Choosing the marketing mix for a small British online store, on margin and repeat purchase rather than revenue band

The question most small British retailers ask is which channel works best at their size. Revenue band is the wrong input. Three numbers decide which channels can pay for themselves in your store, and all three are already in your own records. Contribution per order, which is what is left after the cost of goods, fulfilment, payment fees and returns. Repeat purchase rate, which is the share of customers who come back inside a year. And whether anybody is already searching for what you sell. Work out those three and the channel decision makes itself, at any turnover.

The three numbers, taken from your own data

Contribution per order

Start with your average order value. Take off the cost of the goods, the pick, pack and delivery cost you actually bear, the payment processing fee, and an allowance for returns based on your own return rate. What is left is the most money you can spend to win one order and still be level.

Do this before you look at a single advertising platform, because it is the ceiling on every channel at once. Most disappointing ad accounts in small retail are not badly run. They are being asked to acquire a customer for less than the market charges.

Repeat purchase rate

Count the customers who bought in a month a year ago, then count how many of them bought again within twelve months. That share is what decides whether you can afford to lose money on the first order.

A store where a third of customers come back can pay more than its first order contribution to acquire one, because the second and third orders carry almost no acquisition cost. A store where almost nobody returns has to make the first order pay on its own, and that single fact rules out several channels.

Existing search demand

Does anybody type a phrase that describes your product. For an established category the answer is yes and the work is to be found. For a genuinely new product the answer is no, and no amount of search work creates demand that does not exist yet.

The decision, in four cases

The fourth row is the one most often skipped, and it is the most valuable. A store in that position that spends a quarter on pricing, bundling or delivery cost usually unlocks more than a quarter of advertising would have.

What each channel is actually good at

• Google Search and Shopping. Captures people who already want the thing. Depends on an accurate product feed with current price and stock, and it is the channel most affected by the quality of your product data rather than your creative.
• Paid social. Creates demand and finds people who were not looking. It is a creative supply problem rather than a targeting problem, so the real question is whether you can produce new material every week.
• Organic search and product content. Slow and it compounds. Category pages usually carry more weight than individual product pages, because a category page can answer a comparison question and a product page can only describe one item.
• Email and lifecycle. The highest contribution channel in almost any store with repeat purchase, and it only works if you are capturing addresses at a rate worth measuring.
• Creators and affiliates. Buys credibility and reach on a variable cost, which suits a thin contribution store. Disclosure rules apply to the content and both sides are responsible for them.
• Marketplaces. Additional demand rather than brand building. Treat the fee as a cost of sale and compare contribution after it, not revenue.

Why the platform's number is bigger than your bank's

Every platform reports the orders it touched, so running three channels produces three reports that can add up to more orders than you shipped. Attributed revenue is a claim about which touchpoint was last seen. Incremental revenue is the difference between running a channel and not running it.

The practical fix in a small store is cheap. Judge every channel on total contribution for the whole business, month by month, against the month before you turned it on. One number, from your own accounts, that no platform controls.

A ninety day test that produces a decision

1. Week one. Calculate contribution per order and repeat purchase rate. Write both down.
2. Weeks two to five. Turn on one channel only, at a budget you can run for three months. One channel, because two tell you nothing about either.
3. Weeks six to nine. Fix the thing the first month exposed. In search it is usually the feed. In paid social it is usually the creative. In email it is usually the capture rate.
4. Weeks ten to twelve. Compare total contribution against the baseline month. Then decide to scale, change or stop, and write the reason down before you start the next channel.

Four British details that change the arithmetic

• Consumer prices include VAT. Advertise the figure a consumer pays. A price that excludes VAT in a consumer advert creates a claims problem and a conversion problem at once.
• Delivery cost is part of the offer here. British shoppers compare delivered price, so delivery pricing belongs in your contribution calculation and in your advertising.
• Returns are a line item, not an afterthought. Your own return rate by category belongs in the contribution figure. Clothing and footwear behave completely differently from hard goods.
• The trading calendar is local. Bank holidays, the autumn peak and the January sale period fall where British shoppers put them, so a test run across a peak is measuring the peak.

Questions about marketing a small British online store

What is a healthy marketing spend as a share of revenue?

Share of revenue is a result rather than a target, and a published average for it would describe somebody else's margin structure. The number to work to is your contribution per order, which sets the most you can pay to acquire one, and your repeat rate, which says how much of that you can spend on the first order.

Should a small store do SEO or ads first?

Ads first where contribution supports it, because they answer the question of whether people will buy at your price within weeks. Organic search alongside it from the start, because it takes months and the work begins with product data you need anyway.

How much should we spend to acquire a customer?

Up to your contribution per order if customers rarely return, and above it where a known share come back, limited by how long you can wait for the cash. Both numbers come from your own records rather than from a benchmark.

Is email worth it at low order volume?

Yes, and the work at low volume is capture rather than campaigns. A store with few addresses has nothing to send to, so fix the capture rate on the pages that already get visitors first.

Do marketplaces cannibalise our own store?

Some of it, and the question is answered by contribution rather than by revenue. Compare contribution per order after the marketplace fee against your own site, and keep the channel where the number holds up.

How often should the mix be reviewed?

Quarterly, with the three numbers recalculated each time. Contribution moves when supplier costs or delivery pricing move, and the decision above moves with it.

The shopping question with no platform behind it

There is one route to your product that none of the channels above reports on. A shopper asks an assistant for the best option for a specific job, with a constraint attached, and gets three brands back. That answer is assembled from readable text, which means it depends on whether your current price, stock and specification sit in plain language on a crawlable page rather than inside a script. AI Knows Us measures which brands those answers name across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, and the product data side of it, in the order the work actually pays, is set out in our note on product data first. The wider counting method behind readings like that is in the 597 reading benchmark.