Here's an uncomfortable truth most authors never hear: Amazon is always watching how readers respond to your book, and every ad you run is teaching the algorithm something. The question is whether you're teaching it the right lesson. When your targeting is broad, your metadata is muddled, and your traffic doesn't match your book, you're not just wasting budget — you're actively training Amazon to conclude that readers aren't interested in what you've written.
This matters because Amazon's advertising and organic systems reward relevance above almost everything else. A book that converts the clicks it receives gets shown more, ranked higher, and served cheaper impressions. A book that pulls in curious-but-uninterested browsers who never buy sends the opposite signal. Over weeks, that signal compounds into a quiet form of suppression — your ads cost more, reach fewer of the right readers, and your organic visibility erodes.
This article explains how irrelevant impressions and unconverted clicks weaken the signals Amazon uses to decide where — and whether — to show your book, and exactly how to laser focus your targeting so the algorithm learns to work in your favour instead of against you.
When you run a Sponsored Products campaign, Amazon isn't just charging you per click. It's building a behavioural model of your book. Every impression is a test: it shows your book to a reader and watches what happens. Did they click? Did they buy? Did they bounce back to search within seconds? Each of these micro-events feeds a relevance score that determines how competitive your book is in future auctions and how often it surfaces organically.
The mechanism most authors misunderstand is click-through and conversion rate as a quality signal, not just a performance metric. Amazon interprets a low conversion rate as evidence that your book is a poor match for the query or audience you're bidding on. A book that converts 12% of clicks on a tightly matched keyword tells Amazon "this is exactly what these readers want." A book converting 1% across a scattergun of loosely related keywords tells Amazon "this book disappoints the readers we send it to."
The consequence is not neutral. When your book earns a weak relevance signal, Amazon raises the effective cost to show it. Your bids buy fewer and worse placements. Meanwhile, the organic algorithm — which draws on the same behavioural data — deprioritises your title in search results and "Customers also bought" carousels. You end up paying more to reach fewer of the right people, while your free visibility quietly shrinks.
This is why broad targeting is so dangerous early on. Authors are told to "gather data" with wide auto-campaigns, and that's reasonable for a week or two. But broad targeting left running for months doesn't gather data — it manufactures a track record of irrelevance. You're not exploring; you're teaching Amazon that your book underperforms.
Consider a mid-list thriller author who launches with a single auto campaign and a broad-match keyword list of forty terms. Some of those terms — "psychological thriller," "detective novel" — are relevant. Others crept in from Amazon's auto-matching: "true crime documentary," "mystery box," "crime scene cleaning." The book gets thousands of impressions across all of them.
The relevant terms convert at a healthy rate. The irrelevant ones convert at essentially zero. But Amazon doesn't grade each keyword in isolation when it forms an overall impression of your book's market appeal. A flood of non-converting impressions drags down your book's aggregate performance signal. The strong keywords are effectively penalised by the company they keep.
The same failure happens with metadata. If your title, subtitle, categories, and backend keywords describe three slightly different books — part memoir, part self-help, part business — Amazon can't confidently place you. It shows your book to a blurry, poorly matched audience, most of whom bounce. Confused metadata and confused targeting reinforce each other: the metadata invites the wrong readers, and the wrong readers confirm to Amazon that your book doesn't deliver.
Unconverted clicks are the most expensive part. You pay for each one, and each one deposits a small negative mark on your relevance profile. Ten unconverted clicks on a term you should never have bid on costs you money twice: once at the checkout, and again in the form of a degraded signal that makes your good keywords more expensive to win. This compounding is why some authors watch their ACOS climb month after month despite "optimising" — they're optimising bids while the underlying relevance signal keeps decaying.
- Every keyword describes a reader who would genuinely enjoy your book
- Your metadata, categories, and ad targeting all tell the same story
- You harvest converting search terms and promote them to exact-match campaigns
- Non-converting keywords are negated quickly and deliberately
- You measure conversion rate per term, not just overall spend
- Your comp-author targeting matches readers who buy books like yours
- A single broad auto campaign runs untouched for months
- Backend keywords stuff in loosely related or high-volume terms
- You judge success by impressions and clicks instead of sales
- Irrelevant search terms are left to accumulate spend and bad signal
- Your title says one genre and your categories say another
- You bid on bestseller names whose readers won't cross over to you
The good news is that Amazon shows you the same evidence it uses against you. Your search term report is the single most valuable document in your advertising account, and most authors barely open it. It lists every actual search a reader typed before clicking your ad — and crucially, whether that click led to a sale.
Pull the report for a 30-day window and sort by clicks with zero orders. This is your list of accumulated liabilities: terms you're paying for that convert nothing and quietly poison your relevance signal. Some will be obviously wrong-genre. Others will be surprisingly close to your book but still not converting — those are the subtle killers, because they look reasonable enough to leave running.
Next, look at the inverse: terms with a strong click-to-order ratio. These are the readers who want your book. This is where your budget belongs, and these are the terms you promote into their own tightly controlled exact-match campaigns so you can bid on them deliberately rather than hoping auto-matching finds them again.
The discipline here is subtraction, not addition. Most struggling campaigns don't need more keywords — they need fewer. Every term you negate is a small correction to the story Amazon is telling about your book. Do it consistently for a month and you'll often see effective CPCs fall on your best keywords, because you've stopped diluting the signal that makes them competitive.
One practical rule: don't negate on a single click. Wait until a term has had roughly ten to fifteen clicks with no conversion before cutting it — enough of a sample to be confident it's genuinely a poor match rather than unlucky. Judging too early throws away terms that might have converted; judging too late lets bad signal accumulate.
Start by defining the reader you're actually trying to reach — not the widest possible audience, but the specific person who finishes a book like yours and immediately wants another. Every targeting decision should trace back to that reader. If a keyword or comp-author wouldn't attract them, it doesn't belong in your campaign, no matter how much search volume it carries.
Build your campaigns in tiers. Keep a small, capped auto campaign or broad campaign running purely as a discovery engine — its only job is to surface new converting search terms. Set a modest budget and treat everything it finds as a candidate, not a permanent placement. When a term converts there, graduate it into an exact-match campaign where you control the bid precisely. This separates exploration from exploitation cleanly.
Then run the negation loop weekly. Open the search term report, negate the non-converters that have crossed your click threshold, and reallocate that freed budget toward your proven exact-match terms. This single habit — running consistently, not once — is what separates authors whose ACOS trends down over time from those whose ACOS drifts up.
Finally, audit your metadata against your targeting. Your title, subtitle, categories, and seven backend keyword fields should all describe the same book to the same reader your ads target. If your ads chase thriller readers but your categories place you in general fiction, you've created a mismatch Amazon has to reconcile — usually by showing you to a blurrier audience. Alignment across metadata and targeting is what lets Amazon place you confidently, and confidence is exactly what earns you cheaper, better-matched impressions.
Do these four things — define the reader, tier your campaigns, run the negation loop, align your metadata — and you stop training Amazon to ignore you. You start teaching it that your book is a reliable match for a specific reader, and Amazon rewards reliability with reach.
Every unconverted click costs you twice — once at checkout, and again as a signal that makes your best keywords more expensive to win.
— ScribandoWe start with a search term audit, not a bid change. Before touching budgets, we pull every term a book has paid for and separate the converters from the dead weight. That report tells us exactly what Amazon currently believes about the book — and it's usually the opposite of what the author intended.
From there we rebuild campaign structure in tiers: a capped discovery layer whose only job is to find new converting terms, and controlled exact-match campaigns where the proven winners live. We negate accumulated non-converters deliberately, using a click threshold rather than guesswork, so we're correcting the signal without cutting terms prematurely.
Then we align the listing itself — title, subtitle, categories, and backend keywords — so the metadata and the targeting tell Amazon a single, coherent story about who this book is for. Finally, we run the negation-and-reallocation loop on a regular cadence, because focus isn't a one-time fix. It's a habit that keeps the relevance signal sharp as the market shifts.
Focused targeting isn't about spending less — it's about teaching Amazon the right lesson about your book. That's what we do at Scribando: The Intelligence Layer of Book Marketing.