Decide what the promotion needs to teach you
A placement delivers 120 clicks. Your book records six orders that week. Did the promotion generate all six, some of them or none? Without a defined tracking method, the honest answer is unknown. Traffic, total orders and attributed orders are different records; putting them in adjacent spreadsheet columns does not establish causation.
Start with one title, one edition, one marketplace and one offer. Write the decision before booking: “Is this newsletter worth testing again at this price?” or “Does this social post bring readers to the sample?” A test can answer a traffic question without proving profitable sales, but its success condition must say so.
Record the placement cost, creative version, destination, send date, active dates and planned review date. Choose a maximum cost you can afford for this experiment, not a universal budget borrowed from another author. Note other activity running at the same time: an existing ad campaign, a launch announcement or a price change can complicate the result.
Use a tracking method that can see the destination
For your own author website, Google Analytics campaign parameters identify the source, medium and campaign behind a visit. Use a consistent naming scheme and confirm that your site actually records those fields. Adding a label to a URL does not install analytics or create access to another company's checkout.
A fictional landing-page URL might use ?utm_source=genre_newsletter&utm_medium=email&utm_campaign=harbor_test_01. Open the final published link, not just your draft. Confirm the intended page, marketplace and edition after any redirects. Do not put reader email addresses or other personal data into campaign parameters.
If the destination is Amazon, ordinary UTMs do not give your website analytics access to Amazon purchases. Amazon Attribution provides eligible KDP authors with measurement for non-Amazon marketing. Confirm access and marketplace/product eligibility in your advertising account, then create separate tags for the placements you need to distinguish. Keep Amazon's generated tracking link intact and verify the destination before sending it.
If Attribution is unavailable, record the source's click report and the book's total orders separately. Mark attributable orders as unknown. You can investigate timing and delivery, but you cannot repair missing attribution by assigning every order that week to the campaign.
Do not turn every counter into a conversion rate
| Record | What it can show | What it does not establish |
|---|---|---|
| Publisher-reported clicks | Clicks counted under that publisher's definition. | Unique readers, site sessions or purchases. |
| Author-site sessions | Sessions your analytics observed on your own site. | Every person reached, or subsequent Amazon orders. |
| Outbound store clicks | Recorded attempts to open the retailer link. | That the destination loaded or the person bought. |
| Attributed orders | Orders credited under the reporting tool's rules. | The number of incremental orders caused by the promotion. |
| KDP total orders | Recorded orders within the selected book/report filters. | Which outside placement generated each order. |
For any ratio, name the numerator, denominator and source. Repeated clicks, multiple orders by a buyer, incomplete tracking and different attribution windows can affect the calculation. Do not divide all-book KDP orders by one newsletter's clicks and call the result that newsletter's conversion rate.
A fictional comparison: cheaper clicks do not settle the decision
Suppose two outside placements have mature, separate attribution reports for the same book and edition. For this simplified teaching example, each attributed order contains one copy. Both placements cost $40; the author supplies a fictional net royalty of $2.40 per copy. That royalty is an input, not a rate inferred from the book's retail price.
| Measure | Placement A | Placement B | Placement C |
|---|---|---|---|
| Promotion cost | $40 | $40 | $40 |
| Recorded tag clicks | 120 | 80 | 60 publisher-reported clicks |
| Attributed one-copy orders | 6 | 8 | Unknown; no attribution |
| Cost per recorded click | $40 / 120 = $0.33, rounded | $40 / 80 = $0.50 | $40 / 60 = $0.67, rounded |
| Attributed orders / tag clicks | 6 / 120 = 5% | 8 / 80 = 10% | Not calculable |
| Attributed royalty contribution | 6 × $2.40 - $40 = -$25.60 | 8 × $2.40 - $40 = -$20.80 | Unknown |
A brings more recorded clicks for the same cost, while B has the better observed order-to-click ratio. Neither covers its placement fee with the attributed royalties in this example. The contribution calculation excludes other costs, taxes, returns and unmeasured downstream value; it is not a full business profit statement. Unknown attribution for C is not zero orders.
Eight orders versus six is too little evidence to declare B a permanent winner. The placements may reach different audiences, and attribution does not supply a randomized control group. One additional order changes A's ratio by about 0.83 percentage points and B's by 1.25 points. Preserve the counts and uncertainty instead of presenting 10% as a reliable forecast.
Review mature reporting, not just today's numbers
Amazon's current Attribution documentation describes a 14-day attribution window. That is a measurement rule, not a promise that every sale appears immediately. Set the final review after the last relevant click has had time to complete the applicable window, then allow for reporting updates. Keep an earlier operational check for broken links or incorrect pricing.
KDP's reporting documentation notes different processing timings for eBooks, shipped print orders and Kindle Unlimited reads. Its advertising help also distinguishes all-book KDP reporting from ad-attributed reporting. Export matching book, format, marketplace and currency filters; record each report's time zone, date basis and download time rather than expecting unlike totals to match.
A sale that appears later does not automatically belong to this test. Conversely, a temporarily empty report does not prove that a placement failed. Keep the raw exports and annotate later revisions instead of silently overwriting the first snapshot.
Separate attribution from incremental lift
An attributed order meets a tool's crediting rules. An incremental order is one that would not have happened without the promotion. Those are not interchangeable. A reader might already know your book or encounter another campaign before buying. A normal author campaign log cannot reconstruct that counterfactual for each person.
Likewise, a before-and-after comparison is affected by weekends, launch timing, price changes and other traffic. Treat it as context, not a controlled experiment. Do not add overlapping reports from different platforms and call the sum unique sales. Keep platform totals separate unless you have a defensible deduplication method.
Choose the next action from the weakest observable step
- Wrong or broken destination: stop using the link and correct it before interpreting performance.
- Delivery is unclear: ask the placement provider for its actual delivery/click definitions; do not substitute promised reach.
- Clicks but unclear purchase outcomes: improve measurement before increasing the spend.
- Clicks with weak mature order results: inspect edition availability, audience fit and the book's promise before rebooking.
- Encouraging but small results: repeat within a defined allowance while preserving the same measures; do not assume the ratio will hold at a larger scale.
If the uncertainty is the book's presentation, write a focused reader-feedback brief. For selecting an outside placement, compare promotion sites and their actual conditions; for the wider plan, use the marketing checklist.
Virtalibry does not provide an integrated Amazon sales-attribution dashboard. Keep campaign measurement in the relevant external reports and your own log. Return to the blog for worked editorial examples. Sources checked on 16 October 2026; all campaign figures are fictional and are not promised results.