Lyft vs Uber pay for drivers: run the test yourself, in your own city

Every driver forum has a thread arguing about which app pays better, and every one of them is people comparing cities without noticing. A Phoenix driver's answer is worthless to a Cleveland driver. Your car burns fuel at a rate the other guy's does not. Your usual pocket generates pickups his never sees.

The good news is that this is a measurable question, and measuring it takes one week and a spreadsheet.

The catch is that almost everyone measures it wrong, in the same specific way, so start there.

The trap: you cannot compare only the trips you took

The obvious method is to look at last month's earnings statements, divide by hours, and see which app came out ahead.

That measures your habits, not the apps.

If you accept most Uber pings and only take Lyft when the payout looks big, your Lyft average will look wonderful and your Uber average ordinary, and you will have learned nothing except that you are pickier on one app. The offers you declined are part of what an app is worth to you. An app that sends you eleven trash offers and one good one is not the same as an app that sends you four decent ones, even if the trip you completed was identical.

So the unit of measurement is the offer, not the trip. Every ping either app puts on your screen goes in the log, accepted or not.

What to record

Seven fields per ping, and they all fit on the card before it times out:

• app
• time of day, to the nearest hour
• payout
• trip miles and trip minutes
• pickup miles and pickup minutes
• what you did with it

That last field is for your own review later. It is not part of the comparison.

A filled row looks like this. These figures are illustrative, not a shift I am reporting.

| App | Hour | Payout | Trip | Pickup | Action | |---|---|---|---|---|---| | Uber | 18:00 | $14.20 | 6.8 mi / 19 min | 2.4 mi / 6 min | took | | Lyft | 18:00 | $9.60 | 3.1 mi / 12 min | 0.9 mi / 3 min | took | | Uber | 19:00 | $23.40 | 17.2 mi / 31 min | 7.1 mi / 15 min | passed | | Lyft | 19:00 | $11.05 | 4.4 mi / 15 min | 3.8 mi / 10 min | passed |

Four rows takes ten seconds at a red light if you keep a note open. Most of the drivers I know who have tried this gave up on typing and started screenshotting the card instead, which works as long as you transcribe the same evening while you remember which app it was.

The arithmetic, applied identically to both

Score every row the same way:

Total miles = pickup miles + trip miles Net pay = payout minus (total miles times your cost per mile) Net dollars per hour = net pay divided by (pickup minutes + trip minutes, in hours)

Take the first row above at $0.26 a mile. Total miles 9.2, cost $2.39, net pay $11.81, total time 25 minutes. That is $28.34 an hour. The third row: total miles 24.3, cost $6.32, net pay $17.08, total time 46 minutes, which is $22.28 an hour despite a payout nine dollars larger.

Pick one cost per mile and hold it for the whole week. The number itself matters less than using the same one on both apps, though if you want it to be right rather than merely consistent, build it properly using what a mile actually costs you and swap in your car's EPA rating from the vehicle pages.

One warning about the default. DriverSignal ships $0.26 a mile, assembled from 28 MPG at $3.20 a gallon plus around $0.15 a mile of wear. The gas half is a setting, not a reading of the pump. EIA's published average for regular gasoline was $4.096 a gallon nationally for the week of 2026-07-27, which at 28 MPG works out to $0.146 a mile of fuel and a total nearer $0.30. Use whichever is true where you fill up. Houston came in at $3.600 that same week and San Francisco at $5.578, so the honest answer is genuinely different by city.

Reading the result

At the end of seven days you have two piles of scored offers. Three numbers come out of them, and they answer different questions.

Median net dollars per hour across all offers seen. This is what the app is putting in front of you. Use the median rather than the mean, because one airport run will drag a mean somewhere unrepresentative and you will believe it.

Share of offers that cleared your floor. An app whose median is lower but which sends twice as many pings above your floor may still be the better app to keep open, because you are not paid for offers, you are paid for the ones you take.

Net dollars per hour of time online. Divide the net pay of the trips you actually completed by the hours you had the app open. This is the number your bank account agrees with, and it is the one that punishes long pickups and dead zones.

Those three can disagree, and when they do the disagreement is the finding. An app with a strong median and a low clearance share is sending you a few good offers and a lot of noise, which is a case for keeping it open alongside the other rather than favouring it.

How many offers before you believe it

Around fifty per app.

That threshold is not arbitrary. It is the same one DriverSignal uses before its adaptive floor takes over from the fixed one, and it comes from the same problem: with twenty samples, two odd pings move the picture enough to reverse your conclusion. There is more on how that floor is built in pick a floor and hold it.

Fifty offers per app is roughly a busy week for a full-time driver and two or three weeks part-time. If you cannot reach it, say so in your own head rather than treating a thin sample as an answer.

Three ways to wreck the test

Alternating days. One app Monday, the other Tuesday. This measures Monday against Tuesday. Demand varies far more across the calendar than it does across the two apps, and you will end up with a confident conclusion about the wrong variable. Run both open at once and compare within matching hours.

Changing your behaviour mid-week. If you get excited on day three and start accepting things you would normally pass on, the log stops describing your ordinary driving. Keep driving how you drive. The log is an observation, not an intervention.

Comparing across different weeks. A promotion running on one app during week one and not week two will produce a difference that has nothing to do with either app's base pricing. If a bonus is live, note it on the affected rows and consider them separately.

The part you cannot measure, and I am not going to pretend otherwise

This test tells you what each app offered you. It cannot tell you why.

I cannot see either company's dispatch. I do not know whether being selective changes which offers reach you, how often, or at what price, and I have no way to observe it from outside the car. Neither does anyone else running an app like mine. If someone assures you that declining has no effect on what you get shown next, they are stating something they cannot know.

What the test does give you is your own record, which is more than the forum thread has. Two piles of scored offers from your city, your car and your hours beats any confident stranger.

While the test runs

Running both apps means declining on one to accept on the other, and that has a price worth naming. Uber ties acceptance rate to Uber Pro tier status and the perks attached to it, and drivers report a short stretch without requests after several declines in a row. What public evidence does not show is accounts terminated for declining. Skipping protects your take-home; it may cost Pro-tier perks or a brief cooldown, not your account.

Decline before you accept, never after. In 9,572 trips over three years and three months I have kept a 4.96 rating and a 2 percent cancellation rate while being fussy about what I take, which is one driver's record rather than a forecast for yours, but the shape of it is the point: passing is cheap, cancelling is not.

Doing the scoring without the spreadsheet

Scoring fifty offers a week by hand is the part where most people quit, which is fair. You can run any single card through the is-this-ride-worth-it calculator to check your arithmetic, and the longer explanation of the method is in the 20-second math I run on every offer.

DriverSignal does the same scoring automatically on both Uber and Lyft cards as they appear, keeps the samples, and uses them to build the adaptive floor. It never touches Accept or Decline. The decision on every one of those fifty offers stays yours, which is the only arrangement I am willing to ship.

And if the answer comes back that neither app is clearly better, that is a real result too. It means the question was never Uber against Lyft. It was this offer against your floor, all along.