Philippines staffing research
How stale are knowledge-base articles at their review date?
Contents
A documentary study of knowledge-base review dates, source currency, and the gap between an article's stated review schedule and the evidence available when the date arrives.
Question and scope
This study asks how often a knowledge-base article reaches its review date without a current source, an owner, or a decision to keep, revise, or retire it. It measures review readiness, not article popularity. The unit is one article at one scheduled review date. An article can be old and still correct; the question is whether the review produced evidence for that conclusion. The scope includes the article metadata, the cited source, the owner, the last substantive change, and the review record. It excludes the quality of the answer in a live customer interaction unless a separate sample is linked to the article. A review that only changes the date without checking the source is recorded as a metadata-only review, not as a substantive review.
Method and evidence frame
Use a structured document review. For each article in the sample, identify the scheduled review date, the last substantive change, the current source, and the named owner. Check whether the review record shows what was checked, who approved the outcome, and what changed. Record the result as current, revise, retire, or unresolved. A missing review record is a finding even if the article happens to be correct. The evidence frame separates the article content from its review evidence. The content may be accurate, but the review may be absent. The source may be current, but the owner may have left. The review may be complete, but the article may still be wrong because the source itself changed. Each layer is recorded separately so the report does not collapse them into a single stale/not-stale label.
Population and unit of analysis
The population is all active knowledge-base articles in the selected queue or topic area during the observation window. The unit is one article at its scheduled review date. An article reviewed twice in the window is two units if the second review is a separate scheduled event. A retired article is outside the active population but is recorded in a separate retirement sample. Define whether articles without a scheduled review date are included. Include them as an 'unscheduled' category, because an article with no review date cannot be assessed for review readiness. Report the count of unscheduled articles separately rather than excluding them from the population.
Definitions and measures
The primary measure is the share of articles whose scheduled review produced a documented decision with a source check and a named owner. A secondary measure is the share whose review was metadata-only. A third measure is the share whose cited source could not be found or was older than the article's own review rule. Report counts and denominators for each. Supporting measures include the age of the last substantive change, the number of broken links, the number of articles with no owner, and the time between the scheduled date and the completed review. Record whether the article was used in a customer-facing answer during the window, because usage changes the risk of a stale answer but not the evidence of the review itself.
Sampling and collection
Sample across topic areas, risk levels, and review dates. Include articles that are overdue, articles reviewed on time, and articles with no review date. Include at least one high-risk article involving money, access, safety, or a legal or customer commitment. Record the sampling frame and every exclusion. A sample of only recently reviewed articles will not show the overdue population. Collect the article metadata, the review record, the cited source, the owner record, and the change history. Capture the page as it appeared at the review date if the system supports versioning. If it does not, record that limitation and use the closest available version. Do not reconstruct an old article from memory or from a search snippet.
Analysis and reporting
Report review readiness by risk level and topic. A high overall rate can hide a low rate for the articles that matter most. Group defects by no review date, no owner, missing source, broken link, metadata-only review, review without approval, and overdue review. Show the count and share for each category and the number of articles affected. Separate observation from interpretation. A metadata-only review does not prove the content is wrong; it proves the review did not test it. A missing source does not prove the answer is false; it proves the reviewer cannot verify it. The report should avoid turning a review-process defect into a claim about customer harm without a separate case sample.
Limits and sources of error
Knowledge-base platforms vary in their version history and review workflows. A review may have occurred in a meeting or a chat and never reached the article metadata. The reviewer should record the system limitation. Small samples are sensitive to one large topic area, and a recent migration may reset dates without reflecting a real content review. The study cannot establish that an article is legally compliant, factually correct in every case, or safe for every customer. It tests whether the review left evidence. NIST, CISA, and the U.S. Government Accountability Office provide general guidance on information quality and control, not a content standard for a specific queue.
Operational use
Use the result to add an owner field, require a source check, or retire articles that no longer support a decision. A defect concentrated in one topic area may point to a missing subject-matter reviewer. A defect spread across the library may point to the review template or the cadence. Choose one change and re-sample the same risk tier after the change. A useful review is short enough to complete and specific enough to defend. The article can be kept, revised, or retired, but the record should say which one and why. That is more valuable than a date change that makes the library look current without testing it.
Review cadence
Review the knowledge-base process quarterly and after a major product, policy, or system change. Keep the sample rule stable across the comparison. If the review template changes, label the new period and do not compare the new readiness rate directly with the old one without explanation. Close the review with one owner, one change, and a re-test date. Preserve the sampled articles and the review records so a later reviewer can repeat the method. An article that is retired should also have a record of where its replacement lives and when the old link stopped being active.