Introduction — Optimising in the Dark

Six thousand eight hundred rides, a rating that moved, and no record of any of them.

Six thousand eight hundred rides and no record of any of them

Sasha had been driving for four years and was, by every measure the trade offers, good at it.

Six thousand eight hundred completed rides. A rating of 4.91. No complaints that had gone anywhere, no accidents, no gaps in the licensing, a clean car. If you had asked another driver about Sasha they would have said: solid, does the airport run, does not mess about.

⚠ And when the rating slipped from 4.91 to 4.87 over about three weeks, Sasha had no way at all to find out why.

That is not a failure of effort. Sasha tried extremely hard. Water. Mints. Two charging cables in the back seat. A new air freshener, then no air freshener, because somebody might have hated the new one. Silence unless spoken to, then a single friendly sentence at the start of each ride, because silence might have read as rudeness. The car cleaned twice as often. Motorway route, then the shorter route, then back again.

⚠ Fifteen changes in a month, made simultaneously, in response to a signal that identified no ride, no passenger, no date, and no reason.

The rating recovered. Sasha does not know why, cannot know why, and has no way of knowing whether any of the fifteen changes did anything at all — or whether four hundredths of a point moved because two passengers out of six thousand eight hundred happened to have a bad evening.

⚠ A month of careful, expensive, well-intentioned optimisation, performed entirely in the dark.

Then, a fortnight later, a passenger made an accusation about something that did not happen.

⚠ The account was restricted for nine days while it was investigated. It was resolved. Sasha was reinstated and nothing further came of it — and that is genuinely the good version of this story.

But in those nine days Sasha sat down to write out an account of that ride, and found there was nothing to write it from.

⚠ No dashcam. No log. Not one written note about any of six thousand eight hundred rides.

⚠ Four years of full working weeks, and the only thing Sasha had accumulated was a decimal that could move without explanation, could not be interrogated, and belonged to a platform.

Optimising in the dark

⚠ This is worth being precise about, because it is the specific structural problem of this trade and almost nobody names it.

Feedback here has four properties, and each one removes something.

⚠ It is anonymous. You do not know who gave it. You cannot ask.

⚠ It is delayed. A rating may be given hours after a ride, from a sofa, in a completely different mood.

⚠ It is aggregated. What reaches you is an average across hundreds of rides, in which a single one is worth a rounding error.

⚠ It is non-specific. No reason is attached. Where a comment exists, it is frequently a category rather than an account.

⚠ Feedback with all four properties cannot teach you anything, because learning requires knowing which behaviour produced which outcome — and every one of those four properties severs that link.

What drivers do instead.

⚠ They change everything at once, observe an aggregate figure move, and attribute the movement to whichever change they liked best. Which is not learning. It is superstition with a spreadsheet, and this trade is full of it.

The uncomfortable part.

⚠ Much of what moves a rating is not about you at all. The passenger's evening. Traffic. The price they paid. Whether the app estimated four minutes and you took nine because of a closed road. Whether they were rated badly themselves last week. A driver taking a rating drop personally is reading noise as a message.

⚠ Which does not mean the rating does not matter. It matters enormously, because it can end your access to the work. Chapter 9.2.

⚠ It means you cannot use it to improve — and so you need something else to improve from.

Why the transaction closes before you can fix anything

Compare it to almost any other work.

A plumber who leaves a tap dripping gets a phone call. A pet sitter who worries a client gets a message they can answer. A delivery driver who arrives with a spilled drink can say so at the door, before it becomes a complaint.

⚠ A rideshare driver's transaction ends when a door closes. There is no follow-up, no relationship, no second visit, and no channel through which anything can be addressed.

Three consequences.

⚠ One: everything has to be right during the ride, because there is no afterwards. Which is what rule three is about — the decisions that cannot be made once the car is moving. Chapter 6.

⚠ Two: your account of what happened has to exist at the time, because you will not be able to reconstruct it. Six thousand eight hundred rides are indistinguishable in memory after a fortnight. Chapter 5.3.

⚠ Three: the only person who can produce that account is you. The passenger is not going to. The platform holds a route and a timestamp and nothing about what was said or done. Chapter 24.1.

⚠ So the whole of this book's practical argument comes down to one line: the rating is the only thing the trade gives you, it cannot be interrogated and is not yours, and the only asset you can actually build is a record of what happened.

The four rules

Four sentences, deliberately short.

⚠ Rule 1: You can end any ride, at any point, and you never owe anybody a reason.

⚠ First because it is the one drivers do not believe they have.

Everything in the working arrangement pushes against it — cancellation rates, acceptance metrics, the fare you lose, the confrontation you avoid by continuing, and the entirely reasonable feeling that ending a ride is a dramatic act reserved for dramatic circumstances.

⚠ It is not. It is an ordinary professional action available at any moment, and a year of night driving in which it was never once used is not a lucky year. Chapter 4.5.

⚠ Rule 2: You are the only professional in the car, and the only person who can record what happened.

⚠ Two accounts will exist of any disputed ride and only one of them will be yours. Build it as you go. Chapter 5.

⚠ Rule 3: Nobody gets in until the car is ready for them.

⚠ Belts, doors, capacity, condition, child seats. Every one is decided before the ride, because it cannot be decided during it. Chapter 6.

⚠ Rule 4: You count the hours you gave, not the hours you were paid for.

⚠ You will be online far longer than you are engaged, and the difference is the whole economics of the job. Chapter 7.

How to read this book

Read the introduction and Chapters 4 to 7 first, in order.

⚠ Then read Chapter 15 before your next shift. It is licensing and insurance, and carrying passengers for hire is regulated almost everywhere, frequently with requirements specific to you personally. That chapter is where the stop-work items live.

⚠ Then Chapter 26, which builds your cost per mile and per engaged hour, and Chapter 7, which is what to do with them.

The rest can be read in order or used as reference. Part Two is the vehicle, compliance and records. Part Three is the work. Part Four is systems and money. Part Five is growth and responsible AI.

The resources. Sixty-eight of them, in six parts. Every financial figure is blank and in U.S. dollars.

The prompts. One hundred and forty-eight, four per chapter. ⚠ Not one contains a passenger's name, pickup or destination, and not one is intended for use while driving.

⚠ The warning markers throughout are load-bearing. They flag genuine risk — on the road, in an enclosed space with a stranger, to your licence, or to your ability to answer a question later.

⚠ One last thing about Sasha, because it is easy to read the story as being about somebody who was disorganised.

Sasha was not. Sasha turned up, kept the car clean, held a licence, drove carefully, and treated people decently for four years — and by every standard the trade actually uses, did an excellent job.

⚠ The problem was that the trade's standard is a number nobody can explain, and Sasha had spent four years trying to move it instead of building the one thing that would have been worth having.

⚠ Which takes about forty seconds a ride.

©2026 James Henderson / https://localhandyman.work

Chapter 1 — Understanding the Industry and What You Are Actually Selling

You are selling a vehicle, your whole time, and your presence in a small space with strangers.

⚠ You are selling the use of a vehicle, your time, and your presence in an enclosed space with a stranger. Only one of those is on the fare screen.

1.1 What you are actually selling

Four things.

⚠ One: the depreciation of a vehicle you own. ⚠ The largest input, never invoiced, and higher per mile in passenger work than in ordinary driving. Chapter 26.2.

⚠ Two: your whole time. ⚠ Not the engaged minutes. The waiting, the repositioning, the drive to the pickup, the queue at the airport, the cleaning, and the drive home. Chapter 7.1.

Three: fuel and consumables.

⚠ Four: your presence, alone, in a confined space, with people you have never met, frequently at night. ⚠ This is the input the trade never names and it is genuinely part of what you are providing. Chapter 23.4.

⚠ The fare screen shows a payment for a trip. It shows none of the four.

What this reframing changes.

⚠ It makes utilisation the central figure rather than fare-per-trip. Chapter 7.2 ☐ ⚠ It makes declining a legitimate professional action rather than lost income. Chapter 4.2 ☐ ⚠ It makes the fourth input something you are entitled to protect, which is what rule one is. Chapter 4.1 ☐

1.2 What a passenger is actually buying

⚠ Not what drivers assume, and knowing the difference is most of Chapter 21.

In order.

One: getting there. ⚠ Reliably, in the time they expected, without incident. Almost everything else is downstream of this ☐

Two: not having to think. ⚠ They want a car that arrives, a driver who knows where to go, and a door that opens. Every additional decision you ask them to make is a small cost to them ☐

⚠ Three: feeling safe. ⚠ Unspoken, and it is a large part of the rating. It is produced by driving smoothly, being findable, confirming who you are, and not being unpredictable — not by conversation ☐

Four: a clean, comfortable, quiet-enough car. ⚠ Temperature and smell matter more than anything you can buy to put in it ☐

Five: not being talked at. ⚠ Or, for some passengers, being talked with. Reading which is a skill and getting it wrong in either direction costs you. Chapter 21.3 ☐

⚠ Notice that water bottles, mints, and phone chargers appear nowhere on that list. They are pleasant and they are not what is being bought, and a driver optimising them in response to a rating drop is optimising the fifth-most-important thing. Introduction.

1.3 The four costs nobody shows you

⚠ 1. Depreciation. ⚠ Silent, per-mile, arriving once as a number that has nothing obviously to do with any particular shift. Chapter 26.2.

⚠ 2. Unpaid online time. ⚠ The hours you were available and not engaged. In many markets this is a third to a half of the time you gave. Chapter 7.2.

⚠ 3. The drive to the passenger. ⚠ Frequently unpaid or barely paid, and it is miles on your car every single time. Chapter 7.1.

⚠ 4. Tax and reserves. ⚠ Every fare arrives gross, and three portions of it belong elsewhere before you feel paid. Chapters 18.5 and 31.6.

⚠ And a fifth that is not a cost in money: the cumulative effect of night hours, strangers, and constant low-grade vigilance. It is real, it is why people leave this trade, and no figure captures it. Chapter 23.2.

1.4 What the trade competes on, and what it should

What drivers talk about. ⚠ Which platform pays best ☐ ⚠ Surge and where to find it ☐ ⚠ Hourly averages ☐ ⚠ Ratings ☐ ⚠ Airport queues ☐ ⚠ Bonuses ☐

⚠ Almost all of it is a fare-per-hour conversation, and fare per online hour is the number that hides the car.

What actually determines whether this works.

⚠ Your cost per mile, built from your own figures. Chapter 26.1 ☐ ⚠ Your utilisation — engaged time as a proportion of online time. Chapter 7.2 ☐ ⚠ Your earnings per online hour, not per engaged hour ☐ ⚠ Whether you drive tired. Chapter 23.2 ☐ ⚠ Whether you have a record of anything. Chapter 5 ☐ ⚠ Whether you have ever ended a ride. Chapter 4.5 ☐

⚠ Two drivers in the same city on the same platform with identical gross can be having completely different years, and none of the difference is visible from outside.

1.5 Where the money actually is

⚠ And the largest lever nobody uses: raising utilisation. A driver who moves from fifty per cent engaged time to sixty-five has increased earnings by nearly a third without a single extra hour or a single better fare. Chapter 7.2.

1.6 How AI Helps

Use it for. ⚠ Utilisation and cost-per-mile arithmetic from your own figures ☐ Analysing your own request log ☐ Building record templates ☐ Drafting the licensing and insurance questions ☐

Never use it for. ⚠ Any decision while driving ☐ ⚠ Anything containing a passenger's name, pickup or destination ☐ ⚠ Telling you what platforms pay in your market ☐

1.7 Common Mistakes

⚠ Tracking fare per trip and nothing else. Section 1.4.

⚠ Counting engaged hours as your working hours. Section 1.3.

⚠ Optimising the fifth thing a passenger wants. Section 1.2.

⚠ Chasing surge. Chapter 11.4.

Spending gross fares. Chapter 18.5.

⚠ Believing fuel is the main vehicle cost. It is not. Chapter 26.2.

⚠ Taking a rating movement personally. Introduction.

1.8 Safety, Legal, Ethical, and Professional Considerations

⚠ Establish before your first ride whether carrying passengers for hire requires a licence, permit, or registration where you live — of you personally and of the vehicle. It does almost everywhere and the requirements are specific. Chapter 15.2. Stop-work.

⚠ Establish whether your insurance covers carrying passengers for hire. Ordinary personal policies very frequently exclude it entirely. Chapter 15.3. Stop-work.

⚠ Establish what a platform's insurance covers and during which periods, because it frequently differs between waiting, driving to a passenger, and carrying one.

⚠ You may decline or end any ride for your own safety, at any point, without explaining it to anybody. Chapter 4.

1.9 Do This Now

☐ ⚠ Find out today whether you need a licence to carry passengers for hire here. Resource 3. Stop-work. ☐ ⚠ Ask your insurer whether carrying passengers for hire is covered. Resource 19. Stop-work. ☐ ⚠ Photograph your odometer. Chapter 18.2. ☐ ⚠ For one shift, record online time and engaged time separately. Resource 16. ☐ Write down the four costs from Section 1.3. ☐ ⚠ Start a ride record tonight, even if it is three fields. Resource 31.

1.10 Chapter Summary

You are selling a vehicle's depreciation, your whole time including the unpaid majority of it, fuel, and your presence alone in a confined space with strangers — and the fare screen shows none of them. Passengers are buying arrival, not having to think, and feeling safe, in that order, which is why the items drivers buy in response to rating drops sit fifth. The trade talks in fares per hour, which is the number that hides both the car and the idle time. And the largest unused lever is utilisation: moving from half your online time engaged to two-thirds raises earnings by nearly a third without one extra hour or one better fare.

Prompt 1 — My utilisation, measured

"For one shift I recorded: total time online [A] minutes; time spent waiting for a request [B]; time driving to passengers [C]; time with a passenger in the car [D]; and gross earnings [E]. Calculate: utilisation as engaged time over online time; earnings per online hour; earnings per engaged hour; and the difference between those two figures as a percentage. Show the arithmetic simply."

Prompt 2 — The four costs, estimated

"Help me build a rough picture of what a week of rideshare driving costs me. My figures: odometer miles driven [A], hours online [B], hours with a passenger [C], fuel spent [D], gross earnings [E]. Calculate fuel cost per mile, gross per mile, gross per online hour, and gross per engaged hour. Do not estimate depreciation — flag that it is missing and note that it is likely the largest single cost."

Prompt 3 — Licensing and insurance questions

"Draft the questions I must ask before starting app-based passenger driving, covering: whether a licence, permit or registration is required to carry passengers for hire, of me personally and of my vehicle; what vehicle inspections or standards apply; whether my current insurance excludes carrying passengers for hire; what cover I would need instead; and what a platform's insurance covers and during which periods. Name the kind of body to ask for each. Phrase them as questions only and do not state what any jurisdiction requires."

Prompt 4 — What passengers actually want

"A passenger is buying, in order: arriving reliably and without incident; not having to make decisions; feeling safe; a clean and comfortable car; and the right amount of conversation. Here is what I currently do to try to improve my ratings: [list]. For each item, tell me which of those five it addresses, and flag anything on my list that addresses none of them. Do not suggest new items yet."

©2026 James Henderson / https://localhandyman.work