When a New Sales Channel Is Actually Working

A new sales channel has a flattering first month. There are screenshots, orders, and a reason to tell the team that the next phase has begun. The difficult question comes later: would you buy the same month again at the same cost?

I ask that before calling a channel a success. Not because early orders are meaningless. They tell you that someone can be persuaded to buy. They do not yet tell you whether the business can afford to keep persuading them.

Perfume Lounge taught me how easy it is to confuse a working campaign with a working model. The distinction matters whether you are testing paid social, a marketplace, a retail counter, a reseller, or a new city.

Define the test before the first order

Before spending, decide what would count as a result. Which customer are you trying to reach? What is the offer? How long will you run the test? What is the most you can spend? Which number would make you stop?

If those questions are unanswered, the test will continue as long as there is a promising screenshot. Founders are very good at finding one. I know because I have done it.

The test also needs a comparison. A launch discount can produce sales without proving that customers want the product at its normal price. A marketplace can move inventory while leaving you with a thin margin after fees and promotions. A retail placement can look exciting while the goods are still sitting with the retailer. Write down the price and terms you intend to use after the test. Otherwise you are measuring the launch offer, not the channel.

Follow a cohort to the end

Choose the customers acquired in a particular week and follow those orders through delivery, returns, refunds, and settlement. Keep the acquisition spend with that same group. Then look at net revenue and contribution after the costs of the channel.

For paid acquisition, this means the ad bill belongs against the customers it brought, not against total monthly sales. For a marketplace, include its fees, shipping arrangements, promotion costs, and the timing of its payout. For retail, include the margin given to the store, staff time, display costs, unsold stock, and the possibility that a sell-in to the store is not a sell-through to the customer.

If the data is too messy to trace a cohort, that is useful information. It means you cannot yet scale with confidence. Fix the measurement before spending more to magnify the uncertainty.

Ask what the channel took from elsewhere

Some "new" revenue is an old customer choosing a new door. A person who would have bought from your own site may buy through a marketplace because the listing is more convenient or the price is lower. Your total orders rise less than the channel's order count suggests, and your margin can fall.

You cannot know the exact counterfactual for every customer. You can look for clues: repeat customer share, geographic overlap, branded searches, order timing, and whether direct sales change when the new channel runs a promotion. Do not claim every marketplace order is incremental. Do not assume none are. Keep the uncertainty visible.

Capacity is another cost. If a new channel ties up stock or support attention that would otherwise serve a better channel, its apparent contribution overstates its value. A small team can be busy and still be moving backward.

Separate the first order from the second

People often defend a loss on the first order by promising repeat purchases. Sometimes that is a sensible investment. But "they will come back" is a hypothesis until a cohort does.

Track how many customers place a second order, how long it takes, and what you spend to bring them back. Look at the share who return without another large discount. If the channel brings bargain hunters who only buy when you subsidise the price, the second order may be just as expensive as the first.

This is why I would not scale a channel on a projected lifetime value borrowed from the whole business. Customers from different channels behave differently. Use the behaviour of this channel's customers, even if the first answer is a range rather than a precise number.

Decide what you are buying

At the end of the test, I would write a short decision memo. How many retained customers did we acquire? What did they contribute after channel costs? How much cash did the channel tie up, and for how long? What happened to other channels? Which assumptions are still guesses?

Then choose one of three paths. Scale it with a spending limit and a date to review again. Change one variable and run another small test. Or stop.

Stopping is not a verdict on the team that ran the experiment. It is the point of running an experiment. The expensive mistake is to keep paying for an answer because you liked the first week.

A channel is working when you can explain why it should keep working after the launch offer ends, the returns arrive, and the full invoice is on the table. That answer is less exciting than an order graph. It is also the one you can build a business on.