One Hundred Thousand Page Views
One hundred thousand page views, two thousand five hundred purchases and a quarter million in revenue: as a report, accurate. As an answer, empty. The same figures produce a conversion rate of 2.6%, a revenue per visitor of 260, an average order value of 100, a growth trend of 15% month over month and a confidence level that says the season was real. Nothing in the data changed between the report and the answers. Only the question was asked properly. Standard metrics tell you what happened. Calculated metrics tell you what it means.
The Conversion Rate That Was Not the Same
The governance problem begins with a word. Ask five dashboards for their conversion rate, and you will receive five definitions: purchases over visits, orders over unique visitors, first orders over new sessions, revenue over visits, and one of them quietly counting only half the site. Each is defensible. Each is different. At the quarterly review, the business hears one word and reads five numbers, and the argument that follows is not about data at all. It is about definitions, and definitions are the one component of analytics nobody has ever appointed an owner for.
The trap runs deeper, into the platforms. A calculated metric carries its formula across report suites, and the formula only checks whether a variable with the same technical name exists. If eVar1 is traffic source in one suite and product category in another, an embedded segment still executes, the metric still renders, and the results are unpredictable and meaningless, delivered in the same format as always. The logic validates the name and never the meaning. Someone has to govern semantic consistency, and that someone is usually nobody until the definitions are governed by design.
Two Lives of a Metric
Adobe Analytics gives every metric two lives, and the discipline lies in the promotion between them.
The first life is project-only. A metric is born inside a freeform table, assembled on the fly by right-clicking a selection, and it lives within that Workspace project and dies with it. This is the rapid hypothesis life: build the metric, test its usefulness against the current analysis, and let the project delete it without ceremony if it proves nothing.
The second life is permanent. A promoted metric is saved to the account library, owned by its creator, transferable and shareable, and depended on by every dashboard in the building. Conversion Rate and Revenue Per Visitor belong here: reusable definitions that many reports can hold in common.
The act between the two lives is governance. A metric that survives its project earns promotion, and promotion isn’t a save button. It is a title that says what it measures, a description that says how it is calculated, tags that make it findable in a growing library and a human-readable summary a reviewer can understand at a glance.
The Weights You Choose
Calculated metrics also encode judgment, and the weights are where it hides. A page view can be weighted one, a video view five and a social share ten, and the resulting engagement score says something real about the brand, but only if those weights say what the business believes. The same view is worth more than a share to no brand on earth, and a weighting copied from another organisation imports another organisation’s idea of value.
The formats do quieter work in the same spirit. The same base metric can serve several audiences by changing its dress: Average Time on Page as a decimal for the analyst, and in minutes and seconds for the executive briefing, without touching the underlying report. And the uptrend control asks the author which direction is good, one of the rare places a tool admits that a number has no meaning until someone declares the meaning. Green for conversions, red for bounces. The declaration is the governance.
The Library Is the Language
Promoted metrics live in the library, and the library is where the organisation’s definitions become a language. When the analyst and the stakeholder read the same definition, the debate moves from whose number is right to what the number says about the business, which is the only debate worth having.
The library also carries one honesty note that most reports never learn. When a calculated metric nests another one inside its formula, the builder expands the nested metric’s full definition, but the changes made inside the nesting will not update the original. Treat a nested metric as a snapshot of logic at the time of insertion, not a live reference, and the drift that follows will never surprise anyone.
The Audit of Definitions
Two lists settle the state of the library. The first is the inventory: every conversion rate, revenue per visitor and engagement score at work in the current dashboards, each with its definition written in one line. Count how many agree. The second is the library itself: check that each permanent metric carries a title, a description and a human-readable summary, and that every definition has an owner.
Then three verdicts. Duplicates are merged into one promoted metric. Project-only metrics that earned their place are promoted with their paperwork. And definitions nobody wrote down, the numbers living in dashboards with no record of their logic, are retired before they confuse another meeting. After the audit, the library is the single source of truth the stakeholders trust, because they have read every definition and signed them.

Want to Know More?
If you want to know more, check out Adobe Analytics: A Champion’s Handbook. The calculated metrics chapter goes deeper than this post can, across the full function library of statistical functions, regression and confidence intervals, and the governance frameworks, validation techniques and reusable formula patterns that turn a builder into a library.


Leave a Reply