Owned Customer Relationships: Why Leakage Kills Startups

An owned customer relationship means the core interactions, payments and records between a business and its customers happen inside its own product. The business can reach those customers again without paying a toll to another platform. When users meet through you and then move to WhatsApp, email or someone else's marketplace, you have lost the relationship and your leverage.

This is one of the tests behind every moat. A product can have users, revenue and a good brand, and still own almost nothing if the relationship lives somewhere else.

Why it matters more for startups than for big companies

Big companies usually own their customer relationships by default. Amazon, Uber and Airbnb sit in the middle of every transaction and spend heavily to keep it that way. A startup has two specific problems.

First, startups often begin as a thin layer of discovery. They introduce a buyer to a seller, and after the first match there is little reason for either side to come back. The a16z marketplace glossary defines this leakage as users who use "the marketplace for discovery, but then complete the transaction outside of the platform."

Second, startups often build on someone else's channel. A shop on Amazon, a bot inside Slack or a community on Discord can grow fast, but the platform holds the customer list, the payment and the rules. When those rules change, the startup has no fallback. That is closely related to platform risk.

A large company can absorb a leak. A startup with thin margins and one acquisition channel usually cannot.

Types of owned customer relationships

Ownership shows up in a few distinct forms. A strong product usually has more than one.

• Owned transactions. Payments run through your product, so you see every order and take your fee automatically.
• Owned communication. Buyers and sellers message each other inside your product rather than by phone or email.
• Owned records. The history (jobs, invoices, reviews, files) lives in your system, not in a spreadsheet or inbox.
• Owned contact data. You hold the customer's email or account and can reach them directly, without paying a platform for access.
• Owned identity and reputation. Reviews and ratings earned in your product only count in your product.

The opposite of each is a form of leakage: payment off-platform, contact details swapped in chat, records kept elsewhere, customers who only exist inside another company's dashboard.

11 examples of owning (or losing) the customer relationship

1. Homejoy (lost it). The home cleaning marketplace shut down on July 31, 2015 after raising nearly $40 million. Economists Andrei Hagiu and Julian Wright describe how its most popular cleaners left once they had enough regular customers, leaving a weaker pool behind. Reporting at the time said it could not hold on to repeat customers.
2. Upwork (fights it constantly). Its annual report lists as a risk factor that "customers circumvent our work marketplace" to avoid fees, and that the losses are "difficult or impossible to measure." In May 2023 it replaced its 5% to 20% tiered freelancer fee with a flat 10%.
3. Airbnb (designs against it). Its off-platform policy requires all communication before booking to stay on Airbnb, and payments for current, future and repeat bookings to stay on the platform.
4. Thumbtack (prices around it). Instead of taking a cut of each job, it charges professionals for leads, so there is no transaction fee to avoid. The tradeoff is that it earns nothing from repeat work between the same pro and customer.
5. Uber (masks contact details). Uber says it uses technology so that neither drivers nor riders see each other's phone number when communicating through the app.
6. Amazon sellers (do not own it). Amazon's seller rules ban marketing emails to buyers and restrict contact to its Buyer-Seller Messaging service. A seller can do millions in sales and still have no customer list of its own.
7. Shopify merchants (own it). A store on Shopify collects first-party data, meaning what customers buy, how they shop and their preferences, directly from its own site. That list is the merchant's to email, segment and retarget.
8. Substack writers (own it, by design). Substack tells writers they will "always own your readers' email addresses and be able to export them at any time". This lowers the lock-in Substack has over writers, but it is the reason many writers trust it.
9. Faire (shares ownership with brands). The wholesale marketplace charges brands a commission on marketplace orders but waives it for retailers a brand brings itself through Faire Direct. It gives up the fee on relationships the brand already owns in exchange for running the orders and payments.
10. Toast (pulls ownership back to restaurants). Toast markets its commission-free online ordering as a way for restaurants to own guest relationships and data instead of handing them to third-party delivery apps.
11. Meta (learned what renting means). Meta's ad targeting depended on tracking users inside apps it did not own. After Apple introduced App Tracking Transparency, Meta's CFO said the iOS changes would be a headwind "on the order of $10 billion" for 2022.

The pattern is consistent. Whoever controls the payment, the messaging and the contact data holds the leverage. Everyone else pays rent.

When leakage is worst

Not every market leaks the same amount. The a16z glossary separates monogamous transactions, where the same buyer and seller transact repeatedly, from polygamous ones, where each transaction needs a new match. Monogamous categories such as home cleaning, tutoring, therapy and long-term freelance work leak the most. Once you have found a good cleaner, you do not need a marketplace to find them again.

Leakage also rises with the fee. The higher the take rate relative to the value the platform adds after the first match, the stronger the incentive to go direct. Upwork notes in the same filing that circumvention is likely to increase in a downturn, when customers are more cost-sensitive.

How a startup builds an owned customer relationship from zero

Hagiu and Wright list the tactics marketplaces use: investing in transaction benefits, limiting communication, charging referral fees, price-parity clauses, seller competition, and hiding sellers who push too much leakage. Most of these are defensive. The durable approach is to make staying inside the product worth more than the fee.

1. Put the work itself inside the product. Scheduling, invoicing, contracts, file sharing and messaging give both sides a reason to stay after the match. A marketplace that only matches is easy to leave.
2. Own the payment. If money moves through you, so do the records, refunds and disputes. Tools like Stripe and Square make this possible from day one.
3. Add protection that only works on-platform. Escrow, insurance, guarantees and dispute resolution apply only to transactions you can see. Airbnb, Amazon and eBay all use this lever.
4. Make reputation non-portable. Reviews, ratings and job history earned inside your product should matter to the next customer. That turns the relationship into a switching cost.
5. Collect your own contact data. Get the email, the account and the login. Do not let a marketplace, app store or social network be the only way to reach your users.
6. Price for the value you add over time. If you only add value at the first match, charge for the match (as Thumbtack does) rather than taxing every repeat job.
7. Design for repeat need. Favor categories where users come back for a new match, or add products that create one.

Common mistakes and fake versions

• Counting a marketplace listing as a customer. Thousands of sales on Amazon or the App Store are not thousands of relationships you own.
• Building a community on rented land. A large Discord server, Facebook group or WhatsApp group belongs to that platform. It is useful distribution, not an owned relationship.
• Blocking without adding value. Hiding phone numbers and banning off-platform deals slows leakage, but Upwork's filing warns that anti-circumvention changes can cause dissatisfaction and even more circumvention.
• Owning the data but not the transaction. You may store the customer's email, but if the work and payment happen by text message, the relationship has already moved.
• Assuming it will come later. "Once we have scale, users will stay" is a hope, not a design. Homejoy had scale and still leaked.

Owned relationships also feed proprietary data: when transactions happen inside your product, you see outcomes that competitors cannot buy or scrape.

How Moat Check scores it

Moat Check asks one question for this test: "Do the relationships and transactions live inside the product, rather than in WhatsApp, Slack, email or a marketplace someone else owns?" The reason is simple: if the relationship or transaction happens on another platform, that platform holds the leverage.

• Pass (10 points). Core interactions, payments or records happen inside the product today, or by design. Example: a booking tool where clients schedule, pay and message the provider in the app.
• Partial (5 points). Ownership is plausible but not yet in place. Payments might run through the product while communication leaks to WhatsApp, or the plan relies on adding payments "at scale." Under the scoring rule, "at scale" is a partial at best.
• Fail (0 points). Users connect through the product and then move to WhatsApp, email or a marketplace someone else owns. A directory that hands out phone numbers is the typical case.

Each score comes with one line of evidence. This is one of 10 tests, each worth 10 points, for a total out of 100. Moat Check uses AI and web research to apply a fixed framework, so treat the result as a starting point for your own thinking, not investment advice. The full method is on the methodology page, and the other tests are on the moats hub.

Check your startup's moat

Questions

Keep reading

• Copycat competitors: how to protect a startup from being copied
• Proprietary Data Moats: When Data Actually Protects a Startup
• The Data Flywheel: When Usage Makes the Product Measurably Better
• All 10 types of moats
• What is a moat in business?