The Smarter You Get,
The Less They Pay You

Here's something nobody tells you before you start gig work: getting good at the job doesn't mean you'll earn more. In fact, for a lot of gig workers, the moment they figure out how to work smarter is exactly when the platform starts working them harder for less.

This is the efficiency trap. And it took months of firsthand experience to see it clearly.

Smarter than the app

The Spark app — like most gig shopping apps — routes you through the store in whatever order the algorithm decides. Pull up an order and you might see deodorant, shampoo, and soap on the list. Common sense says grab all of those together since they're in the same aisle. But the app would send you to a grocery item first, then bounce you across the entire store to the vitamin and over-the-counter medicine section, then back again.

It made no sense. The app wasn't routing efficiently — it was routing arbitrarily. So the obvious solution was to ignore the app's sequence entirely.

The better system looked like this: open the full order list before touching a single item, mentally group everything by store section, then shop in a logical physical sequence through the building.

This system cut shopping time dramatically. Orders that used to take 25 minutes started taking 14. Customers got fresher groceries. Ratings went up. Repeat customers started requesting the same driver specifically — not just for speed, but because this driver actually picked good produce, checked expiration dates, and evaluated meat by marbling and cut thickness the way a personal shopper would.

That's skilled work. A personal shopper or a specialty butcher gets paid a premium for exactly that judgment. In the gig economy, it went unrewarded — and then it got punished.

The algorithm noticed

As efficiency improved and ratings climbed, something else started happening. The pay per order quietly started to drop. The baseline shifted. Offers that would have paid $14 for a comparable order six months earlier were now coming in at $10 or $11. The platform had data showing this driver completed orders faster than average — and the algorithm apparently used that to justify squeezing the offer price.

Then came the double orders.

Two orders, one cart, one nightmare

Spark started bundling two separate customer orders into a single offer. On the surface it looked like a normal offer — you'd glance at the mileage, quickly divide the offer price by the miles, and it seemed reasonable. But buried in the details was the fact that it was two completely separate orders for two different customers, both of which needed to be shopped simultaneously.

The physical solution was to put one order in the main cart body and the second order in the child seat at the front — that small flip-down seat designed for toddlers. It worked, barely. But the mental load was enormous. You're constantly switching between two lists on the app, verifying which item belongs to which customer, making sure a frozen item for Customer A doesn't end up in Customer B's section and start thawing before delivery.

A couple of mix-up errors happened — not many, but enough. And when they did, the only acceptable response was to drive back and fix them. Not because the platform required it. Because taking pride in the work required it. That drive back ate 20–30 minutes of unpaid time.

What Walmart actually gets out of this

Step back and look at this from Walmart's perspective. They have a skilled, experienced driver who:

The platform captures all the upside of a skilled, experienced worker — faster throughput, happier customers, better ratings — while bearing none of the traditional employment costs. And as the driver gets more efficient, the algorithm's response isn't to reward that efficiency. It's to extract more value from it by lowering the per-order rate and bundling more work into each offer.

The $1/mile myth revisited

New gig workers are told that $1 per mile is the baseline for a worthwhile order. Here's what that actually looks like on a double order run:

That $10 offer — which looked like $1/mile at a glance — paid $4.45 per hour of real working time when everything was accounted for. Below minimum wage. Below what a parking lot attendant makes. And it required navigating a Walmart with two separate shopping lists, a cart with items stuffed into the child seat, and the mental overhead of keeping two customers' orders perfectly separated.

What to do with this

None of this means gig work is impossible to do profitably. But it requires going in with your eyes open — something the platforms are not incentivized to help you do.

Before you accept any offer, ask:

Is this one order or two? What are the actual total miles including the return trip? How long will the shop take? What does that leave per hour after gas, wear, and taxes?

The platforms show you gross offer price and one-way mileage because those are the numbers that make the offer look best. Your job is to do the math they don't do for you.

Read more