Improve KDP book discoverability by matching the book's real contents to the language its intended readers use. Then test a bounded change and record what you can actually observe. Metadata is a way to communicate relevance; it is not proof of demand or a guarantee of rank.
You will finish with a metadata sheet and a promotion experiment log. Start with a quality-checked edition, current listing details, dated royalty records where available and a budget you can afford to spend without recovering it. Allow an initial research session and a defined observation window; the right window depends on your traffic and publishing context.
1. Establish what the book can honestly promise
List the reader, situation and useful result before opening a keyword field. A beginner workbook for renters with shaded balconies should explain those constraints in reader language. It should not imply professional horticultural training, structural assessments or guaranteed harvests.
Read the description alongside the sample and contents page. Identify where each promise is delivered. A discovery campaign that attracts the wrong readers can produce disappointment even when it increases visits. Repair a misleading promise before measuring how many people respond to it.
If the edition still has unresolved navigation or layout defects, return to the first-title quality checklist. Otherwise you will spend scarce attention promoting a reading experience you already know needs work. Keep quality repairs and discovery changes in separate logs so their effects are not confused.
2. Gather reader language without copying competitors
Collect phrases from reader questions, public reviews, library classifications and comparable descriptions. Note the source, date, reader context and whether the phrase describes a problem, desired result or format preference. Paraphrase your findings and avoid copying protected expression.
For the balcony example, your research sheet could distinguish “plants for low light,” “small-space container plan” and “beginner watering routine.” Each phrase needs a content match. If the book does not cover indoor plants, a popular indoor-plant phrase is an irrelevant lure, not an opportunity.
Include conflicting evidence. Readers asking for a comprehensive species encyclopedia are expressing a different need from those seeking a short planning workbook. Decide which need your edition serves rather than filling metadata with every related phrase.
Amazon recommends keywords that accurately describe the book and reflect customer search language. Its guidance also identifies terms to avoid, including misleading or inappropriate metadata. Check the current rules before saving your selection. KDP keyword guidance.
3. Write a metadata sheet with a reason for each choice
Keep your title and subtitle, description, keyword concepts and category rationale together. Add the exact edition and date. A saved sheet lets you compare versions and recover if a later edit becomes less accurate.
Use a description structure that answers three questions: who is this for, what will they be able to do, and what is included? A plain description of exercises is more useful than a string of adjectives. Mention genuine prerequisites or scope constraints that materially affect suitability.
Here is a hypothetical evidence-to-copy map:
| Reader evidence | Actual book coverage | Metadata decision |
|---|---|---|
| Renters struggle to assess available light | Seven-day light observation sheet | Describe planning around observed conditions |
| Beginners ask what to buy first | Container and supply checklist | Mention the shopping worksheet |
| Some readers want advanced plant pathology | No specialist disease reference | Exclude that promise and those targeting terms |
| Readers need a manageable routine | Weekly care-planning exercise | Explain the routine without promising results |
Choose categories for genuine subject fit. Do not chase an unrelated low-competition category or add another author's name to attract their audience. A ranking gained through mismatch does not establish a durable reader proposition.
For a keyword update, open the book format’s details from its Bookshelf menu, edit the keyword fields, then follow Save and Continue through to Publish on the pricing page. Merely saving the sheet does not submit an update. Check the live listing after processing and record when the change became visible before starting the observation window. KDP keyword update steps.
4. Define one bounded experiment
Write a hypothesis that links a specific change to an observable result. For example: “A clearer description of the light-observation worksheet may attract readers who want a planning workbook rather than a plant encyclopedia.” The change is the description; the mechanism is better expectation-setting.
Record the baseline version, change date, observation window, budget ceiling and a stop condition. Avoid changing the cover, description, price and promotional channel simultaneously if you want to learn which element mattered. If a major quality repair must happen at the same time, make it and label the experiment as confounded.
Choose the evidence you can actually obtain. Do not invent organic search impressions, customer-level attribution or a complete funnel when your available reports do not provide them. Log royalties, orders or other available platform measures using their actual definitions, and distinguish them from external page visits or reader replies.
A test can end with “insufficient evidence.” That is useful when the alternative is treating three purchases as proof of a scalable channel. Set the next observation decision before seeing the result so enthusiasm does not move the goalposts.
5. Use promotion channels that fit the reader and platform
Choose one channel where the intended audience already asks the relevant question and promotion is permitted. This might be an owned newsletter, an appropriate community with explicit permission or an eligible platform tool. Explain the book's practical use and show an original sample rather than manufacturing urgency.
Review the conditions for any KDP promotion you propose to use, including applicable enrollment or format requirements. Availability in a help article does not prove your title qualifies. Keep the actual eligibility check and start/end dates in the experiment log. KDP promotional resources.
Do not commission favourable reviews, reimburse readers for positive ratings or hire a provider to manipulate platform programs. A contractor's promise to “handle everything” does not remove the publisher's responsibility. Ask what the provider will do, where traffic comes from and whether the method follows current rules; stop when the answer is vague or depends on artificial engagement.
You can ask readers for honest feedback through permitted methods without scripting approval. Separate private usability feedback from public customer reviews. A free sample reader's comments can inform revision, but they should not be presented as a verified customer testimonial without an accurate basis and appropriate permission.
6. Complete the experiment log
Consider a hypothetical author who spends $80 in cash and four hours promoting a workbook through one permitted channel. During the observation window, recorded royalties are $110 and ongoing software allocated to the project costs $10. The royalty-based result for those stated items is $20 before production recovery, tax and other expenses. It is a cash result only if the $110 has actually been received and the $90 of expenses paid in the same accounting window; keep report-period royalties and bank receipts in separate columns.
If those four hours are valued internally at $25, the same exercise consumes another $100 of owner work. The resource result becomes negative $80 on that basis. This does not make the experiment useless; it shows why royalties, cash contribution and compensation for work should not be used interchangeably.
Do not attribute all $110 to the promotion unless the evidence supports that attribution. Existing discovery, other reader activity and timing may contribute. A before/after change is an observation, not automatic causation.
| Log field | Completed hypothetical entry |
|---|---|
| Version | Description D2; cover and price unchanged |
| Hypothesis | Clearer workbook promise improves reader fit |
| Channel | One permitted community placement; new channel plus description change, so effects are confounded |
| Dates and baseline | Hypothetical baseline: October 1–14; observation: October 15–28; actual prior-window royalties not supplied, so no lift calculated |
| Ceiling and stop rule | $100 promotion cap; stop earlier for a rights, quality or platform-compliance defect |
| Spend and effort | $80 cash; four hours |
| Available result | $110 recorded royalties in the window |
| Attribution limit | Other discovery continued; causal lift unknown |
| Decision | Keep accurate copy; pause and reconcile reports. Only $20 of the promotion cap remains; no causal or scaling conclusion |
7. Diagnose the failure before buying more attention
If the promotion reaches few suitable readers, inspect the channel and message before rewriting the book. If people misunderstand the offer, repair the description or sample. If readers buy but report a missing promised result, return to the content and quality process.
If no reliable attribution exists, improve the log rather than claiming a conversion rate. Use dates, channel identifiers and available reports without collecting unnecessary personal information. Preserve original exports so later comparisons are not based on memory.
Stop the campaign when spending reaches the ceiling, the method becomes noncompliant or reader feedback exposes a material defect. Do not keep spending merely to make the launch look successful. A paused experiment with a clear repair action protects cash and reputation.
For an existing catalog, repeat this work title by title before assuming a catalog-wide promotion will help. Separate a temporary sales event from ordinary performance and record any ongoing workload. These records later support the catalog operating dossier, but they do not establish a sale price or account-transfer permission.
Should I replace accurate wording if sales remain low?
Only when you have a reason to believe the wording obscures the book's value. A low-volume observation may reflect limited reach, seasonality, audience mismatch or weak demand. Keep accurate scope and test one plausible cause.
Does a proof or author-copy order belong in the sales result?
No. KDP explains that those orders do not generate royalties or appear in KDP Reports. Record them as production or distribution costs, not independent demand. Proof and author-copy rules.
What is the next useful step?
Keep the metadata version that communicates the book honestly, close the experiment with its limits and choose one affordable repair. Once you have repeatable operating records, use the digital-asset business plan to decide whether to maintain, expand or stop the catalog.

