The first product was a warning
I wrote about Postmates in a July 12, 2020 First 1000 issue, shortly after the Uber acquisition made the story timely. The source separates an early courier service from the later “buy anything” proposition.
The first beta launched in September 2011. Through the old Postmates era, described as September 2011 to April 2012, Bastian focused on retailers and couriers. Merchants said they would use the service 1,000 times a day, but the source says they produced only one or two orders at the end.
At the same time, users began asking whether Postmates could purchase things for them. It took seven months for Bastian to test that request. The strongest demand signal was not the market-size slide. It was the repeated question from users.
Change the side you are making liquid
For the GetItNow test, Bastian shifted the model. Instead of relying on merchants for demand and distribution, the team focused on consumer demand and couriers while hacking the supply catalog.
The source says the team listed businesses from public directories and used Visa gift cards for couriers. That is a historical description of how they tested a proposition quickly, not a recommendation to copy unpermissioned data or bypass merchant relationships today.
The operator move is the reallocation: stop forcing the original supply side to create demand when customers are already asking for a different transaction. Make the missing side good enough to test the behavior.
Stage the test and count quality
The source reports a pre-beta with 300 customers and $5,000 of GMV. In the invite-only beta, users got five invites, a two-day weekend offer required a #GetItNow tweet, and Bastian offered free delivery for feedback. The issue says the first launch followed 27 days later.
Those mechanics did two jobs: they created a bounded way to get demand and they gave the team feedback before building every piece of infrastructure. The source says Bastian tracked how much users spent per month and cared about service quality rather than only user count.
This is the part worth carrying forward. A marketplace test should have a short loop, a clear quality signal, and a reason for people to tell you what broke.
The marketplace pivot test
If a marketplace is not behaving as planned, I would ask:
- What user request keeps appearing even though it contradicts your first supply plan?
- Which side can you fake or constrain safely for one bounded test?
- What is the smallest transaction volume that can reveal quality?
- Which incentive gets feedback before you build the full system?
- Are you measuring value per customer, or only counting signups and orders?
The takeaway
Postmates’ useful move was not a clever invite loop. It was listening to a request that invalidated the original marketplace shape, then testing the new side with constrained supply and a quality metric.
Sources and original research
Adapted from Postmates, published July 12, 2020. The September 2011 launch, September 2011–April 2012 early era, seven-month delay, 20 gift cards, 300 customers/$5,000 GMV, five invites, two-day offer, 27-day launch timing and quality-per-customer framing are historical claims from the issue. No current Postmates/Uber metrics or new marketplace test is claimed.
Updated September 21, 2026. Based on original First 1000 reporting and the sources listed above.