Why most landing page benchmarks fail a comparability test, and what to do instead

Landing page software benchmark research should begin with a comparability test. No England-wide benchmark for landing page conversion exists, so a reader cannot borrow one and must build a controlled internal reference.

A published rate is unusable for an England decision when the source does not define its eligible pages, geography, traffic, event, denominator, collection period, missing data and privacy conditions. Familiar terminology is not enough.

This article provides a study design for deciding whether external evidence can be used or whether the buyer should build that reference.

What to take away

Write the benchmark question

Specify the decision before searching. Name the proposed metric, eligible organisation and page population, England geography rule, page purpose, traffic sources, device scope, event, numerator, denominator, period, timezone and minimum evidence condition. Define what difference would lead to an action, but leave the threshold blank pending buyer and statistical approval.

A useful question is narrow: for example, whether one current page differs from a comparable internal group under the same measurement rules. "What is a good conversion rate?" has no stable population or event.

Publish the search protocol

Record the research cut-off, databases and websites searched, search strings, language, date range, inclusion and exclusion rules and reviewer. Prefer original methods and data over summaries. Keep England, UK and global sources in separate columns.

The UK Statistics Authority's Standards for Official Statistics govern official-statistics producers. Their quality practices still offer a sound screening discipline for commercial evidence: check source suitability, method, comparability, limitations and uncertainty. Using the framework does not make the resulting benchmark official statistics.

Named sources fail the comparability test for an England landing page decision. The table records the verdict for each source.

Take Unbounce's published median landing page conversion rate. The comparability test fails at several fields. Geography is not England. Eligible pages are self-selected users of one platform.

Traffic mix and device scope are not fixed. Event and denominator are not defined for the reader's pages. Collection period and privacy conditions are not stated for England. A named statistician should record that verdict before use.

Extract a complete evidence row

For every candidate source, capture:

Reject a rate when a decision-critical field is unavailable. Do not infer an England subset from a UK or global aggregate. A supplier benchmark can describe its participating accounts under its own method, but it cannot establish an external market norm without evidence of coverage and independence.

ONS quality and methodology information explains, for official statistics, how relevance, accuracy, timeliness, accessibility, coherence and comparability help users assess fitness for use. These dimensions are more useful than copying a headline average.

Test privacy and technical comparability

Two studies may observe different portions of the same visitor population. Cookies, pixels, server logs, consent choices, browser controls and identifier joins can change coverage. The ICO's final analytics exception guidance sets conditions for a particular statistical-purpose use and excludes online advertising from that exception. The actual configuration needs qualified review.

Record whether each source measures people, browsers, sessions, starts, button clicks or accepted submissions. Do not merge them. If the source changed its collection rule between periods, treat the series as broken unless its method explains a defensible bridge.

Build an internal reference when external evidence fails

Pre-register the internal page population and metric card. Use consistently defined historical or concurrent pages only when their audience task, offer, traffic mix and collection are close enough. Preserve every page-period observation rather than reporting only a mean.

Illustrative internal reference table. All figures are examples, not published data.

State the distribution, sample count, missing pages, exclusions and uncertainty without inventing a target.

HM Treasury's Magenta Book covers evaluation in central government. Its emphasis on designing evidence around the question and interpreting results within limitations is applicable as a method discipline, not as endorsement of a commercial benchmark.

Report the evidence gap

The final record should list included sources, rejected sources and reasons. If none pass, say that no comparable benchmark was found by the cut-off date. The next step is to improve the study or internal reference, not to average incompatible figures.

Before you act

Common questions

What is the main output of the comparability test?

The test records a verdict for each source. A pass requires every decision-critical field to be defined; otherwise the source fails or remains unclear.