← harrison.buildMANAGED BY Q2019–20

Case study · Two-sided marketplace

Making the Managed by Q marketplace self-service

Harrison Johnson, product manager · Design by Rebecca Erickson · Mar 2019 to Jan 2020

Office managers came to Managed by Q to book vendors: cleaners, handymen, IT, movers. In early 2019 our support team still did most of the matching, and a request could wait a full business day before a vendor saw it. Rebecca and I owned the rebuild end to end. We let customers pick their own vendor, and the share who did went from 11% to 83%. Then we turned the support option off.

The redesigned Services page on desktop, with search, project conversations and service categories, beside a vendor card on mobile.

Where we started

A marketplace that used to be the whole product

Managed by Q began as office cleaning and grew into a marketplace for every office service. By 2019 we had bought a small startup and WeWork had bought us. The marketplace was turning into one feature inside a bigger SaaS tool for office managers, and a smaller team had to keep it running.

Customers felt the old model. Booking took hours at best and days on average. We kept four booking flows alive: the main one, instant booking for light maintenance, re-ordering from a past vendor, and messaging a past vendor until they sent a work order. Customers also couldn't tell what Managed by Q did in the middle.

Before · the main booking flow

  1. Search for a service
  2. Enter service details
  3. Enter building details
  4. Wait up to 1 business day
  5. Get matched with 1 to 3 vendors
  6. Message vendors, get quotes
  7. Approve a quote

After · one flow for everything

  1. Search for a service
  2. See available vendors
  3. Message one or more
  4. Approve quotes when ready

The bet

Show customers the vendor network and let them choose

Customer interviews pointed at two gaps. Office managers didn't know what would happen after they asked for a service, and they didn't trust a vendor they had never heard of. Self-service would only work if we fixed both. We shipped the pieces one at a time so we could watch the 11% move and change course between releases.

What we shipped

Five changes, released one by one

Vendor profiles vendors could edit

Vendors wrote their own profiles. We added standard tags like quick turnaround, union friendly and veteran owned, so two cleaners could be compared side by side.

We also pulled in Google and Yelp reviews. New vendors had no reviews with us yet, and research showed office managers were googling vendors anyway before they hired one.

Vendor profile for Shield Facility with standard tags, a 92% would-hire-again score, Yelp rating and customer reviews.

Reviews up front

The line we heard most in research was that office managers trust other office managers. So reviews moved earlier in booking, and we made reviews easy to leave: five yes-or-no questions, starting with "Would you hire this vendor again?"

A vendor card with a 97% would-hire-again score next to a five-question review form.

Projects built around messaging

Each project page became a conversation with the vendor, with the quote and an approve button on the side. Customers talked to vendors sooner, and the web app caught up with the mobile app.

A project page showing a message thread with a vendor, an estimated price, and an Approve quote and hire vendor button.

Words that describe a tool

The old copy made us sound like a concierge. We rewrote the labels so customers could see they were the ones choosing.

  • Request quotes → Find a vendor
  • Submit request → Message vendor
  • Orders → Projects
  • Create an order → Start a conversation
  • Questions? We're here to help → Message your vendor or contact support
The language audit: old button and page labels mapped to new ones.

One booking flow instead of four

Instant booking, re-ordering and vendor work orders all folded into the same search, message, approve path. A smaller team now had one flow to maintain.

The hard call

Turning off the safety net

Support-assisted matching was the last piece of the concierge model. As the self-service share climbed release after release, we got the confidence to switch it off. From then on, every new project was self-service, and a smaller team no longer had to run the old model next to the new one.

We then tested mobile concepts with customers and started tying the marketplace to the rest of the app, starting with the Vendor Directory. Users and our sales team had both asked for that.

Wireframes for the marketplace on mobile, beside a vendor card on a phone.

What comes next

Agents will run the busywork, and buyers will pay for it

At Q we took the human matchmaker out by giving the buyer better tools. I think the next step takes the buyer's busywork out too. An agent will search the vendor network, message vendors, compare quotes and book, and bring the office manager in at the decision.

When an agent does that work, pricing should follow the work. At Q most of the effort came before any deal closed: matching, messaging, chasing quotes. A take rate only pays when the deal closes.

TodayThe marketplace takes a cut of each closed deal. Searching and quoting are free to the buyer and paid for out of the cut.
NextThe buyer pays for the agent's usage on each task, and the take rate shrinks or goes away.

I was the product manager on this project from start to finish. Rebecca Erickson led design, and we worked with an engineering lead, two engineering squads, and partners in product marketing and marketing. Screenshots are Rebecca's, used with her permission. Her design case study is at rebeccaherickson.com.

Claude Code built this page from my notes. The story and the calls are mine.

← Back to harrison.build