Teams can look at the same document analytics and reach different conclusions. Marketing may treat completion as a content-quality score, sales may read a revisit as interest, and the content team may see long viewing time as evidence that a page is difficult. The number is the same, but the unit, period, and meaning attached to it are not.
Before adding another score, create a metric dictionary. Define what was observed, which scopes may be compared, what the signal may support, and which conclusions remain off limits.
A list of metric names is not enough
Each dictionary row should include the metric name, where it appears, its observation unit, aggregation period, inclusion and exclusion rules, allowed use, prohibited inference, owner, and last change date. “Time spent” alone is ambiguous: one teammate may mean a viewer’s cumulative time while another means time on a specific page. “This week” may mean the last seven days or Monday through today.
Start with one analysis question. “Which pages in this distribution need more explanation?” and “Which questions should we prepare for one recipient?” are different questions. Even when they use the same metric, keep document-level and viewer-level scopes separate, and fix the period and link scope before comparing anything.
Treat completion as a measure of reached depth
If completion is labeled only high or low, it quickly becomes a quality score. Record which document, period, and viewer scope the displayed completion rate covers. Limit the allowed statement to something observable, such as: “Compare how often viewers in this scope reached later pages.”
A low completion rate does not prove that the content is poor. A reader may have found the answer early, run out of time, or received a document that was not relevant. Completion is not purchase readiness and does not qualify an MQL or SQL. It is a signal that helps narrow the pages worth checking next.
A revisit records another opening, not its reason
A revisit tells you that the document was opened again. It does not tell you whether the viewer liked the pricing, forwarded the material, or could not find a needed answer. Define how the same viewer is grouped and whether the first and latest viewing timestamps are reviewed together.
An allowed use might be: “Which page in the reopened document should we ask about next?” Prohibited inferences include purchase intent, contract probability, and automatic sales-stage advancement. A revisit may change the order of questions, but it cannot answer them.
Keep multiple explanations open for time spent
FeatPaper provides cumulative viewing time by viewer and time spent by page. Do not combine them in one dictionary field. Decide whether the analysis compares total viewing time or focuses on a particular page. Also distinguish an active viewing session from a finalized record after the session ends.
Long time spent may reflect attention, comparison, confusion, or simply an unattended screen. An allowed statement is: “Use the page where time accumulated as a candidate for a follow-up question.” Do not infer that long viewing proves strong purchase intent. Refusing to assign one meaning to this metric is part of the dictionary’s purpose.
A click is a selected path, not a conversion result
A click on a link inside a document records which next path was selected. It does not establish that an inquiry was submitted, a meeting was booked, or revenue was created. Define the click using the actual item available in the product view, and keep the CTA name and destination together.
Do not combine clicks, completion, and time spent into an invented buying score. Instead, ask whether the clicked and unclicked paths were clear to readers. Revenue contribution and conversion cannot be guaranteed or attributed from document behavior alone.
Give every exclusion rule an effective date
Internal QA, production checks, and sales-team tests can distort a baseline when mixed with external viewing. For any analytics exclusion setting, record the owner, effective date, excluded scope, and the results of one internal and one external control test. Distinguish the rule that excludes a logged-in document owner’s own views from any additional range your team configures.
Do not assume that changing an exclusion today recalculates past data. If a comparison crosses the change date, mark the boundary and describe the two periods as operating under different conditions.
The dictionary changes the conversation, not the numbers
A good metric dictionary is not an answer key. It is an operating agreement that gives everyone the same scope and the same limits. Before the next analytics review, choose four metrics and complete five fields: unit, period, exclusions, allowed statement, and prohibited statement.
When someone says “interest is high,” ask which record, period, and scope support that phrase. If the team can answer in the same format, it has not merely collected more numbers. It has built a shared language for overstating them less.
