Barrow and Pike reconditions garden tools in Dundee and sells them online: spades with the handles replaced, secateurs sharpened and reground. The shop takes 9,400 visits a month, turns 1.1 in every hundred into an order, and the average order is £34. Those three numbers are the entire business, and their product is £3,516 a month.
Revenue is visits times conversion rate times average order value. Nothing sold online escapes that identity, which means every decision under this topic is a decision to move one of three numbers, and the useful question is never which one matters most. It is which one is cheapest to move from where you are standing.
The same thousand pounds, bought three ways
Suppose Barrow and Pike wants another £1,000 a month. Through visits, that is roughly 2,700 more of them; the cheapest click the shop has ever bought against its own phrases was £0.71, so 2,700 visits cost £1,917 and the thousand pounds arrives at a loss of £917 before a single spade is reconditioned.
Through conversion, it is a move from 1.1 to 1.4 in every hundred. That is 131 orders instead of 103, which at the same £34 is £4,476, and the work is one week spent on the three screens after the basket. Through order value it is a move from £34 to £43.70, which a free postage threshold set just above the average will do on its own. The same £1,000 costs £1,917 in clicks or one week of checkout work, and that comparison is available to anybody willing to spend an afternoon with a calculator, which is roughly nobody.
Which lever is cheap depends entirely on size
The reason the arithmetic above favours the last two is scale, not virtue. Conversion rate is the cheapest of the three to move below about ten thousand visits, because at that size it is still made of specific broken things that one person can find in a week: a delivery cost that appears three screens in, a postcode field that rejects its own valid input, a card form that reloads and empties on a slow connection.
Above that, those have all been found, and the remaining conversion work is statistical rather than diagnostic, which needs volume to read and money to run. At that point buying visits starts to look sensible for the first time. Most shops read the advice written for the second situation while standing in the first, which is the single most expensive misreading in this topic.
The postage threshold, which is the one free lever
A postage threshold moves average order value without touching traffic. Set free delivery at £45 against a £34 average and a measurable share of orders climb to meet it, because a customer holding a £34 basket and a £4.95 delivery charge is being offered £10.05 of tools for £6. The threshold has to sit above the average and within reach of it; set at £75 it is ignored, and set at £36 it gives away the delivery on orders that were already coming.
What it costs is real and it is not the delivery. It is that the shop now posts heavier parcels, and a reconditioned spade is heavy, so the gain is partly eaten by the carrier. Barrow and Pike found the threshold worth £6.10 an order gross and £3.40 after postage, which is still the best return anybody there has found for one afternoon's work.
One month
Four planks across the same thousand pounds
One shop, one target, four routes to it, priced. The planks are not alternatives in principle and they are in practice, because there is one week.
- 01Buy 2,700 more visits£1,917 at the cheapest click the shop has ever bought. Arrives at a loss.
- 02Move conversion to 1.4 in a hundredOne week on the three screens after the basket.
- 03Set free delivery at £45An afternoon, and the carrier takes some of it back.
- 04Sell a second order to somebody who already boughtMonths, and no acquisition cost at all. The slowest of the four to arrive.
Priced in this shop's money at this shop's size. The first plank gets cheaper and the second gets dearer as a shop grows, and they cross somewhere nobody can tell you in advance.
Where this stalls: nobody in the shop has ever paid for anything
The stall in ecommerce is that the owner looks at the part of the shop they are proud of. The product page gets rewritten and rephotographed, because it is the page that represents the work, and the money is being lost on three screens that the owner only ever walks through as a logged-in administrator with a test card, on a desktop, on a fast connection, in a browser where the address is already filled in.
The instruction that fixes it is embarrassing and it takes twenty minutes: place a real order, on a phone, on mobile data, paying with your own card, once a quarter. Not a test order. A real one, with a real address typed in with a thumb. Everything that is wrong with where most of the money is lost is discoverable that way and almost none of it is discoverable any other way.
What to open in 2026
Shopify is the default for a shop this size and it replaced running a store as a plugin on a site that was built for articles; it is a monthly fee with no free tier, and the honest version of the pitch is that the fee buys a checkout somebody else keeps working. WooCommerce is the alternative for anyone already on WordPress, and it is free in the sense that the software costs nothing and the hosting, the payment gateway and the afternoons do not.
For the measurement half, Google Analytics 4 replaced Universal Analytics, which stopped processing data on 1 July 2023, and its funnel exploration is the free way to see which of the three screens after the basket people leave from. It will undercount, for reasons that belong to measuring it, so decisions are not guesses, and it will undercount consistently, which is enough to compare one month against another.
Five clusters, and the first one is the hardest to undo
The decisions that are hard to reverse later is genuinely first, because a platform choice, a URL shape and a product taxonomy are cheap on day one and expensive in year three. Then the page that has to close the sale and the checkout, in that order, which is the order the customer meets them in.
Selling on Amazon and its equivalents is a different business wearing the same coat: the visits are not yours, the customer is not yours, and the margin is set by somebody else. It belongs here because ignoring it is a decision rather than an oversight. Last is where the profit actually is, which is last in the list and first in the arithmetic, because the second order from a customer costs nothing to acquire and every number on this page improves when it exists.
Sideways, the traffic half of the identity comes from being the result someone clicks, without paying for it and from buying attention, and not overpaying for it, and the cheapest repeat orders come through the channel you own, and nobody can take away. All of it sits under getting what you make in front of the people who want it.
The number nobody at Barrow and Pike has questioned is the £34. It is an average across a catalogue that runs from a £9 hand fork to a £180 restored scythe, and averaging those together produces a figure that describes no customer who has ever bought anything. What the shop would find by splitting it is not a better average but two different businesses, and only one of them can afford to buy visits.