Shopify Experts Marketplace

One of my first projects at Shopify was making it easier for partners and merchants to manage store access—limiting permissions, monitoring activity, that kind of thing. The feature took off like a weed. What struck me wasn't the feature itself, but what the data revealed: there was an enormous amount of collaboration happening. Partners logging in, doing quick tasks, logging out.

I started looking at what these accounts were actually being used for. Quick jobs. Login once, set up SEO, configure AdWords, update product descriptions, fix images. Menial, small tasks. That was the insight: there's real demand for help, but it's about micro-tasks, not long engagements.

This is really a story of the product that should have been versus the product it became.

The original insight was sharp: merchants need quick, fixed-price tasks in the $50-500 range. I built the first version with one or two devs over a Christmas holiday. We launched it, it worked, and it should have stabilized within six months with maybe a couple of engineers maintaining it.

But any company that sees initial success looks at it and asks, "Why can't this be bigger? Why don't we do more with this?" And that's how Experts Marketplace eventually became this Upwork-like system—which is a completely different undertaking. You're dealing with disintermediation, money transfers in the thousands of dollars, misaligned incentives at scale. The original insight got buried.

Merchants usually did not know what they wanted. The questionnaire's job wasn't just intake—it was translation. It extracted just enough information to give a partner a well-scoped task.

Without that questionnaire, partners would get "I need more traffic" or "I need help"—completely unactionable. The branching logic asked: What stage is your business? Have you tried these things? How much detail can you provide? Depending on the answers, we'd route them to an AdWords task or a product descriptions task or something else entirely.

The value was converting vague merchant pain into a specific deliverable. That translation work would otherwise require a sales call.

We started with trusted partners, many from Ireland where Shopify had a large ecosystem presence. The other factor was willingness to do small jobs—which skewed toward partners in India, Indonesia, places where labor costs were lower but quality could still be high if we curated well.

As we expanded, we built reputation systems. Reviews, response rates, close rates—all factored into who got matched more often. We also had an anti-disintermediation lever: partners who took things off-platform got matched less frequently because we had no data on them. You can't punish people for going off-platform, but you can reward staying on by having better data to match them.

I was pretty junior, so it took me a while. The moment I remember clearly was a meeting with the CPO, maybe two years into the investment. The CPO asked, "Why is this a $10 million business and not a $100 million business? Is this going to be the same scale as the App Store?"

Everyone in the room wanted to say yes. My internal compass was saying no. I was too junior to voice it, but that was my moment of realization.

The thing is, the realization should have come earlier. A marketplace is fundamentally a broker of trust. Where is trust needed? In small tasks—you need someone to vouch that this person can do a quick job competently. But for larger tasks, trust isn't brokered by a platform. It's built through meetings, project management, ongoing relationships. A platform focused just on that initial connection is never going to be sufficient for either side. A directory would be good enough for discovery—but that's not a business model.

Almost all of it. The $100-500 range was where everything happened. We managed to push slightly beyond that, but the ceiling was around $500. Once you got into the thousands, a 10% marketplace fee stopped making sense for partners. If you're paying $100 in fees on a $1,000 job, that feels unreasonable—especially when you've lost the repeat relationship.

And the $5,000-10,000 jobs where real money lives? Those were never transacted on the marketplace. The core insight—micro-tasks are what merchants need to get off the ground, and those are what's liquid in a marketplace—that was carrying almost all the weight.

When something has initial success, everyone gets bought into wanting it to keep succeeding. From VP level down to individual PMs, everyone's trying to prove the reason for this thing to continue existing—and to continue existing at the scale it's at. There's always desire to get more, or to pretend it's more successful than it is.

The biggest lesson wasn't about scope expansion specifically. It's that every idea, every step of the way, has to be evaluated on its own merit. You can't let past success validate future ideas. The initial insight might have been good, but that doesn't mean the next idea is good just because the first one was.

If something fits into a longer-term picture—if even a losing investment pays off downstream—that's different. But Experts Marketplace didn't fit a longer-term picture. It was purely about revenue. And when that's the case, each investment needs to stand on current merit, not the borrowed credibility of past success.

I wish I'd had the experience to see that earlier. The CPO was right to ask the question. I just wasn't senior enough to give the honest answer.

Fixed-price micro-tasks only. The $50-500 range. Built by one or two devs over a holiday, stabilized within six months, maintained by a small team indefinitely. We'd expand the types of tasks and maybe the pricing tiers, but tightly scoped.

The insight was real: new merchants need quick help getting off the ground. That need is ongoing—there are always new merchants—but it's not a growing need per merchant. For ongoing support, you want a directory structure, not a transactional marketplace.

A $10 million business that runs lean and serves a real need? That's a good outcome. The mistake was trying to make it something it was never going to be.