A ten-year-old marketplace with collapsing project volume, four departments that each held a different piece of the truth, and a pricing model that charged the side with no reason to stay.
MakersRow launched in 2014 and grew to roughly $9M ARR. From 2018 the line turned and never turned back — decline year on year, down to $25K MRR by the time I was brought in as product owner in 2022.
The company knew revenue was falling. That was the whole of what they knew. There was no account of why, because no single person or department held enough of the picture to build one.
They didn’t have a revenue problem they could see. They had a revenue problem nobody could describe.
Engineering, sales, marketing and operations each held a different section of the knowledge base. Marketing owned the pages, Google Analytics and Search Console. Sales was running independently — booking demos without a clear picture of the product, and promising features that did not exist. Both went to engineering with feature lists, and because of accumulated tech debt, every change took months.
My first job was not design. It was consolidation: pull what each department knew into one knowledge base, then give the CEO a single account of the business.
Every symptom pointed at marketing. The cause was structural.
MakersRow is two-sided. Buyers are brands who need something made — apparel, carpentry, hardware. Sellers are factories who want an additional channel. A marketplace only works when both sides are liquid.
Project volume was falling. Fewer projects meant factories extracted less value, so factories left, which made the platform worse for brands. Then in 2021 the owner raised prices across the board — on both sides at once.
The brand side was structurally broken. A brand posts a project, gets connected to a factory, and is done. There is no reason to keep paying $50–60 a month after that. The pricing model manufactured its own churn, and the price point suppressed the exact activity — posted projects — that the factory side depended on.
We were charging the demand side for the privilege of creating our supply side’s value.
I argued for minimal-to-no pricing for brands, reserving paid plans for genuinely exclusive options. Removing a revenue line in order to grow revenue was the hardest call to get agreed.
Posting a project went to two or three clicks — a post your first project button straight on the landing page, with signup moved to after the project exists rather than before. Posted projects rose within a month or two.
We contacted every factory that had left, with a promo offer and, more usefully, a conversation. We asked what had actually gone wrong and took the answers into the roadmap.
The 2014 UI was a live blocker for sales — they were demoing something they had to apologise for. Rebuilt with UI/UX designers against focus groups: the brands’ experience first, factories second.
One expensive plan became simple, medium and large, with real differentiation at the top — priority bidding and first pick on projects. A free trial let factories see the platform before committing. Conversion rose.
We turned the site and the marketing around it into a knowledge source — how things get made, how to source, tips and tricks — so that arriving at MakersRow wasn’t purely a sales encounter.
A feature was proposed. Someone produced a wireframe. UI/UX turned it into a design. The design went to a meeting to be discussed. It went to development. It came back to a meeting to be discussed again. Then it shipped.
Two to three weeks was a fast cycle. Meetings were spent reacting to drawings of software rather than software. Sales and marketing sat behind a queue they could not influence, which is why they kept escalating feature lists that took months to land.
The backlog wasn’t the bottleneck. The distance between deciding and seeing was.
We rebuilt the platform to be AI-first in how it is developed, with the design system embedded rather than living beside the code. Changes that took two to three weeks now take a day at most.
The meeting changed shape as a result. We no longer present a design and ask what people think. We present the built feature. If it is agreed, it goes live immediately after the meeting and we watch the result the same day.
That made A/B testing cheap enough to be routine, which finally unblocked sales and marketing — they could run campaigns against a product that moved at their speed instead of a quarter behind it. Alongside it: follow-up and drip campaigns, and SDRs reaching out where the data showed friction.
The AI work that mattered most wasn’t a feature. It was the delivery cycle.
| Before | After | |
|---|---|---|
| MRR | $25K, declining year on year | $100K+ per month, growing |
| Feature cycle | 2–3 weeks | 1 day |
| Brand pricing | $50–60/mo, churn by design | Minimal to none |
| Posting a project | Signup wall, then a form | 2–3 clicks, signup after |
| Factory plans | One expensive tier | Three tiers, free trial, priority bidding |
| Product decisions | Reviewed as designs | Reviewed as working software |
| Knowledge | Split across four departments | One shared base |
| Traffic | Falling, high bounce | 4–5×, bounce down |
The cycle time is the one I care about most. Revenue and traffic measure the outcome; the cycle measures whether the organisation got its ability to change back — and that outlasts any single thing on this page.
The knowledge was never missing. It was distributed in a way that made it unusable, and no roadmap built on a partial picture would have survived contact.
Charging the transient side looks like revenue and behaves like churn. Brands were never going to renew, so every dollar taken from them cost us the project volume the factories were paying for.
The hardest thing to get agreement on was taking money off the table in order to restore liquidity. It reads as a loss on every report until the other side of the market recovers.
Nothing else we shipped mattered as much as collapsing the cycle from three weeks to a day, because it changed what everyone else in the business was able to attempt.
The over-promising stopped when there was a product worth demoing and a cycle short enough to answer requests honestly.
MakersRow was not failing because it lacked features. It was failing because a decade of tech debt had made it unable to change itself, and because its pricing model quietly worked against the liquidity the marketplace ran on.
The sequence mattered more than any single change: consolidate what the company already knew, fix the economics, rebuild the product those economics implied, and then rebuild the way the product got built — so the business could keep moving without waiting on engineering.
The marketplace didn’t need to charge more. It needed to be able to change.