An earlier pillar covered the launch quarter, the twelve or so weeks when attention is cheapest and most authors spend everything they have. This pillar covers everything after that window closes, which is where the majority of a title’s commercial life gets lived. The months and years that follow determine whether a catalogue keeps earning or quietly stalls.
What follows looks at what ongoing maintenance actually consists of across a year and beyond, why the returns arrive slowly, and three specific decisions worth tracking in detail: metadata revisits, series consistency and post-publication correction, the same categories Perivane’s published audits of backlist titles document in practice.
The mechanism behind this work is straightforward. Metadata fields age because the market around them keeps moving, categories get restructured, comparable titles reposition, and reader language shifts, so a record that was accurate at launch drifts out of alignment even though the book itself hasn’t changed. Correcting that drift restores visibility, and visibility is a precondition for sales even when it isn’t a guarantee of them, as the eBOUND Canada study found when it measured page-view gains without a matching sales lift. The sections below walk through what drifts, why series metadata deserves its own attention, how to catch errors before a reviewer does, and why the payoff for all of this shows up slowly enough that most authors quit doing it.
What Maintenance Actually Means After the Launch Quarter Ends
How to keep selling books after launch is a question about upkeep, and upkeep has three components that recur on different clocks, each requiring a return visit on a schedule the author sets rather than one imposed by an algorithm.
The first component is the retailer record itself: categories, keywords, description, series fields, contributor details. These fields are editable at any time, but most authors edit them once, at launch, the way Perivane’s audit of The Murder at the Vicarage found a decades-old category assignment left untouched since its original release. The second component is the text of the book, where typos, broken links, outdated front and back matter, and a preorder link pointing at a page that closed eighteen months ago all quietly accumulate. Third comes the internal consistency of a body of work: whether book two contradicts book one, and whether both point at each other correctly in their descriptions and back matter.
It helps to be precise about what the metadata fields actually do. Amazon’s published guidance describes keywords and browse categories as inputs that help customers find a title in search and browse results. That is a description of how the field functions, not evidence that editing it raises sales. A backlist title sells because someone is still maintaining its record months after the launch conversation moved on.
The three decisions the spokes take apart
Each spoke in this pillar handles one recurring maintenance decision in detail, worth tracking on its own timeline.
- Metadata revisits: how often to reopen categories and keywords, and what signal justifies a change
- Series consistency: what to track across volumes so book two agrees with book one
- Post-publication corrections: which errors warrant a new upload and which do not
Why Metadata Drifts When the Book Stands Still
The book is fixed after publication, but the environment around it keeps changing. Subject classification schemes are revised on a published schedule: the Book Industry Study Group issues annual editions of the BISAC Subject Headings list, and codes are added, renamed, or deactivated between editions. A category assignment that was accurate in one year can become outdated the next, without anyone touching the book.
Retailer browse structures move independently of BISAC. Amazon’s own category guidance is maintained on its own schedule, which means a title can sit in a category that no longer matches where comparable books now live. Comp titles drift too. The books that defined a category at launch get replaced by newer entries, and a description written to stand out in month one can read as generic by month fourteen.
It’s worth being explicit about evidence quality here. Figures circulating in industry commentary, such as claims that complete metadata records sell substantially more, appear repeatedly in association and distributor writing, including IBPA’s “Metadata That Sells”, but are not accompanied by a published dataset or methodology. Treat them as plausible and unverified rather than tested.
There is a way to test drift yourself. Change one field at a time, record the date, and compare sales rank and page reads across matched periods before and after. A single title cannot produce a clean result, but a repeated log across several changes will show whether anything moves at all. Perivane’s audit of The Murder at the Vicarage shows this kind of evaluation in practice, judging a long-established title’s category placement against the books currently occupying those shelves rather than against its own reputation.
Keeping a Series Straight So Book Two Does Not Contradict Book One
Series continuity is both a craft problem and a metadata problem, and the metadata half is the one that silently breaks. Consistent details matter: series name spelled identically across volumes, volume numbers in the correct field rather than buried in the subtitle, each book’s description pointing at the others, and a reading order unambiguous to someone arriving at volume three first. These are small fields that carry a large cost when they’re wrong.
The content half is just as exposed. Character names, timelines, ages, and place names all have to hold across volumes in fiction. Non-fiction series carry the same risk through cross-references, chapter numbering, or citation references that must resolve correctly from one volume to the next.
Self-reported data from Draft2Digital indicates series titles account for a large share of indie unit and revenue volume in its network, though this comes from one distributor’s own data rather than an independent audit and should be read as directional. Series continuity fails quietly: readers who spot the contradiction usually leave a review rather than send an email.
Perivane’s audit of J.C. Ryle’s Expository Thoughts on Matthew demonstrates this kind of cross-reference checking on a multi-volume non-fiction work, where internal references are the main continuity surface. Continuity tracking is cheapest when built during the drafting of book two and most expensive when reconstructed after publication, once errors have already reached readers.
Catching the Errors That Shipped Before a Reviewer Does
Every published book contains errors that survived proofreading. Perivane’s audit of J.C. Ryle’s Expository Thoughts on Matthew, for instance, turned up broken cross-references that had shipped for over a century before anyone flagged them. The question after launch is which ones are worth a new upload and how you find them without rereading the whole book every quarter.
A useful sweep covers several categories at once: typographical errors readers report, broken or expired hyperlinks, outdated author bio and back matter, incorrect front matter copyright or edition statements, formatting faults that only appear on certain devices, and metadata that contradicts the book’s actual content. Reader reviews often function as a defect log, whether or not you treat them as one.
A fixed schedule works better than reacting only when someone complains: twice in year one, then annually after. Keep a running defect list between uploads so corrections ship in batches instead of one at a time.
Common mistakes worth avoiding: republishing for trivial single-character fixes, letting back matter links rot for years, and changing the manuscript without updating the version record, so you can no longer tell which file is current. Errors do not become less expensive with time. They become more public. Both Perivane sample reports illustrate what a full manual audit surfaces on titles that were already published, and in one case, already canonical.
Why the Returns Are Slow, and Why That Is the Reason It Gets Skipped
Be honest about the shape of the payoff. Launch activity produces a visible spike within days. Maintenance produces a slightly better baseline that is hard to attribute and impossible to screenshot.
The available evidence supports the modest version of the claim. Data from eBOUND Canada and Ontario Creates measured a significant lift in Amazon discovery from keyword metadata supplied through distribution, and found no sales effect over the quarter tested, across roughly 600 Canadian ebooks. It tested ONIX keyword fields delivered through distribution, not KDP’s seven keyword boxes, so the finding does not transfer directly to a self-published KDP title.
Survey data from the Alliance of Independent Authors shows higher median incomes among authors with longer publishing histories, though such surveys are self-selecting and cannot separate backlist depth from experience, genre, or output rate. Learning how to keep selling books after launch is mostly learning to keep doing work that gives you no feedback for months.
The attribution problem is structural: the eBOUND Canada study tracked roughly 600 titles for a single quarter and still could not isolate a sales effect, which shows how noisy the data stays even under controlled conditions. Accept that and keep a dated log anyway, because the log is the only evidence you will ever have.
Why the Long Push Matters
A book has one launch and many years. The decisions that determine whether those years produce anything are made quietly, by whoever is still paying attention. This stage separates a catalogue from a pile of finished files, and maintenance is the only part of publishing that gets cheaper the longer you keep doing it.
Conclusion
How to keep selling books after launch comes down to three recurring maintenance jobs: metadata revision, series consistency, and error correction, performed on a schedule rather than in response to panic. The evidence supports discovery gains more strongly than sales gains, and most of the confident statistics in circulation cannot be traced to a published study. Do the work because the alternative is a record that decays, not because a number promised you a return.
A reasonable next step: pick one title, open its retailer record, and compare every field against the book as it exists now. Most authors find something wrong within ten minutes. For more on where launch-window effort tends to go wrong before this maintenance phase even begins, see common self-publishing launch pitfalls, and for context on why paid promotion alone cannot substitute for this kind of upkeep, see why paid ads fail indie books. The two full sample reports at perivane.com show what a complete manual audit of a published title actually contains.