I Ran 21 Didi Rides to Check the Commission. The Average Was 13.22%. Drivers Still Don’t Buy It.

After 21 rides, Lao Yu sat in his car and pressed the calculator. 13.22%. He remembered one evening-rush ride: 23.9%. The two numbers stayed in his head. Neither won.

He cleared the calculator and entered the numbers again. 13.22%. A water bottle sat on the passenger seat. The cap was still off.

I

A Tsinghua University team surveyed 894 drivers. 768 of them photographed income reports for verification. Average monthly commission: 15.3%. 71.8% overestimated their commission. More than a third guessed above 25%. Of the 768 bills, one exceeded 25%.

Didi’s public data is close: the 2024 average across all orders was 14%.

Taken together, the numbers show a gap. Drivers feel “above 20%.” The monthly average sits lower.

Averages lose the memory contest. They have no name, no scene, no fuel gauge ticking in rush-hour traffic. Drivers remember one ride: the passenger pays this much, the driver gets that much, and the middle disappears. A 23% cut replays in the head. A commission-free ride leaves no trace because nothing was lost.

Someone in a driver group posted a screenshot with a 28% commission. A row of “+1”s followed. No one posted a 13% screenshot.

II

Drivers have reasons not to trust 13%.

Passenger fare and driver pay follow separate pricing rules. Drivers cannot check the gap. They see one line: platform service fee. No explanation follows.

High-commission rides exist. The per-ride cap fell from 29% to 27% only in 2025. In 2020, the company said orders above 30% accounted for 2.7%. Those rides may be rare. One hit becomes evidence in the group chat.

Subsidies arrive late. Incentives often settle the next day, not with the ride. Lao Yu opened his bill and found an extra incentive. The ride’s effective commission was lower than he remembered. His memory kept yesterday’s number.

Ride-by-ride reconciliation always feels like a loss. Monthly bills rarely get opened.

III

Where does the money go?

An official breakdown says drivers receive about 79.1% of passenger payments. The remaining 20.9% splits three ways: passenger subsidies and discounts about 10.9%, operating costs and taxes about 6.9%, net profit about 3.1%.

More than half of the commission returns to passengers as coupons and discounts. The platform keeps 3.1% as profit.

This breakdown does not settle every doubt. A platform’s own numbers need outside review, and the method deserves scrutiny. It does show that commission is not one transfer into the platform’s pocket. Subsidies change the final ratio. Driver incentives in hot zones and passenger discounts in cold zones move the gap between what riders pay and what drivers receive.

Drivers cannot see that flow. They see the platform service fee line. Subsidies, costs, profit: none of it appears on the bill. One visible deduction carries every complaint about cost.

IV

Commission is not the whole cost.

Deadhead miles do not appear on the bill. Fuel or electricity does not appear. Depreciation, insurance, maintenance, tires: none of it appears. Some drivers run empty 35% to 45% of the time. Nearly half their daily fuel cost burns on empty runs. The dashboard fuel number climbs. The bill omits the item.

Those costs eat take-home pay. “Commission is low” does not mean “drivers are fine.” Commission is the visible line. Much of the pressure sits in costs the bill never lists. Drivers pile every frustration onto platform service fee. It becomes the outlet and the stand-in.

V

Lao Yu’s test and the Tsinghua survey do not prove drivers imagine things. They do not prove the platform is innocent. They raise a narrower question: transparency should let drivers check the math, not just read an average.

Pressing the per-ride cap from 29% to 27% helps. Clear monthly bills, subsidy details, and a line-by-line match between passenger payment and driver receipt would help more. Without that, the platform says 14%, drivers say 20%, and the group chat stays split.

Transparency will not fix deadhead miles or income expectations. It can show how much was taken, where it went, and when subsidies arrive. Visible rules do not solve everything. Invisible rules breed distrust.

VI

Lao Yu drove home. The calculator stayed on the passenger seat. The screen had gone dark. He will drive again tomorrow. He can calculate 13.22%. He cannot calculate the pause of his finger on 23.9% during evening rush hour.