ARC review rates: what the data actually says

Ask five indie authors what share of ARC readers actually leave a review and you'll get five numbers, all of them confidently stated and none of them sourced. The folklore says something like "send a hundred, get a handful." The measured campaigns say something quite different.

This is the reference page we wanted and couldn't find: what the research says reviews are worth, what review rates real campaigns actually produce, and — the part that causes most of the confusion — which denominator anyone is using when they quote you a percentage. Every figure below links to where it came from, and we've labelled which are independent research, which are a platform's own marketing figure, and which come from a single author's launch.

First: what a review is actually worth

The economic case for running an ARC campaign at all rests on one well-known piece of research. Northwestern's Medill Spiegel Research Center analysed purchase and review data and found that the purchase likelihood for a product with five reviews is 270% greater than for the same product with none.Spiegel The important part is the shape of the curve, not the headline: nearly all of that lift happens between zero and five reviews. After that the returns shrink quickly.

Three more findings from the same work are worth knowing:

• Price changes the stakes. For lower-priced, low-consideration purchases, reviews lifted conversion by around 190%. For higher-priced or higher-consideration purchases, closer to 380%.Spiegel A novel from an author a reader has never heard of behaves like the second category — you can't evaluate the writing until after you've bought it.
• A perfect rating hurts you. Purchase likelihood peaks in the 4.0–4.7 range and then declines as the average approaches 5.0.Spiegel A flawless 5.0 reads as manipulated or as too small a sample to trust. A few honest 3- and 4-star reviews make the whole set more credible.
• Verified buyers count for more. A verified-buyer badge improves the odds of purchase by about 15%.Spiegel Verified reviewers also skew less extreme — anonymous reviewers are considerably more likely to leave 1–2 stars.

For calibration on that last point: BookSirens reports that the average rating its readers give, across every book on the platform, is 4.1 out of 5, as of January 2026.BookSirens pricing That sits inside the credible band rather than above it, which is what you want.

Why reviews don't arrive on their own

Written Word Media's 2026 reader survey (3,589 respondents) found that 22% of readers leave reviews often and 30% do so sometimes, with 17% saying they never leave one at all.Written Word Worth reading that with a caveat the survey itself implies: respondents came from Written Word Media's own promotions list, so these are engaged, active readers. The general book-buying public almost certainly reviews less, not more.

So roughly half of even a highly engaged readership reviews rarely or never — after finishing a book they enjoyed. If you wait for reviews to appear organically, you wait through exactly the window where they'd have done the most good.

The number everyone quotes wrong

Here's the source of most of the disagreement. "Review rate" can mean at least three different things, and they produce very different percentages from the same campaign:

• Reviews ÷ people who requested a copy. The most pessimistic, and the one that includes everyone who raised a hand and then evaporated.
• Reviews ÷ copies actually downloaded. The most common industry framing, and the one most platform figures use.
• Reviews ÷ people who finished the book. Flattering, unmeasurable, and best ignored.

Susan Illene published the full funnel for a six-week romantasy campaign, which lets us calculate all of them from the same launch.Illene

Source: Illene, September 2025. One campaign — treat as a worked example, not an industry average.

Same campaign, three defensible headlines: 59.8% (reviews per download, at six months), 50.6% (reviews per download, at twelve days), or 42.3% (reviews per original request). None of them is "a handful out of a hundred." When someone quotes you a review rate, the first question is always: over what?

Rates by platform

Because that campaign ran across five platforms plus a direct list simultaneously, it also gives us something rare — a same-book, same-author, same-window comparison, which strips out the variables that usually make cross-platform numbers meaningless.Illene

Source: Illene, September 2025. Single campaign, romantasy. Platform performance varies by genre and by how well the book is matched to each audience.

Two things stand out. The first is that the floor is around 38%, not 4%. Even the weakest channel in this campaign returned reviews on well over a third of copies. The second is the gap inside a single platform: on BookSirens, readers invited from the author's own list reviewed at 93% against 72% for strangers who found the book by browsing. Roughly two-thirds of her ARCs came from that direct list.

For a second data point on the ceiling: BookSirens states a site-wide review rate of 75% across its reader community, while being explicit that it can't guarantee that for any individual book — real books range from above 90% to below 50%.BookSirens That's a vendor's own figure and should be read as one, but it's corroborated rather than contradicted by the independent campaign above.

Your launch-week number is not your final number

In that campaign, reviews grew from 263 at twelve days to 311 at six months — another 48 reviews, an 18% increase, arriving long after anyone would have declared the campaign finished.Illene Some platforms are built for launch-week velocity and some for the long tail, and if you close the books on day twelve you'll misjudge which is which.

This matters for how you evaluate spend. A platform that looks mediocre on launch day may be your best performer by the time the sequel comes out — and for a series author, reviews still accumulating six months later are landing exactly when book two needs them.

So where does "I sent 100 and got 4" come from?

It's real, and we hear it constantly. But notice what all the measured campaigns above have in common: they ran on platforms that set a deadline and chase the reader with reminders. Nothing in those numbers required a better book — the accountability is doing the work.

What the low numbers seem to describe is the self-run campaign: a spreadsheet, one email, no deadline, no reminder, no memory of who ghosted last time. We should be straight that there is no published dataset measuring self-run spreadsheet campaigns — nobody's collecting that data, so the 10–20% figure that circulates for them is craft knowledge rather than evidence. What the evidence does establish is the other end of the range: managed campaigns with deadlines and follow-up land between roughly 38% and 75%.

Read that gap as the opportunity it is. The difference between a 15% campaign and a 60% campaign isn't the book. It's whether anything in the system remembers who took a copy and whether anyone follows up before the deadline.

What to do with all this

• Set your target on downloads, not invitations, and say so when you report it. 50–60% of downloads is a realistic goal for a well-run campaign.
• Invite your own list first. The 93%-vs-72% split inside one platform is the single biggest lever in the data. See how to build an ARC team from zero if you don't have a list yet.
• Aim for five reviews on day one, then keep going. The Spiegel curve says those five carry most of the conversion benefit.
• Don't chase a 5.0. The credible band is 4.0–4.7, and honest mid-range reviews help you get there.
• Keep counting for six months. Judge platforms on their long-tail numbers, not their launch-week ones.
• Keep a record that survives the launch. Everything above depends on knowing who downloaded and who delivered — which is what a permanent reviewer CRM and a reliability score exist to do.

Next: how the major ARC platforms compare on cost and completion, when to start sending by genre, and what an ARC campaign is actually worth once you account for read-through.