Redefine Web
SEO

Benchmark SEO against your sector, not a published table

Benchmark SEO tables rarely state their sample. What four publishers actually measured, why you cannot combine their numbers, and what to compare instead.

· 14 min read
Benchmark seo illustration
Key takeaways
A benchmark table is a sample of one vendor's customers, not a census of your sector.
Of four publishers ranking for this search, only Similarweb states its sample size, and it gates the figures.
First Page Sage and Databox both publish readable tables dated 2023, neither stating how many sites contributed.
Never average two publishers or multiply one's traffic by another's conversion rate. The result describes nobody.
Five competitors you named yourself beat any industry row, because you know why each one is in the set.

To benchmark SEO is to answer one question. Is this number normal for a business like mine. Suppose your own conversion rate is 2 percent. That figure means nothing on its own, and it means two opposite things depending on whether you sell legal services or software. So people go looking for an industry table, find one, and treat the row for their sector as a target.

We sell audits, so treat this as interested testimony and check it. Every figure below is attributed to the publisher that produced it, in the same sentence, and was read off that publisher’s own page on 16 September 2026. Where a publisher does not say how big its sample was, that is stated too, because it turns out most of them do not.

What benchmark SEO means, and what this article leaves alone

One boundary first, because three neighboring questions get tangled together and answering the wrong one wastes an afternoon.

What a given number means, and how a position differs from a tool’s score, is a separate subject, handled in our guides to SEO ranking and SEO analytics. Which numbers belong in a monthly report is another one, and our guide to what goes in an SEO report covers that side of it. This article is only about the third question, what counts as normal for your sector and who you should be comparing yourself against.

SE Ranking draws the underlying distinction cleanly on its own benchmarking guide. “KPIs measure what you are trying to achieve, while benchmarks tell you how far you’ve come and what’s needed to reach your objectives.” A KPI is a target you set. A benchmark is a comparison you make. The rest of this article is about where that comparison should come from.

The short answer, argued for below, is that it should come from a set of competitors you name yourself rather than from a published table, and that the published tables are most useful for showing you how wide the spread between sectors is.

Where published SEO benchmarks actually come from

Before quoting any table it is worth knowing what kind of thing you are quoting, because these numbers are not collected the way a government statistic is.

An SEO benchmark table is a sample. Somebody took a pool of websites they had access to, computed an average or a median for each sector, and published it. The pool is almost always the publisher’s own customers or its own crawl data, which means the sample is whoever happened to buy that product. That is not a criticism, because there is no other way to get the data, but it does change what the number can tell you.

So the first question to ask of any benchmark is whose sites are in it and how many. The second is when it was measured. The third is exactly how the metric was defined, because engagement rate and session duration are computed differently by different analytics setups. Of the four publishers examined below, only one answers the first question, and it is the one that does not publish the figures.

There is a quieter distortion in all of them worth naming. A sample drawn from a vendor’s current customers has already lost everybody who tried the product and left, and everybody whose results were bad enough that they stopped paying. What remains skews toward businesses for whom the thing was working. Nobody is hiding this and there is no way around it, but it means a published sector figure is closer to a picture of the reasonably successful than a picture of the typical.

What First Page Sage publishes and what stands behind it

This is the most detailed public table in the set, and it is the one most often quoted without its caveats.

First Page Sage publishes SEO benchmarks across 19 industries on five metrics, average session duration, engagement rate, annual percentage change in organic traffic, number of autofill transactional keywords on page one, and visitor-to-lead conversion rate. The spread across sectors is the useful part. First Page Sage puts visitor-to-lead conversion at 7.4 percent for legal services and 1.1 percent for B2B SaaS, and engagement rate at 83 percent for ecommerce against 52 percent for real estate.

Now the caveats, both of which are printed on the page. First, the firm describes the source as its own proprietary pool of data covering the industries it serves, and it is an SEO firm, so the sample is its clients and prospects rather than the sector at large. Second, the report is dated 26 October 2023 and the page carries the same date as its last update. It does not state how many websites are in the sample.

None of that makes the numbers useless. It makes them one firm’s client base, measured almost three years ago, which is a very different object from an industry standard.

What Databox publishes and what stands behind it

The second readable table takes a different approach to the same problem, and its choice of statistic matters more than it looks.

Databox publishes medians rather than averages across 15 industries, on metrics including impressions, position, clicks, click-through rate, sessions, session duration, engagement rate and conversions. A median resists being dragged by one enormous site in the sample, which for this kind of data is the better choice and worth noticing.

Databox puts the median click-through rate across all industries at 1.56 percent, with ecommerce and marketplaces at 0.88 percent and education at 2.73 percent. It puts median impressions across all industries at 107.81 thousand, with health and wellness at 202.31 thousand and ecommerce and marketplaces at 57.62 thousand.

Two caveats again. The figures are labeled for August 2023 on the page, so they are of similar vintage to the set above. And the table has gaps, with some sector cells carrying no value at all, which is honest of them and a reminder that a benchmark table is a sample with holes in it rather than a census. Databox does not state how many companies contributed data.

What Similarweb publishes, and the one thing it does better

The third publisher inverts the problem, and the comparison between it and the first two is the most instructive thing in this whole exercise.

Similarweb’s SEO Benchmarks Report page states its sample outright. It says the company analyzed key SEO data across 10,000 websites in 10 industries on 14 key metrics. That is exactly the disclosure the other two are missing, and it is what lets a reader judge whether a number is worth anything.

The figures themselves sit behind a download form. So the publisher that tells you the most about its method is the one whose numbers you cannot read without handing over your details, and the publishers whose numbers you can read do not tell you how many sites produced them. That trade is worth knowing about before you go looking.

BrightEdge also ranks for this search with a benchmarks article. Its page builds its content in the browser rather than serving it in the page, so no figure from it could be verified from the page source and none is quoted here.

The three problems every published table shares

Read across those publishers and the same three weaknesses turn up every time, regardless of who produced the data.

The sample is whoever bought the product. A benchmark built from one vendor’s customers describes companies that were already spending money on the problem, which is not the same population as your sector. The age is usually older than the page looks, and two of the tables above carry 2023 dates while ranking today. And the sector labels are coarse, so a specialist manufacturer and a mass-market retailer land in categories that were never designed around how people search for what they sell.

The third one does the most damage in practice. Your sector row aggregates businesses whose customers behave nothing like yours, and the more specialized you are, the further the row sits from anything you could act on.

Definitions are the trap underneath all three. Engagement rate is not one thing, it depends on what the analytics install counts as an engaged session and on thresholds a site owner can change. Conversion rate depends entirely on what somebody decided to count as a conversion, and a form view, a form submission and a qualified lead are three very different denominators wearing one name. Two publishers can report the same metric for the same sector and be measuring genuinely different events.

Which is why a benchmark you cannot trace back to a definition is not really a number yet. It is a number shaped object, and the further it travels from its source, the more confident and less checkable it gets.

Use the tables for the spread rather than the target. The useful lesson from First Page Sage’s numbers is not that legal services should hit 7.4 percent, it is that the same First Page Sage table puts B2B SaaS at 1.1 percent on that same metric, and two sectors sitting that far apart tells you how meaningless a single cross-industry figure is.

Why you cannot combine two publishers into one number

This is the most common mistake made with these tables, and it is worth stating plainly because it looks like diligence.

Faced with two sources, the instinct is to average them, or to take one publisher’s traffic figure and multiply it by another’s conversion rate to estimate leads. Both operations destroy the thing that made the inputs worth having. Averaging two samples that were drawn differently, measured in different periods and defined differently produces a number that describes no population at all, and it now carries the authority of two sources while being traceable to neither.

The multiplication is worse, because the traffic sample and the conversion sample are different sets of websites. You would be applying one group’s behavior to another group’s volume and presenting the result as a forecast.

So keep every figure attached to its publisher, its date and its definition, and never let two of them merge. That is why the figures in this article are always introduced by the name of the organization that produced them, and why no combined figure appears anywhere in it.

Benchmark against named competitors instead

Here is the version of this exercise that actually changes decisions, and it is the one the published tables cannot do for you.

Benchmark SEO. A spreadsheet window holding a benchmark export, six columns of cells with a highlighted read column on the right, one row per competitor being measured.

Pick five businesses that genuinely compete with you for the same buyers, measure the same small set of things about each of them, and compare yourself to that. The sample is tiny, which sounds like a weakness and is the entire strength, because you chose every member of it and you know why each one is in there. A sector row of 15 industries cannot tell you that the firm two towns over is beating you on the four terms that matter. Five named competitors can.

This is also the search most people are really running when they look for benchmarks. The related searches around this topic are dominated by competitor comparison rather than industry tables, which suggests the market already knows the named set is the useful one.

Our walkthrough of SEO performance step by step covers the measurement side of this, and our guide to analyzing a website covers running the same assessment on your own site first.

How to choose the five, and who to leave out

Choosing the comparison set is the part that decides whether any of this is worth doing, and most people choose wrong in the same two ways.

The first mistake is picking the biggest names in the industry. A national brand with a twenty year old domain is not a benchmark, it is a different weight class, and every gap you find will be explained by history rather than by anything you could change this quarter. The second is picking whoever ranks first for your single most important term, because that one page may be an outlier propped up by a link profile that has nothing to do with the rest of their site.

Pick businesses of roughly your size that sell roughly what you sell to roughly who you sell it to, and that appear repeatedly across the searches you care about rather than once at the top of one. Repetition across a term set is the signal. A single first place is noise.

Include one business clearly ahead of you as a direction marker, but do not set targets from it. Set targets from the middle of your chosen set.

Write down why each of the five is in the set, in one line each, at the moment you choose them. Six months later that note is what stops somebody quietly swapping in whoever ranks first this week, which turns a stable comparison into a moving one and destroys the trend you were building.

Measure everybody the same way or do not bother

A comparison is only a comparison if the same instrument was pointed at every subject, and this is where most competitor benchmarking quietly falls apart.

You have first party analytics for your own site and nothing but external estimates for everybody else’s. So comparing your true session count against a third party tool’s estimate of theirs is not a comparison, it is two different measurements wearing the same label. Either use the external tool’s estimate for all six sites including your own, accepting that the absolute numbers will be wrong, or restrict the comparison to things that are equally visible from the outside.

Externally visible things work best anyway. Which terms each site ranks for and where, how many pages each has on the topics that matter, how fast the pages load, whether the pages answer the question the search asked. Those can be measured identically for anyone, including you, which is what makes the resulting gap real.

Our roundup of free SEO tools and our comparison of website ranking software both cover what these external tools can and cannot see about a site you do not own.

The sector spread, and why one good number does not exist

It is worth sitting with how far apart the sectors are, because it settles a question people keep asking in the wrong form.

Free SEO competitor analysis. A spreadsheet window with a chart card over the table it was built from, showing a spread of values rather than the single number a published benchmark table prints.
Metric and publisherLow sectorHigh sector
Visitor-to-lead conversion, First Page Sage1.1 percent, B2B SaaS7.4 percent, legal services
Engagement rate, First Page Sage52 percent, real estate83 percent, ecommerce
Median click-through rate, Databox0.88 percent, ecommerce and marketplaces2.73 percent, education
Median impressions, Databox57.62 thousand, ecommerce and marketplaces202.31 thousand, health and wellness

Read down the rows and notice that ecommerce is at the top of one and the bottom of two others, from two different publishers measuring different things. A sector is not simply good or bad at search. It is shaped differently, with different buying cycles and different volumes behind each conversion.

Which is why the question of what a good number looks like has no answer in the abstract, and why the two halves of the table above must stay in their own rows rather than being blended into a single figure.

When your sector is not in anybody’s table

Most businesses are not in these tables, and the honest instruction for that situation is not to find a closer row.

Between them the publishers above cover 19, 15 and 10 industry categories. That is a coarse grid over the whole economy, so a specialist business will usually find its nearest label describes a neighbor rather than itself. The temptation is to pick the closest row and treat it as an approximation. It is better to treat it as unrelated, because a wrong benchmark is worse than none. A wrong one produces confident targets.

The damage from a wrong row is specific rather than vague. It sets an expectation somebody will be judged against, and when the site misses a target that was never relevant, the conclusion drawn is usually that the work is failing rather than that the target was borrowed from a different kind of business. Budgets get cut on that reasoning. A missing benchmark leaves people asking what changed, which is the better question anyway.

What replaces it is your own history plus the named competitor set. Your own site last quarter is the single most relevant comparison available to you, it needs no sample disclosure, and it is measured with exactly the same instrument as your site this quarter.

Start there, and add the competitor set when you need to know whether a flat quarter was you or the whole market. Our guide to SEO visibility covers establishing that baseline and watching it move.

What we would check first

If you are about to benchmark SEO for a site, this is the order that avoids the most wasted work.

Write down your own numbers for the last four quarters first, because that comparison is free, exactly relevant and measured consistently. Then name five real competitors and pick three or four externally visible things to measure about all six sites including yours. Only then go and look at a published table, and read it for the spread between sectors rather than for your own row.

When you do quote a published figure to anybody, carry the publisher’s name, the date and the sample with it every time. If the sample is not stated, say that too. A benchmark whose provenance travels with it can be checked by the person you are showing it to, and one that arrives as a bare number cannot.

If what you want is somebody to establish the baseline and name the competitor set for you, that is a fixed piece of work rather than an ongoing one. Our SEO audit starts there, and our website audit covers the same ground plus the speed and security side that never shows up in a rankings table.

Frequently asked questions

A benchmark is a comparison point that tells you whether a number is normal, as distinct from a target you are aiming at. SE Ranking puts the distinction well on its benchmarking guide, saying KPIs measure what you are trying to achieve while benchmarks tell you how far you have come. In practice a benchmark is either your own past performance or somebody else's current performance.

It depends heavily on sector, and any single figure should carry its source. Databox published a median click-through rate across all industries of 1.56 percent for August 2023, with ecommerce and marketplaces at 0.88 percent and education at 2.73 percent. Treat that as one publisher's sample rather than a standard, and compare against your own previous quarters before comparing against any table.

They answer different questions, so compare what each discloses. Similarweb states a sample of 10,000 websites across 10 industries but puts the figures behind a download. First Page Sage publishes 19 industries openly from its own client pool. Databox publishes medians across 15 industries. Prefer whichever one tells you its sample size and measurement period for the metric you care about.

The phrase comes from general business benchmarking rather than search work, but the sequence that applies here is short. Pick the metric that reflects revenue. Pick the comparison set, either your own history or named competitors. Measure everyone with the same instrument. Identify the gap. Set a target from the middle of the set rather than from the leader.

It is benchmarking against businesses you name rather than against an industry table. You choose a handful of companies competing for the same buyers, then measure the same externally visible things about each, such as which terms they rank for, how many pages they have on a topic and how fast those pages load. The value comes from having chosen every member of the set deliberately.

Start by picking five businesses of roughly your size selling roughly what you sell, favoring ones that appear repeatedly across the searches you care about rather than once at the top of one. Then pick three or four things measurable from outside any of the sites, and measure all six including your own with the same tool. Read the gap, not the absolute numbers.

Google Search Console shows the queries your own pages already appear for, which is the honest starting point and costs nothing. For rivals you need an external estimate, and free tiers of the major platforms will show a limited sample of terms per domain. Treat every external keyword list as an estimate, and use it to spot themes you are missing rather than as an exact inventory.

No single tool wins, because the useful part is choosing the comparison set and the metrics, which no tool does for you. What matters is that whichever platform you pick is run against every site in your set including your own, so that a consistent instrument produces the gap. A cheaper tool used consistently beats an expensive one applied unevenly.

Measure against your own previous periods first, because that comparison is exactly relevant and uses the same instrument throughout. Add a named competitor set when you need to know whether a flat quarter was your site or the whole market. Use published industry tables last, and read them for the spread between sectors rather than for a target in your own row.
Found this useful? Share it.
Keep reading
FREE · WRITTEN IN 24 HOURS · NO PITCH

Get your free website audit.

A written report in your inbox within 24 hours, with three fixes you can ship the same week, whether or not you hire us.

WRITTEN IN 24 HOURS · 10,000+ SITES RUN · 300+ CLIENTS SINCE 2021