August 2026 Earnings Index: the market got wider, not better

August didn't get better. It got wider. The median offer slipped to $1.22 per mile, and the share of offers priced under a dollar rose to one in three — but the top quarter of the market actually improved, and the projected hourly on analyzed offers didn't move at all. The bottom sagged, the top held, and the gap between a good offer and a bad one grew. That's a market that pays for filtering more than it did in July.

Key findings

The market spread out

Half of August's offers landed between $0.91 and $1.79 per mile. In July that range was $0.96 to $1.74 — so the bottom of the middle fell five cents while the top rose five. The same driver saw more offers that don't pay for the car and more offers that pay very well, in the same shift. Autopilot averages those together and gets a mediocre month; a floor keeps only the second kind.

The floor itself moved this summer. The IRS raised its vehicle cost basis from $0.725 to $0.76 per mile effective July 1 (the mid-year change, explained). Against that number, the bottom quarter of August's offers paid the car and left about 15 cents a mile — or less — for the driver, before taxes.

More MAYBEs, fewer clean calls

In GigReal you set two minimums — pay per mile and pay per hour — and every analyzed offer is checked against your own floors. TAKE cleared both. MAYBE cleared one. SKIP cleared neither. August's split, with July in parentheses:

Clear takes barely moved. What changed is that clear skips became MAYBEs: offers that pass one floor and fail the other grew five points to 62% of everything analyzed. That's the widening spread showing up in verdicts — a long run at surge that pays the hour but not the mile, a short hop that pays the mile but wastes the hour. Nearly two of every three decisions in August were a judgment call, which is why how you call a MAYBE mattered more than ever.

Spark, out of beta and into the numbers

Walmart Spark drivers analyzed 1,094 offers in August, nearly three times July's 405. The median came in at $1.94 per mile, up from $1.82, and the projected net hourly — which counts realistic in-store and waiting minutes — rose to $17.44 from $13.54. Part of that is the reader getting sharper as drivers register their usual stores; part of it looks like a real improvement in the offer mix. Either way, Spark remains the richest per-mile platform we track, with the usual caveat that the hourly is where shopping time gets paid for. New to it? Here's how Spark support works.

A change in how we count — and July, restated

Not every offer the reader sees can be fully read. When a screen is captured mid-animation or a platform changes its layout, the app sometimes gets the pay but not the miles — and an offer with no mileage can't have a real per-mile figure. Those partial reads were 30% of August's analyses, up from 21% in July, and until now they were counted as $0.00-per-mile offers in this Index. That dragged the medians down and pushed the under-$1 share up in a way that measured our reader, not the market.

Starting this month, the Index only counts offers where both pay and miles were read. On that basis, July's figures restate to a $1.27 median, 28% under $1, and a TAKE / MAYBE / SKIP split of 18 / 57 / 24 — versus the $1.09, 44%, and 14 / 66 / 20 we published in the July report. Every July comparison above uses the restated numbers, so month-over-month movement is apples to apples. The July post stays up as published; treat this section as the correction of record. Improving the reader's hit rate is a product priority — the 2.4 update added a way to send a missed offer from inside the app, and every one sent helps.

A note on volume

Total offers analyzed fell from 10,843 in July to 6,841 in August. That tracks the late-summer lull in hours driven — fewer shifts, more vacations — not a measured drop in platform demand. As always, treat the per-offer figures, not the totals, as the market signal.

What this means for drivers

A wider market is a worse market for anyone accepting on feel and a better one for anyone with a floor. The offers above your bar paid as well in August as they did in July — the projected hourly on analyzed offers was identical to the cent. What got worse is the cost of not filtering: more sub-$1 offers to decline, more MAYBEs to think through, and a cost floor that's 3.5 cents higher than it was in June. Set the floor against the new 76-cent basis, decline below it, and let September's richer mix do the rest.

Methodology

Figures are drawn from 6,841 trip offers analyzed with Trip Identifier in the GigReal app by US drivers between August 1 and August 31, 2026, compared against 10,843 offers analyzed in July. Beginning with this report, per-mile, per-hour, and verdict figures include only offers where both pay and mileage were successfully read (4,782 in August; 8,561 in July); July figures are restated on that basis. Per-mile figures divide total offer pay by total offer miles, including pickup distance. Projected hourly is the app's estimate at analysis time — offer pay against realistic trip minutes — and is not net of vehicle costs or time between trips. Verdict shares compare each offer against that driver's own configured per-mile and per-hour minimums — TAKE clears both, MAYBE clears one, SKIP clears neither — so they measure offers against drivers' standards, not a fixed benchmark, and month-over-month movement can reflect drivers adjusting their minimums as well as the offer mix. Spark figures come from the separate Spark offer reader and use the same readability rule. All data is aggregated and anonymized from an opt-in panel of drivers; no individual driver's data is identifiable or published, consistent with our Privacy Policy.