Kestrel Bindery in Leeds sells one thing online: a kit for repairing the spine of a hardback. It goes out at £34, the materials and the box cost £11, and the postage costs £4. Profit per sale is the price minus what the thing cost and what it cost to send, which makes it £19, and the two people who own the shop had been trading for two years before either of them did that subtraction.
That £19 decides which of the eight topics under this pillar are open to Kestrel and which are shut, and it decides it before anybody opens an ad account. The expensive mistake here is not a thin page or a badly written ad. It is choosing the channel first, because the channel is the only decision on this page that can be made while knowing nothing.
Only four of the eight topics here are channels
Strategy, content, ecommerce and analytics are not places to be seen. They are the decision about who is being sold to, the supply of material to send, the page where money changes hands, and the count afterwards. SEO, paid ads, social and email are the four that carry something to somebody, and they are the only four an outsider would recognise as marketing at all.
What separates those four from each other is not reach. It is what happens when you stop. A page that ranks keeps arriving for months after the last edit, while a channel you rent stops the day you stop paying for it, and paid search is the pure case: the traffic ends the hour the card is declined. Social sits closer to rented than most people believe, because the platform decides how many of the followers you collected are shown anything. Email is the only one of the four where the list is a file you hold.
Of the eight pillars on Knowledge has a shape, this is the one where nearly every competing page was written to sell the tool it recommends, which is worth knowing before reading any of them.
The order that cost Kestrel a quarter
The order that works starts in the same place every time: what a sale leaves behind, then who buys, then what to say, then where to say it. Kestrel ran it backwards. Somebody suggested the shop should be running search ads, an account was open by lunchtime, a page was written to catch the clicks, and £447 went out over a month. The question of what a sale was worth arrived after the money had.
Run forwards it is arithmetic and it takes ten minutes. Break-even cost per click is profit per sale divided by the visits one sale takes. Kestrel's kit page turns one visit in fifty into an order, measured across a quarter rather than across a good week. £19 in fifty visits sets the ceiling at £0.38 a click. The cheapest click the account would quote against the phrase people actually type was £0.71. Paid search was shut to this shop before a single ad was written, and no amount of writing could have opened it.
The cost is not in any one of those steps. It is that the channel gets chosen twice: once at the start with nothing behind it, and once at the end with the £19 in front of you, and only the second choice had any information in it. Everything in between was work done in service of a decision that had not been made yet.
The wrong order
The route Kestrel took, drawn as a route
Six stops, and the sixth is the same decision as the first. Most shops walk this route more than once, with a different channel each time.
- 01Pick the channelSomebody suggests search ads, and that is where the work starts
- 02Build something to sendA page written to catch clicks, before anyone has said who is clicking
- 03Buy the clicksA month, and an account that will happily spend more
- 04Read the reportSessions up, and the number underneath not looked at yet
- 05Work out what a sale is worthNineteen pounds, and it was nineteen pounds the whole time
- 06Pick the channel againThe same decision, now taken with the arithmetic in front of you
Nothing in the first five stops was done badly. That is what makes this route hard to see from inside it, and what makes £447 a normal price to pay for it.
Traffic went up, orders did not, and the report led with traffic
This is where people stall in marketing, and it is not a failure of effort. It is measuring the wrong thing and then getting steadily better at it. The metric is sessions. Sessions is the one number every channel reports the same way, which is why every monthly report gets built around it, and it is the one number in the shop that cannot tell you whether anything worked. Kestrel's sessions went from 2,100 to 7,400 across the quarter. Orders went from 34 to 31.
Each topic below has its own version of the same trap, and naming yours is most of the cure. In search it is a ranking for a query nobody buys after. In email it is an open rate on a list that has never ordered. In social it is a follower count, a number the platform can devalue on a Tuesday afternoon. In content it is pieces published. Every one of them is real, every one is easy to move, and moving any of them is perfectly compatible with a shop that is quietly dying.
That month is worth writing out in full, because the shape repeats everywhere. Six hundred and thirty clicks at £0.71 came to £447. At the page's own best rate, one visit in fifty, those clicks were worth twelve orders. Twelve kits is £408 in, against £132 of materials, £48 of postage and the £447 of clicks, which is £627 out. The month finished £219 short, and that is the generous reading: it credits the ads with twelve orders while the shop as a whole sold three fewer than the month before, so most of the twelve were sales that were arriving anyway. The sessions chart over the same month had never looked better, and that chart was at the top of the report.
One month
Where a month of paid search sent the money
Every pound that left the shop that month, sorted by who received it. The largest supplier Kestrel had was not a supplier.
One month at Kestrel Bindery, in pounds out.
- The ad platformSix hundred and thirty clicks£447
- Materials and boxesTwelve kits, at eleven pounds each£132
- PostageTwelve parcels, at four pounds each£48
It leaves out the hour a day the account took to run, which is the largest line in a two person shop and never appears on a chart like this.
What to open in 2026
Google Search Console is the free floor, and the only place your own queries exist at all. It replaced the keyword data Google removed from analytics in 2011, and nothing has taken its place since. The thing people misread is the position column: average position in Search Console is an average over impressions, not a rank, so a figure of 8.4 can mean third for half the queries and twentieth for the rest. It also knows only the searches you were shown for. The ones you never appeared in are not missing from the report. They were never in it.
Google Analytics 4 does the counting. It replaced Universal Analytics, which stopped processing data on 1 July 2023, so every tutorial older than that is describing a product that no longer runs, and it is free at the size Kestrel is. The error people make is treating a session as a person and a channel label as a fact. Analytics knows about the visits it was allowed to see, and after consent banners and blockers that is a sample with a slant in it rather than a census.
Ahrefs Webmaster Tools is the honest answer for a shop with no budget: free for a site you have verified, and it replaced paying for competitor data you did not need yet. What people get wrong there is the difficulty score, which is mostly a count of the links pointing at whoever ranks now, and that is a statement about the past rather than about the page you are going to write.
For email, the useful advice names the category rather than the product, because these change hands and change price often: any service that will let you export the list as a file. You own a list exactly as far as you can export it. Free tiers usually end somewhere between one and a few thousand addresses, and the export is worth testing on the first day rather than on the day the pricing changes.
The first of the five you cannot answer is where you are
Nothing about getting better at this feels like anything, and the standard substitute is counting the work: posts published, ads written, reports sent. It is checkable instead. What one sale leaves behind is a number, £19 at Kestrel, and most people running a shop have never once taken the postage off it. What the page does with a visit is a second number, one in fifty. Those two hand you the third for nothing. The fourth is the one you write down before you start rather than after: the cost per sale above which the ads get switched off, agreed while nobody is losing anything yet.
The fifth costs a week of sales to answer honestly, and it is the one that separates somebody three months in from somebody one month in. Which orders would have happened anyway takes a week with the ads off to answer. Kestrel ran that week in July 2026: seven kits with the ads off, against eight in the week before with them on. The account had been claiming all eight. Deciding which touch gets the credit is the argument that starts there, and it does not have a tidy ending.
The signal
Five answers, and the ground is not flat
Four of these are answered by looking something up. Only the last one changes what you would do about the other four.
- 01What one sale leaves behindNineteen pounds at Kestrel, after the materials and the postage
- 02What the page does with a visitOne in fifty, over a quarter rather than over a good week
- 03The highest click price that still paysThe first answer divided by the visits a sale takes
- 04The number at which you stopWritten down before the first click rather than argued about after
- 05Which orders would have happened anywaySeven kits with the ads off, against eight with them on
Answering the fifth moves the third. If most of the orders were arriving anyway, what an ad actually adds is less than nineteen pounds, and the ceiling comes down with it.
Eight topics, and which one you are standing in
Almost everyone arriving here is standing in one of two places. If nothing has been settled yet, the decisions that come before any channel is the first topic and the shortest route out of wasting a quarter: who it is for, what it is worth, and what it is not. If something already sells and the question is how to get more of it, the choice is among the four channels, and it turns mostly on what you can afford to wait for.
Being the result someone clicks without paying for it is the slowest of the four and the only one that compounds. It also changed more than the others between 2024 and 2026, because a question answered inside the results page is a question you will not get a click from. Buying attention, and not overpaying for it is immediate, honest about its price, and shut to anyone whose arithmetic looks like Kestrel's. Building an audience on someone else's platform is the one where the landlord sets the rent after you have moved in. The channel nobody can take away is slow enough that it has to be started before it is needed.
Under all four sits writing the things that get found and read, which is what the channels are carrying, and after them the details that decide whether selling online works: the page where money changes hands, and where a shop like Kestrel loses more sales than any ad ever brought it. Measuring it, so decisions are not guesses comes last in the list and should come second in the order, because a decision you cannot check afterwards is not a decision.
For anyone here because the shop is theirs rather than a client's, the neighbouring subject is building something that runs without you selling your hours, which is a different problem sitting on the same arithmetic. The page where the money changes hands is a design question before it is a marketing one, and understanding what people need before designing anything is where that gets taken seriously. The supply problem has an obvious 2026 answer that mostly does not work, and making images, video, audio and text with a model is worth reading for what it costs you as much as for what it saves.
What is left is the question Kestrel should have asked before opening the ad account, and it is not which channel. £19 is what one kit leaves behind, but the bindery has customers who send four books a year for rebinding and have done for a decade, and for them £19 was never the number. Which of those two figures a channel gets judged against decides everything above, and that argument belongs at where the profit actually is.