Running the Operational Backbone of a Seed-Stage Marketplace
Bringing structure to contractor verification, user engagement, and the commercial model at a fast-growing two-sided marketplace — when the operational surface had outgrown the informal processes holding it together.
The situation
A seed-stage marketplace was connecting homeowners and architects with vetted construction contractors. The product worked, with a team that was smart and active. The business was growing fast enough that operational gaps were appearing faster than anyone could close them.
It's the reality of early-stage. There is always more to do than there are people to do it, and ideas are rarely the problem. What you need is someone who can look at the whole system, find what's falling through, and build something that holds. I joined at a moment when the operational surface had grown well beyond what informal processes could support.
The challenge
Running a two-sided marketplace is genuinely more complex than it looks from the outside. You're not managing one customer relationship, you're managing two, simultaneously, with completely different needs, incentives, and reasons to disengage.
On the contractor side, the structural challenge was trust. Construction has a reputation problem. Homeowners and architects arrive already sceptical. To build a marketplace worth trusting, you need to be able to stand behind every contractor in the network, which means your verification and quality processes have to be rigorous, consistent, and scalable — not dependent on one person's memory or a spreadsheet.
On the demand side, the challenge was engagement and responsiveness. Contractors invested time bidding on projects. When homeowners went quiet, without explanation, it corroded confidence in the platform. Every step of the journey, from a homeowner first uploading a project through to a contractor winning the tender, was a potential drop-off point. At seed stage, you can't afford to lose people to friction you haven't noticed yet.
Underneath all of this sat a commercial question that hadn't been fully interrogated: whether the subscription model the business was running could actually be sustained given the supply of available projects in the market.
What I did
The work spanned four areas, simultaneously.
Contractor quality and trust
The verification process existed but wasn't systematic. No consistent methodology for how contractors were assessed, no shared visibility on where each one was in the process, and no way to work through a backlog at pace. I built a scoring matrix in Notion — each contractor assessed against a defined set of criteria, including company registration, years in business, past project history, and reference checks, with an automated score populated from inputs. Worked through the backlog in structured weekly sessions with a clear split and defined criteria, so quality didn't depend on who happened to be checking. Created a trusted analytics report in Mixpanel showing verification status across the contractor base — pass, fail, in progress — so the whole team had visibility and nothing sat in someone's personal to-do list. The result was a step-change in verification throughput: the backlog cleared, quality improved, and there was a consistent process to run at scale.
User engagement and campaign performance
Email and push notification campaigns were running on internal assumptions about when contractors were most likely to engage. One of those assumptions — that Sunday wasn't a useful send day because it wasn't a working day — turned out to be wrong. I designed and ran structured A/B tests across timing, messaging, and audience segmentation using Customer.io. Sunday morning was consistently the highest-performing window: contractors managing live projects during the week used Sunday to do proactive work they couldn't fit in otherwise. I applied the findings across campaigns, including re-engagement flows for contractors who had dropped off mid-onboarding. Open and click-through rates improved by approximately 25% as a direct result of the experimentation.
Marketplace health and user journey
A significant amount of time went into understanding what was actually happening versus what the team assumed was happening. The gap between the two was wider than expected. I mapped the full user journey for three user types — contractors, homeowners, and architects — identifying every friction point, drop-off moment, and step where the experience diverged from the intended flow. Built the chatbot decision tree for Intercom, structured around the most common queries from each user type, to reduce the volume of support tickets hitting the team. Worked on the matching and lead-sending logic: which contractors to prioritise for which tenders, and when to send them, to improve acceptance rates and reduce contractor frustration from irrelevant leads.
Commercial model
At one point, I pulled together data on the subscription model — what it needed to deliver versus what the UK market could realistically supply at any given time. The gap was structural, not a growth problem. I presented it simply, with numbers, to the CEO. They changed the model the following week.
The result
Verification went from ad hoc to systematic, with a visible metric to show it. Campaign performance improved measurably. The team had shared visibility across areas that had previously sat in individual inboxes and informal processes.
And the commercial model was reframed.
Recognise something in this?
Two-sided dynamics, operational gaps, processes that haven't kept up — feel free to reach out.