Tracking what actually works
Here is the problem, stated once so you can stop fighting it. Sales are reported to you after the fact, added together, split across stores and territories, and stripped of everything about where the buyer came from. The store knows. You do not, and there is no setting that changes that.
So the question "which post sold that copy" has no answer. Not a hard answer — no answer. Every hour spent chasing it is an hour spent on the one thing the chain is built not to tell you. Measure what is in front of you instead.
What you can actually measure
Clicks on links you control. If you send people to your book through a link you own — your own site, a page you host, a redirect you set up — you can add a tag to the end of the address saying where it came from, and use a different tag in each place you post it. Then you know that the newsletter produced four hundred clicks and the forum post produced eleven. That is a real number about a real difference in reach, and it is available on the day.
Newsletter opens, clicks and sign-ups. Your list is the one audience you own outright. Opens are soft — devices and privacy features distort them — but clicks are clicks, and sign-ups over time are the clearest measure of whether you are building an audience or renting one.
The shape of your sales around something you did. Not a single figure, and not attributable. But if the week you ran a promotion sold three times what the weeks around it sold, that shape is real information, and it is roughly the best the reporting allows.
Notice what these have in common: they all stop at your own boundary. The moment a reader crosses into a store, measurement ends, and the honest move is to accept that line rather than buy a tool that claims to see past it.
What you cannot measure, whatever anyone says
Attribution through a store's checkout. A reader sees a post on a phone, thinks about it for two days, and buys on a tablet — or in paperback instead of ebook, or in a different store, or through a library. Nothing carries your tag across that gap. Ad platforms produce attribution figures anyway; treat them as one signal with a known lean, not as truth.
Anything resting on sales rank. A rank is a position in a queue, not a quantity. It moves when other books move. It can improve on a day you sold nothing because the books above you sold less, and it decays on its own schedule. Screenshotting a rank tells you where you stood for an hour relative to strangers. It does not tell you how many copies you sold, and no conversion chart makes it do so.
Keep one file
This is the part that turns scattered numbers into something you can use, and it is unglamorous enough that almost nobody does it.
Open one file. Every time you do something — a price change, a promotion, a newsletter, a podcast, a new cover, a category change, an ad switched on or off — write the date and one line about what you did. Nothing else. It takes fifteen seconds.
Six months later a spike appears in your sales report and you can name its cause, because it is in the file with a date beside it. Without it you get the same spike and a shrug, and whatever caused it never gets repeated.
In the Cover Maker. The thumbnail check shows the cover at the size a search result actually displays it, which is the test most readers apply before they read a word of your description. It is also the cheapest thing on this page to change and re-check. Start a book → · about the Cover Maker
Change one thing at a time
A new cover and a price drop in the same week teaches you nothing. Sales go up, and you now own two changes and one result, with no way to separate them — so you keep both forever, including the one that did nothing, or the one that hurt.
This is harder than it sounds, because doing things feels productive and waiting does not. But a change you cannot read is a change you cannot repeat, and repeating what works is the entire purpose of measuring anything.
Use a run of days, not a day. Sales are lumpy. Compare the week before against the week after, or ten days against ten days. A single day either side of an event will show you noise and let you believe it is a trend.
Give it long enough to appear. Store reporting lags, price changes take time to propagate, and a promotion's knock-on sales arrive after the promotion. Judging a change the next morning is judging the reporting delay.
The two levers worth testing
The cover and the description. Both decide what happens to every visitor your page will ever get, both are free to change, and both can be observed over a few weeks against the run of days before them. Neither costs anything per test, which is exactly what advertising cannot say.
A page that converts badly converts badly for paid visitors too, and you pay for each one. Fixing the cover and the blurb costs an afternoon and applies to every reader from then on, whether they arrived from an ad, a promotion, a search, or a friend who mentioned you. Test those first, write down the date you changed them, and read the result a month later.
Make yours
Everything here is built into the tools, and they are free to use: name your book, bring your manuscript or your artwork, and work as long as you like. You only pay when you want the files without the preview watermark.
Back to: Promo sites and price drops · More: Advertising a book, honestly · How to write a blurb that sells the book · The thumbnail test