Research question and scope
This protocol measures how often a knowledge article is past its review date. The unit is one published article in one knowledge base. An overdue article has a recorded review date earlier than the review date and no completed review after that date. The measure describes review discipline, not whether the content is correct.
Set the scope to one knowledge base, one audience, and a fixed period. Record the review interval and the owner rule in force. If the interval changes during the period, split the sample at the effective date. A different interval creates a different denominator.
Keep the measure separate from a repeat-question count. An overdue article may still be accurate, and a current article may still fail to answer the question. The purpose is to test the review calendar. Report repeat questions separately.
Name the decision the result will support before collection starts. The client may use it to add review capacity, change an interval, or retire a page family. A report that only lists overdue pages creates guilt without a choice. Write the intended action beside the scope.
Record the knowledge base and the export format before sampling. A wiki export may include page history, while a help-center export may show only the current state. Note which fields are native and which must be inferred. The tool limits the questions the report can answer.
Population and unit of analysis
Define the population as every article visible to the support team or the customer during the period. Include articles in the internal wiki and the public help center, and mark the audience. The unit is the article, not the section or the link. A long page with five sections is one unit.
Include articles with no recorded review date. Mark them as undated rather than dropping them from the sample. Exclude drafts that were never published. Report the excluded count and the reason. An undated article is a finding, not a missing row.
Exclude a page that is a redirect to another article. Count the destination once and record the redirect. A redirect is a navigation aid, not a separate knowledge claim. State the exclusion rule before sampling so a second reviewer can apply it the same way.
Decide how to treat a page that was updated but not formally reviewed. Mark it as edited and report it separately from a completed review. An edit may fix a typo while leaving the policy example stale. The two states answer different questions.
Decide whether a page with no traffic stays in the population. A page that no one reads can still be a compliance record, while a high-traffic page carries immediate risk. Record the traffic band and the page type. The reader should see both before judging the overdue count.
Definitions that make the measure reproducible
Define the review date as the date on the article record, not the date in the page footer. Define a completed review as a decision by the named owner with a date and a result. Define keep, update, redirect, and retire as the four allowed results. Publish one example of each.
Define overdue in whole days from the review date to the sample date. A page due today is not yet overdue. Use the same time zone for every calculation. Record the formula in the report so the count can be reproduced.
Define the owner as the person or role named on the article record. A team inbox is not an owner unless a named person is accountable for the review. Record the owner field and whether it was blank. Owner coverage is part of the result.
Decide how to classify an article whose owner has left the team. Keep it overdue and add an owner-missing category. Do not reset the date because the owner changed. The reassignment is a separate action that should appear in the report.
Agree how to treat an article that is a duplicate of another page. A duplicate may need a merge rather than a review, and its owner may be the same person. Record the duplicate pair and the chosen action. Counting both as independent overdue pages inflates the workload.
Sampling and data collection
Draw the sample from the knowledge base export for the full period. Record the export date, the audience filter, and the row count. Do not build the sample from pages that received complaints. A complaint-based sample cannot show the quiet pages that are also stale.
Capture the article identifier, the title, the audience, the owner, the last review date, the next review date, the last edit date, and the current status. Add the page type and the traffic band where available. Store the export with the report. The review date and owner are the core evidence.
Check a small overlap sample with a second reviewer. If the reviewers disagree about whether an edit counts as a review, the definition is ambiguous. Rewrite the rule and retest before the full sample. Agreement matters more than sample size.
Record the time spent matching the export to the page status. A report that requires opening every article will not be repeated. Note the fields the knowledge tool does not expose. The collection cost belongs beside the result.
Sample the page status on the same day as the export. A review completed between the export and the check can make a current page look overdue. Record the export time and the check time. The gap is small but should be documented.
Measuring overdue reviews
Report the share of articles current, overdue, undated, and missing an owner. Show the counts behind each share. The four groups describe different failures and should not be collapsed. A single overdue rate hides the page that no one owns.
Break the result down by audience, page type, and review interval. A public policy page may age faster than an internal how-to. A quarterly interval may be too long for a pricing page. Compare only groups with a similar change rate.
Report the age of the oldest overdue article and the median days overdue. A single forgotten page is different from a whole family that missed a quarter. Include the page type and the owner status. The distribution shows where the review capacity is failing.
Report the share of overdue articles that were edited without a recorded review. An edit can create false confidence in a page that still carries an old policy. Include the last edit date and the editor where available. The two signals together describe the real risk.
Compare the overdue rate for pages with a named owner against pages without one. A large gap suggests the ownership rule matters more than the review interval. Record both counts and the ownership status. The comparison points to a fix that does not require more reviewing hours.
Limitations and next action
The measure cannot show whether an article is accurate or whether readers found it useful. It also cannot show a review completed outside the knowledge tool. Treat the result as a measure of recorded review. Note the bases and page types that were not included.
Give the result to the support lead and the content owner with one change to test. The change might be a shorter interval for policy pages, a monthly review block, or an owner reassignment rule. Record the effective date and repeat the measure with the same definitions. A controlled change is easier to judge.
Review the boundary cases before shortening every interval. A few stable how-to pages may not need the same cadence as pricing or policy. Fix the page-type rule first if that is the pattern. The goal is a page a reader can trust, not a lower overdue number by definition.
Schedule the next measurement for the same week in the following quarter. Keep the page types and audience filter unchanged so the comparison is fair. Record any change in the review tool or the owner rule. A trend built from different definitions is not a trend.
