How much should a taxi ride cost?

Search for this and you get a dozen fare calculators. Type an origin and a destination into any of them and you get a number.

They are all doing the same thing: looking up a route you have not taken yet, multiplying its length by a rate table, and predicting. That is genuinely useful before you travel — it tells you whether a trip is a five-dollar trip or a fifty-dollar one.

It is close to useless for the question people actually have while sitting in a taxi, which is not what will this cost but what is this ride worth right now — the ride with the diversion in it, the twenty minutes of traffic, and the route the driver actually chose.

Why predictions and measurements are different numbers

A prediction assumes the route the planner picked and typical conditions. A real ride departs from both, in ways that all push the same direction:

The route taken is not the route planned. Roadworks, a one-way system, a driver's own preference, or a deliberate detour.
Traffic is charged. Under any tariff with a time rate, twenty minutes stationary is money. A predictor assuming free-flow does not include it.
Waiting counts. Time at pickup, a stop on the way.
Surcharges apply. Night rate, airport pickup, luggage, tolls.

A predictor that says "about 1,400" and a meter that says 1,850 are not necessarily in conflict. The prediction described a hypothetical trip; the meter priced a real one.

Which is exactly why "the app said 1,400" is such a weak thing to say at the end of a journey — and why measuring is a different and stronger position.

What you need to work it out

Three numbers from your city's official tariff, which is usually published by the transport authority or printed inside the vehicle:

Flag-fall, and what distance and time it already covers
Per kilometre
Per minute (the waiting or time rate) — often zero for negotiated rides

Plus how far you actually went and how long it took.

The distance is the hard part. Not the meter's distance — that is the number in question — but an independent one. Which is what a phone is for: it has a GPS receiver, and it is already in your hand.

The arithmetic

Do not add a distance charge to a time charge. That is the mistake almost every "how to calculate a taxi fare" answer makes, and it inflates the result on exactly the rides where it matters. For each stretch of the journey a meter charges whichever of the two rates is worth more, not both — the full explanation is here.

A shortcut that gets you close by hand:

Work out the changeover speed: 60 × per-minute ÷ per-kilometre, in km/h.
Any part of the journey faster than that: charge distance only.
Any part slower than that, including stopped: charge time only.
Add the flag-fall. Subtract whatever distance and time it already covered.
Add surcharges last.

With a tariff of 80/km and 20/min, the changeover is 15 km/h. So: kilometres on the open stretches, minutes in the traffic, and never both for the same stretch.

Doing it without arithmetic

Working that out in a moving car is unpleasant, which is what Meterless is for. It measures the ride from your phone's own GPS and applies the tariff you entered:

Open it before the ride starts and enter your city's rates once — they are saved as a named preset, so you do that once per city, not once per ride
Press start when the vehicle moves
Watch the running estimate, the breakdown, and which rate is currently accruing
Press stop for an itemised total

It records like any GPS activity — the recording stays on your phone and nothing is uploaded — and it is an estimate for the passenger. It is not a meter. Taximeters are legally controlled instruments almost everywhere, they carry calibration seals, and charging a passenger with an uncertified one is an offence. This is a second opinion for the person paying, never an instrument for the person charging.

Expect the two numbers to differ. Here is by how much

A meter counts wheel revolutions. A phone measures position. These are different physical quantities and they will not agree exactly.

The gap is not small, either. On a measured ride, a vehicle's own odometer read 16.8% low against two independent measurements that agreed with each other to 0.6% — almost certainly a gearing or tyre-size calibration issue, not a fault.

So a divergence of a few percent means nothing at all. What is worth attention is a divergence that is large, or one that cannot be explained by the coverage figure — the proportion of the ride the phone could not measure, which is reported alongside the estimate rather than hidden behind it.

What "reads low" means, and why it always does

A GPS estimate errs downward, consistently:

Fixes worse than 30 m accuracy are discarded rather than trusted. In an urban canyon that is routine.
A tunnel produces a gap, and a gap accrues nothing — no distance, and no time either, because nobody knows whether the vehicle was moving.
The jitter filter ignores movement too small to distinguish from noise.

Every one of those subtracts. None of them adds. So if the phone reads higher than the meter, that is a bigger signal than the reverse — the estimate had to overcome a systematic bias to get there.

When the honest answer is "the meter is right"

Worth saying plainly, because a tool that only ever produces grievances is not a useful tool:

A few percent apart — the meter, every time. It is the certified instrument and it is measuring the wheels.
The ride was in heavy traffic and the fare looks high for the distance — that is a meter working correctly. Time is chargeable. It is not a distance bill.
Coverage well under 100% — the estimate is incomplete and knows it.

The case worth pursuing is a large gap, on a ride with good coverage, that traffic does not explain. That is a different article.

Where to go next

Work out a fare → Work out what a taxi ride should have cost, measured from your own phone.