Benchmark report · 2026 edition

Website personalization, measured on live traffic

What personalization actually does when it runs on real sites — across 32 live Personyze accounts, 1,046 campaigns, 15,944,591 delivered actions and 65,509,365 sessions over a 90-day window in 2026.

Key findings

8.26% of the visitors shown a personalization campaign click it, at the median account; the top quartile reaches 17.13% (n=32). Counted per impression instead, the same accounts read 6.07%.
The campaign type matters more than the industry: 16.08% of visitors click a recommendation widget they are shown, against 4.18% for a banner or popup — roughly four times the rate, on the same sites.
11.59% of recommendation clicks lead to an order session (n=10), and recommendations carry 3.50% of product-goal value at the median account (n=13).
The median session is shown 1.45 actions, which is the whole difference between the two denominators — and the reason a rate quoted without one is not a number.
The spread is wide on every metric. Any single number quoted without its range — including ours — is hiding something.

Headline numbers

Every figure is the median account, never a pooled total. Pooling would let the largest account in the set — which alone delivered 3.7 million actions — describe the whole platform.

16.08%

of visitors shown a recommendation widget click it

11.06% per impression · n = 16

8.26%

of visitors shown any personalization campaign click it

6.07% per impression · n = 32

11.59%

of recommendation clicks become an order session

20.65% top quartile · n = 10

3.50%

of product-goal value is attributed to recommendations

6.71% top quartile · n = 13

The large figure is the median account, counted per visitor shown — the same denominator an email click rate uses. The line beneath gives the per-impression figure or the top quartile. Both are measured; neither is an average of averages.

Click rates, per visitor shown

Of the visitors who were actually shown a campaign, how many clicked it. The mark is the median account, the bar is the middle half of accounts, and the hairline is the full range. The spread is the finding: these are not tight distributions.

0%6%12%18%24%30%

Product recommendation campaignsvisitors who saw one and clicked

16.08%n=16

Any personalization campaignvisitors who saw one and clicked

8.26%n=32

Banner, popup and inline campaignsvisitors who saw one and clicked

4.18%n=24

Median account Middle half (25th–75th percentile) Full range

Hairlines that reach the right edge continue past 30% — the highest account on this chart is 78.14%. Exact ranges are in the table below and on hover.

Two denominators, and why they differ

Personyze reports click rates per impression, because that is what the delivery counter measures. Email platforms report per recipient. The median session here is shown 1.45 actions, so the same performance reads about a third lower per impression than per visitor. Both are below; neither is wrong, and quoting either without saying which is.

Click rate at the median account, both ways, over a 90-day window in 2026.
Campaign	Per visitor shown	Per impression	Top quartile	Middle half	n
Recommendation campaigns	16.08%	11.06%	27.50%	5.72% – 27.50%	16
Any personalization campaign	8.26%	6.07%	17.13%	3.00% – 17.13%	32
Banner, popup and inline campaigns	4.18%	2.63%	8.89%	1.34% – 8.89%	24
Recommendation widget, item level	9.82%	8.03%	14.37%	2.97% – 14.37%	14

Recommendation economics

What happens after the click, on the accounts that run a product catalogue. These two are session-based already, so they read the same either way.

0%5%10%15%20%25%

Rec click → order sessionof rec clicks that became an order session

11.59%n=10

Rec share of order valueof product-goal value attributed to recs

3.50%n=13

Median account Middle half (25th–75th percentile) Full range

Recommendations versus banners and popups

The same split again, counted per impression — the denominator a display CTR uses, and the one to quote against advertising figures. A visitor clicks a recommendation because it is the content; a banner interrupts the content. The gap is about four to one either way, and it is the most actionable number on this page.

0%5%10%15%20%25%

Product recommendation campaignsrecommendation widgets on the site

11.06%n=16

Targeting & personalization campaignsbanners, popups and inline content

2.63%n=24

Median account Middle half (25th–75th percentile) Full range

Hairlines that reach the right edge continue past 25% — the highest account on this chart is 66.89%. Exact ranges are in the table below and on hover.

Article recommendation campaigns clear the floor at only 4 accounts, below the threshold used everywhere else here, so they are excluded rather than published thin. A/B test campaigns are counted inside the campaign click rate but are not split out for the same reason.

How it compares, denominator for denominator

Display advertising
clicks per impression~0.5%

Product recommendations
clicks per impression8.03%

Marketing email
clicks per recipient~2.3%

Recommendation campaigns
clicks per visitor shown16.08%

Only the last line is ours. Display advertising is the 2026 cross-industry average from WordStream’s display benchmarks; marketing email is the 2026 all-industry average click rate published by Brevo. Three channels with different denominators, so each of ours is placed against the one it matches: per impression against display, per recipient against email. It is still three different kinds of attention — an ad on someone else’s site, an email in an inbox, and a block on a page the visitor chose to open — so read it as an order of magnitude, not a controlled test.

All figures

Median, top quartile and interquartile range across qualifying accounts, over a 90-day window in 2026.
Metric	Median	Top quartile	Middle half	n
Rec click → order session	11.59%	20.65%	7.75% – 20.65%	10
Rec share of order value	3.50%	6.71%	2.32% – 6.71%	13
Mobile share of sessions	53.83%	71.69%	29.36% – 71.69%	32
Focus time per session	34s	65s	18s – 65s	32
Campaigns delivering per account	24	47	9 – 47	32

What each figure means

A benchmark without its denominator is a slogan. These are the definitions every number above was computed against.

Delivered actionOne personalization action rendered to one visitor: a banner, a popup, an inline block, a recommendation widget. The denominator of every click rate on this page.

Click rate, per visitorSessions in which the visitor clicked the campaign, divided by sessions in which the campaign was shown to them. This is the per-person rate, comparable to an email click rate.

Click rate, per impressionClicks divided by deliveries. Lower than the per-visitor rate, because the median session is shown 1.45 actions. Comparable to a display advertising CTR.

Product recommendation CTRClicks on a recommended item, divided by recommendation impressions. An account qualifies at 500 impressions in the window.

Order sessionA session containing a completed product goal. “Rec click → order session” is the share of rec-click sessions that reached one.

Rec share of order valueProduct-goal value attributed to a recommendation, divided by all product-goal value. Attribution is session-scoped, not last-click across visits.

Focus timeTime the page was in the foreground and the visitor was active — not time-on-page from first to last request.

How to use these numbers

Comparing a single campaign against a median account is the most common misuse, so here is the honest reading.

Compare like with like.A recommendation widget belongs against the 11.06% row, not against the 6.07% blended figure. A newsletter popup belongs against 2.63%.

Read the range before the median.The middle half of accounts on campaign click rate runs from 2.54% to 15.73%. Being at 4% does not mean something is broken; being at 0.4% does.

Judge the placement, not the copy, first.The widest gaps in this data are between campaign types and placements, not between one headline and another. Move the block before you rewrite it.

Match the denominator before you compare.An email click rate is per recipient, so compare it with the per-visitor column. A display CTR is per impression, so compare it with the other one. Mixing them moves a number by about a third.

Value beats clicks.Recommendation CTR is easy to move and easy to fool. The 3.50% share of product-goal value is the number that survives scrutiny.

How the population was chosen

A benchmark is only as honest as its denominator. Signing up is not running personalization, so accounts had to earn their way in.

2,649Accounts on the rosterEvery account ever created

2,637Switched onNot administratively disabled and not paused at a page-view limit — a paused account is answered by the tracker with a no-op

117Running personalizationAt least one active campaign carrying a production tag, and at least one live action

32In the benchmarkDelivered at least 500 actions inside the 90-day window

The window for this edition is 21 June to 19 September 2026 — 90 consecutive days ending on the day the figures were read. A re-run replaces this page rather than adding an edition beside it, so what is above is always the most recent set published.

Two filters are applied by the platform rather than by us, which is worth stating plainly: campaigns published only to staging never reach the summaries, and visits an account tags as testing — its own team, on its own office IP — are discarded before archiving. No internal traffic reaches any figure on this page, and no individual customer is identifiable in any of them.

What this report does not claim

A benchmark is judged on what it refuses to say.

No industry breakdown.Site type is inferred by an AI reading of the customer’s own site and is recorded for only 5 of the 32 accounts. Cutting by vertical would mean publishing categories of two or three accounts.

No campaign conversion rate.The per-campaign goal counter is populated for exactly one account in the set. Campaign-dedicated goals are stored as separate rows this pass did not read, so the number is recoverable — but it is not in this edition.

No new-versus-returning split.New users and total users are counted on different bases in the daily summaries; ten accounts return a ratio above 100%, which is impossible. Dropped rather than patched.

Content recommendations are too thin.Only 4 accounts record any article recommendation views above the floor. Product recommendations carry that section.

n is small.32 accounts overall, 10–16 on the recommendation figures. Enough to publish medians with the spread shown; not enough to slice further.

Frequently asked questions

What is a good click rate for a website personalization campaign?8.26% of the visitors who are shown a campaign click it, at the median account, and 17.13% at the top quartile (n=32). Counted per impression rather than per visitor the same accounts read 6.07%. The campaign type matters more than the industry: recommendation widgets are clicked by 16.08% of the visitors who see them, banners and popups by 4.18%.

What is a good product recommendation CTR?16.08% of visitors shown a recommendation campaign click it (n=16 accounts). At the level of the individual recommended item, 9.82% of visitors who see recommendations click one, which is 8.03% of impressions (n=14 accounts with at least 500 impressions in 90 days).

Why are there two click rates for the same campaigns?Because there are two honest denominators. Per visitor counts the sessions in which someone clicked, over the sessions in which they were shown the campaign — the same shape as an email click rate, which is per recipient. Per impression counts clicks over deliveries, the same shape as a display advertising CTR, and reads lower because the median session is shown 1.45 actions. Both are published here; a vendor quoting one without saying which is not telling you much.

How much revenue do product recommendations drive?3.50% of product-goal value is attributed to recommendations at the median account, and 6.71% at the top quartile (n=13). Separately, 11.59% of recommendation clicks lead to an order session (n=10).

How does personalization compare to email or display advertising?On-site product recommendations are clicked at 8.03% at the median account. For scale, published 2026 benchmarks put marketing email click rates near 2.3% and display advertising near 0.5% — different channels with different denominators, so read it as an order of magnitude, not a like-for-like test.

How many accounts are in this benchmark, and why so few?Thirty-two. The roster is 2,649 accounts, but an account only counts if it is switched on, has at least one active production campaign and one live action, and delivered at least 500 actions inside the window. Signing up is not running personalization, and a benchmark that counts dormant accounts is measuring nothing.

Is internal or staging traffic included?No. Campaigns published only to staging never reach the summaries, and visits an account tags as testing — its own team, on its own office IP — are discarded before archiving.

Can I cite these numbers?Yes, with a link to this page. Every figure states its denominator and its n so it can be quoted without losing the context that makes it true.

Citing this report. Every figure may be quoted with a link to this page. Please keep the n and the range attached to any number you use.

Personyze. Website Personalization Benchmarks, 2026 edition. 32 live accounts, 15,944,591 delivered actions, 21 June to 19 September 2026. https://www.personyze.com/website-personalization-benchmarks/

Where these numbers come from

Personyze runs personalization, recommendations, A/B testing and on-site chat for the accounts measured here. The figures are read from the same daily summary tables those accounts see in their own reporting, aggregated so that no individual customer is identifiable. If you want to compare your own site against them, the fastest route is to run one campaign and read the same metrics back.

Start free Book a demo