Building Fare 1 in public.

A private hire company normally needs an office: someone on the phones, someone watching the map, someone chasing drivers. We are building Fare 1 so that one person on shift can run all of it from a phone.

That only works if the system does the routine work on its own and asks a human only when a decision is needed. So every feature gets the same test: does this remove a job from the person on shift, or add one?

The next few posts cover the pieces that passed that test: dispatch, flights, late drivers, and alerts.

If every notification makes the same sound, you have to pick up the phone every time. Most of the time it's nothing urgent, and that is how the urgent one gets missed.

In our admin app each kind of event has its own sound. A new booking, a driver running late, a customer message and an emergency all sound different, with the serious ones loud and the routine ones soft.

It sounds like a small thing. In practice it means the person on shift knows what happened before the screen is on, and can ignore the right things.

A pre-booked job does not wait for someone to remember it. Ahead of the pickup time the system offers it to the nearest available drivers, and if nobody accepts, it widens the search step by step.

A person hears about it only when the system has run out of options. Then the office gets a no-driver alert while there is still time to fix it by hand.

Most jobs never reach that point, so most jobs need no dispatcher at all.

Airport pickups used to mean someone refreshing a flight tracker and phoning the driver when a plane was late.

Now the booking is linked to the flight. When the arrival time changes, the pickup time moves with it, and the driver and the customer are told when the change is big enough to matter.

The driver does not sit at arrivals for an hour, the customer does not land to an empty kerb, and nobody in the office had to do anything.

Traffic happens. The old fix was a customer calling to ask where the car is, then a scramble to find a replacement.

Now the system watches each driver's arrival time on the way to a pickup. If a driver is going to be late and another driver can get there sooner, it offers that driver the job. If they accept, the job moves over and the first driver is released.

The office is told what happened, after it has already been handled.

Off-the-shelf dispatch software is built for an average taxi firm, and it assumes you have an office full of people to work around its gaps.

We built Fare 1's booking platform, apps and admin system in-house. It was slower at the start. But every time the same manual task shows up twice, we can turn it into a feature, and that is not possible with software you rent.

The automation in the posts above came from that one decision.

Over half of Fare 1's search impressions from the last 3 months came in the last 28 days: 12K of 21.6K. Average position moved from 14.8 to 11.7.

We're still on the edge of page one, and a 0.7% CTR says there's plenty left to fix. But the technical groundwork is starting to show up in the data.

Next target: page one, and turning impressions into clicks.

This site used to be a landing page. Now the front page is a running log: short notes on what we are building, what the numbers say, and what went wrong.

No algorithm, no account needed. If you use an RSS reader, the feed link is at the top of the page.