Website audit report overwhelming? Read only page one
A website audit report does not go unread because it is long. It goes unread because nothing in it obviously wins. The research on that, and the one page worth your ten minutes.
"I'll get a 40-page report I never read."
Fair. You have probably been sent one already, and it is still sitting in a folder. But a website audit report is not overwhelming because of its length. It stalls you because nothing in it obviously wins, and those are different problems with different fixes.
So the question to ask before you buy any audit is not how many pages it runs to. It is what sits on page one, and what that one thing is worth.
Why does a website audit report feel overwhelming?
Not because of the count. Scheibehenne, Greifeneder and Todd pooled 63 conditions from 50 experiments covering 5,036 people, and found the average effect of adding more options was close to nothing. They call it "virtually zero". A longer list, on its own, does not stop people deciding. Something more specific does.
Their paper names it. Overload shows up when you have no prior preference between the items, and when nothing in the set obviously beats the rest. Both have to be true. The authors are blunt that they found necessary conditions and no sufficient ones.
Read that against the report in your folder. Two hundred findings, no preference between them because you cannot price them, and no single item presented as beating the other 199. That is the condition, exactly described. The page count is incidental.
Why do two good studies disagree about this?
Because the famous one is an outlier with conditions attached. Iyengar and Lepper ran a jam tasting table at Draeger's Supermarket in Menlo Park, and their result is the one everyone half remembers. The meta-analysis that includes it reports effects running in both directions across the other 49 experiments.
The half nobody repeats is the more useful one. Here is what the original paper actually recorded:
- 6 jars versus 24 jars on the tasting table, on two consecutive Saturdays
- 60% of 242 shoppers stopped at the 24 jar display, against 40% of 260 at the 6 jar one
- Nearly 30% of those who stopped at the small display bought a jar, against 3% at the large one
- 5,036 people across the 50 experiments pooled later, average effect of more choice close to zero
The big display won the attention and lost the sale. Which is also how a 200 finding report gets bought.
More people stopped at the twenty-four jar table. Almost none of them bought anything.
How many findings can you actually hold?
About four. Nelson Cowan's review of the short-term memory evidence puts the central capacity limit at "about four chunks", with three to five as the population average. Miller's famous seven plus or minus two was, in Cowan's reading, probably a rhetorical device rather than a measured ceiling.
Every agency page telling you to pick three fixes arrived at roughly the right number by instinct. None of them says where it comes from, which is why the advice sounds like a preference rather than a constraint. It is a constraint. Four is the ceiling you are working against, and a report that hands you 200 items has ignored it.
What does a ranked report put on page one?
Three things, in Revslip's case. A three sentence verdict on what is wrong, scores out of 100 for conversion and performance, and the three lowest effort highest impact findings tagged as quick wins. Every other finding carries a severity, a revenue impact band and the hypothesis behind it.
That last part is what makes the order arguable rather than decorative. Across the 134 businesses Revslip has audited, roughly 200 signals get checked across messaging, conversion, trust, technical, performance and content, and the checks route by business type, so an ecommerce site and a consultative B2B site get different arithmetic. That sample is not neutral, and the reason is obvious once stated: sites that get audited are sites whose owners already suspected something was wrong. See what lands at the top for your site.
Ranked is not the same as right. The order comes from an estimate of what each finding costs you, so if the estimate misreads your business the order is wrong with it. Revslip prints the hypothesis next to every finding for exactly that reason: read it, and bin any item that assumes a business you do not run. We have written separately about how much weight those revenue estimates carry.
What is the rest of the report for?
Reference, not homework. The tail is there so you can check whether something you already suspected is in it, and so the ranking can be challenged. Nobody is meant to work through 200 items. Treat anything below the top few as an index you search, not a list you complete.
This is also where an unranked report does its real damage. With no order, a missing trust badge arrives at the same weight as a checkout that fails on mobile. One of those is worth fixing this afternoon and the other mostly is not, and we have gone through which badges actually move anything elsewhere. A flat list cannot tell you which is which, so you read both, believe neither, and close the file.
Ten minutes is enough to get the top of any report, whoever produced it:
- Read the summary and write the top item down in your own words. If you cannot, nothing has been ranked.
- Find the number attached to it. A finding with no cost on it cannot be compared with another finding.
- Read the assumption underneath that number. Reject it if it describes a business you do not run.
- Pick one item. Ship it this week.
- Put the rest away and search it when something breaks.
Run the free audit and read only page one. How the audit works if you want the mechanics first, and the feature list if you want the full set of checks.
When is a long report the right thing to buy?
When somebody is paid to work through it. If you have a developer and a marketer with hours booked against the site, the full 200 findings beat a ranked three, because the work can run in parallel. A ranked report is built for the founder doing this between other jobs.
If that is not you, it is the wrong shape. Two more cases where you should not buy this either. If the decision in front of you is whether to rebuild the site, a list of findings is not the thing that answers it, and the order you do those two in matters more than the report. And if nothing is going to change on the site in the next month, no report length helps, because the constraint is your calendar rather than your information.
And there is a sharper limit on our side, built into the very thing this post is recommending. Tagging the top three by lowest effort and highest impact means a fix worth more, but taking two weeks, can sit underneath three cheap ones. That ranking optimises for what you will actually ship, which is not the same as what is worth the most. So before you close the file, read the impact band on the items further down, and if one of them is marked high, put it in next month rather than never.
The 40 page report you never read was not too long. It just never told you which page to start on.