter distance between a problem starting and a problem being seen.
A test you can run this week
Write down every promise your venture makes, including the ones you never said out loud.
Some are explicit: delivery in three days, support response within a day, the price on the website. Most are implicit: that the app will load, that the person who answers knows your account, that the invoice will be right.
Now, for each one, answer honestly: what happens on the bad day? Not the average day. The day two people are sick and the supplier missed and the system is slow.
The promises that collapse on the bad day are the ones you are quietly not making. Either build the capacity to keep them, or stop making them. Both are respectable. Making a promise you keep seventy percent of the time is the option that costs you the most, because you are paying the full price of the promise and collecting none of the trust.
The honest cost
This is unrewarding work.
Reliability generates no story. There is no launch, no announcement, no moment where the thing you have been doing quietly for two years becomes visible. The recognition, if it arrives, arrives as a customer who has stopped comparison shopping and cannot quite tell you why.
You will also watch flashier competitors raise more, grow faster, and get written about, while you are having an argument about whether you can genuinely support a feature. Some of those competitors will still be there in three years. Most will not, and the reasons will look obvious in hindsight and did not look obvious at the time.
I do not have a way to make this feel better. I only know that in this market, the ventures that are still standing after year five are almost never the ones that were most interesting in year one. They are the ones whose customers stopped worrying about them.
