When I told friends in tech that we were six months in and hadn't built a customer-facing app yet, the reaction was always some variant of: "What have you been doing?"
Onboarding cafés. Writing contracts. Walking through pricing models with bakery owners over coffee in Sehar Commercial. Sending Instagram DMs to brands whose only public contact info was a profile bio. Driving out to DHA Phase 6 to sit with the team at PAUL and explain why their existing French-bakery brand wouldn't be cheapened by a 70% discount on a sealed Loot Bag.
This is the part of building a marketplace nobody likes to talk about.
Supply is the hard side.
Most two-sided marketplaces start with the same instinct: the customer side is the fun part. You can make a beautiful onboarding flow. You can A/B test signup conversion. You can put it on Product Hunt. The vendor side is talking to small business owners on WhatsApp at 11 PM about why your platform won't ruin their margin.
So most early-stage marketplace founders default to building the customer app first. Get the brand polished, get the launch event booked, get the press in. Then, on launch day, scramble to fill the supply side because suddenly you have customers and three vendors.
I watched two food delivery startups in Karachi do exactly this in 2023 and 2024. Both are now dead. One of them ran customer ads while having approximately twelve restaurants on the platform. Customer opens the app, sees a menu of nothing they recognize, never opens it again.
Cold-start failure. Solvable only by paying both sides aggressively to show up, which only works if you have the kind of capital that lets you eat eighteen months of customer acquisition cost. We don't. Nobody in Pakistan does.
The hard fact about marketplaces: customers come and go in an afternoon. A vendor relationship, once you've earned it, is sticky. If you've spent six weeks with the team at HobNob figuring out what their average end-of-day waste looks like, what bag size makes sense for their Sehar branch versus their Phase 6 branch, what their fridge constraints are at 9 PM, what their POS staff need to be told when a Loot Bag pickup walks in — that's six weeks you don't have to repeat. They're on the platform. The flow works. They show up.
A customer who downloads the app, finds nothing in their area, deletes the app — you've lost them, possibly forever. Karachi customers are not patient. They tried Foodpanda, they tried Cheetay, they tried whatever else, and the moment one of them disappointed, the app came off the phone. Once they've deleted yours, getting them back costs more than the original acquisition.
So if you don't have supply when you launch, you're paying twice — once for acquisition, once for re-acquisition. Most early-stage startups can't afford the second one.
The other reason we built the supply side first is that the supply side is the actual product.
The customer experience is "I bought a Loot Bag, I picked it up, it had good things in it." That's it. That is a four-step user flow. The thing the customer is buying isn't the app. It's the bag. The bag is determined entirely by the vendor.
If our vendors aren't aligned on quantity, quality, packing logic, brand standards, refund handling, weekly disbursement schedules, SST treatment, the precise moment a Loot Bag becomes available in the app — the customer experience collapses, regardless of how good the app looks.
We could have built the prettiest onboarding flow in Pakistan and it would have meant nothing if a customer turned up to a partner café and the staff at the counter had no idea what a Loot Bag was.
So we worked backwards. We spent four months doing nothing but vendor interviews, contract drafts, and pricing negotiations. Things we learned in that period that you don't read on a slide:
- Most independent Karachi cafés don't have a public email address. Their websites are either empty or they redirect to Indolj.pk, the listings aggregator.
- Instagram DM is the actual contact channel for most independent F&B brands in this city. We sent over cold DMs across April alone.
- The SECP and SRB tax structure for a marketplace principal-seller model in Pakistan is genuinely complex. You cannot build this product correctly without engaging an actual chartered accountant. We did, and the structure changed twice.
- Most café owners do not think of their end-of-day waste as a number. They think of it as "what we have left." Helping them see it as inventory was a real, slow, in-person conversation. It happened once per vendor.
The result: when we went live in April, we had 8 signed vendors. Contracts were already in place. Pricing was negotiated and locked. Branch staff had been trained. The first Loot Bags went out from day one.
That's not glamorous. There was no launch event. No "1,000 signups in 24 hours" tweet. No press release. We told a small group of friends, the vendors told their regulars, the app went live, and bags started moving.
This is the boring version of a startup launch. It works.
The version that doesn't work — the one I've watched at least four Karachi startups try in the last two years — is "build a beautiful customer app, run paid Instagram ads, hope vendors come on later because we have user numbers to wave at them."
Vendors do not care about your user numbers. Vendors care whether you can sell the specific Loot Bag they're going to pack at 8:55 PM on a Wednesday. A user number on a deck does not sell that bag. A clean contract, a 24-hour disbursement record, and a phone you'll pick up when something goes wrong — those sell that bag.
We built that first. The customer app came after. By the time we wrote a single line of customer-side code, we already knew what the customer was going to buy, who was going to make it, what time it would be packed, where it would be picked up, and at what price.
You can call this "boring sequencing." I'd call it the only sequencing that actually works in this market.