Close to 30% of the qualified leads in DoorDash's merchant-acquisition account were coming from a segment spending about 1% of the budget. Nobody decided that. The money was going where it had always gone.
DoorDash only makes money when there are stores to order from. Behind the consumer app sits a business-to-business sales engine that has to convince liquor stores, florists, pet shops, and neighborhood grocers to put their catalogs on the platform, an expansion DoorDash called New Verticals.
From late 2021 through the third quarter of 2022 I ran the paid media that recruited those merchants, on an agency engagement through Right Percent, along with the weekly analysis that decided where the next dollar went.
Every finding below came out of one habit: split the account into segments, price each one, and let the budget mistakes announce themselves.
One blended number hides the answer
The reflex on a large paid account is to read a single cost per lead and call the whole thing healthy or sick. I priced each segment separately over the same four-week window in early 2022 instead, the same teardown I run in the growth audit. The segments were telling completely different stories.
New Verticals produced 595 marketing qualified leads, store owners who clicked an ad and asked to hear more, at $300 each on roughly $180,000 of spend. Eighty-six of those merchants signed and went live.
The flagship Merchant segment, the largest and most established, paid $801 per lead. The restaurant SMB segment paid $844. New Verticals also turned clicks into leads at more than double the restaurant rate, so it was cheaper at the top of the funnel and tighter the whole way down.
The mismatch was in where the money went. The flagship absorbed about 70% of the spend across these segments while producing roughly a third of the leads and a third of the signed merchants. It was buying the most expensive lead in the account, in bulk, because it had always been the segment that got fed.
Budgets follow history until someone makes them follow efficiency.
New Verticals bought a lead for less than half of what the flagship paid. The growth was sitting in the categories nobody had prioritized.
The segment that ran on almost nothing
The segment from the opening line was a small group the account tracked as self-delivery, merchants who sell through the app but handle their own deliveries. It generated close to 30% of all qualified leads in the window on just over 1% of the spend, signed 71 merchants, and had the lowest cost per lead in the account by a wide margin.
The cheapest growth anywhere in the account was already running. It just was not funded.
Nobody starves a segment on purpose. The flagship gets the money by default, and default is the hardest budget line to beat.
The recommendation wrote itself. Move money off the flagship and into the segments already converting for a fraction of the cost.
Inside New Verticals, flowers kept returning signed merchants at a rate the other categories could not match, so the plan fed the categories the numbers pointed at rather than the ones that happened to be biggest.
Creative that spoke a store owner's language
None of it worked without creative built for the person on the other side of the ad. A store owner is a different audience from the diner the app talks to every day, and the ads had to sound like someone who understands how a small business makes money.
We shipped a fresh flight of nine static variants most weeks, and the copy led with margins: "Raise your liquor store's bottom line" is a headline for someone counting the till at the end of the month.
Variants were scored and rotated on the method behind the creative scoring system. Winners earned another flight and tired variants came down on schedule.
A store owner clicks for their bottom line. The ad that recruits a merchant shares nothing with the ad that sells dinner except the logo.
A result worth caveating
The other half of the job was holding back from declaring victory early. We tested a structural change in the retargeting: one consolidated full-funnel campaign against the older setup that split prospecting and retargeting apart.
The consolidated version came back roughly 7% cheaper per signed merchant with about 18% lower CPMs, the price of a thousand impressions. It looked like a clean win.
I wrote it up as a lead rather than a conclusion. The two setups had run months apart, and seasonality or shifting auction prices could account for a swing that size on their own.
The honest recommendation was a true A/B split, same weeks and same audiences, before real money moved on the strength of it. Gordon and Zettelmeyer's comparison of 15 Facebook experiments found that the observational methods advertisers normally rely on often fail to recover the true effect of advertising.
A finding you caveat is worth more than one you oversell, because the caveat is what stops you scaling a mirage.
The year underneath the snapshot
Zoom out from the four-week window and the account had been improving for a year. The cost to sign a New Verticals merchant fell from about $6,400 in the first quarter of 2021 to roughly $2,250 by the fourth.
The lesson holds at every altitude. Inside a large account the growth sits in the segments, where one category quietly returns leads at a third of the cost while another soaks up the budget on reputation.
Rank the parts, then move the money. That discipline anchors the growth operating system.
If you are running a big paid account and suspect the cheap growth is buried in a segment nobody has ranked, this is the work clients bring me in for. Let's talk.