What it is
cohort_dimension says which dimension a cohort is cut on — the axis its members are grouped along.
There are seven, and a cohort is cut on one of them or on several together, never on anything outside the set.
This value gives you the axis and not the point on it. It tells you a cohort is cut on acquisition period. It does not tell you which period. The value that names the point is not published, and a reader who needs the specific month reads the cohort’s own identifier instead.
How it is calculated
Measured. The dimension is part of the cohort’s definition — recorded when the cohort is created, not worked out from its members afterwards. Where no dimension is specified, acquisition period is used, wherever the account has one. Several dimensions can compose one cohort — enterprise segment, bought on the Pro tier, acquired in the third quarter. Each dimension added makes the population smaller, and small populations carry weaker conclusions. Below 20 accounts in a segment tier, results for that tier are not treated as reliable on their own: they are pooled with comparable segments and labelled as an estimate across segments, with the account count shown.Where it comes from
Each dimension is supplied by a different part of Beacon, and each depends on the sources behind it:
Where a source is not connected, cohorts cannot be cut on the dimensions that depend on it. The dimension is not approximated from something else.
Two dimensions can be missing for an account even with every source connected. An account with no won deal — usually a self-serve customer — has no acquisition period. An account with no first-value milestone set up for its segment has no activation date. Neither is filled from another date: not sign-up, not first payment, not first login. Each of these two dimensions is the month of the same event as one cohort type — acquisition period of Became customer, activation date of Reached first value — so a dimension and its type never disagree. See cohort type.
No read interface or MCP tool serves this value yet — the table above publishes the shape ahead of the doors.
How fresh it is
Fixed when the cohort is created and unchanged afterwards. Cutting the same population on a different dimension produces a different cohort, not an updated one. The underlying values move for individual accounts — an account changes segment, a fit score is recalculated, an adoption pattern shifts. Cohort membership does not follow them. It is frozen at the moment the cohort opens, which is what makes two cohorts comparable years apart.Currency and rounding
Neither applies. The dimension is a label.What changes it
Nothing after the cohort is created. A dimension change is a new cohort under a new definition, and the original is retired rather than edited.What it is not
- Not the value on the axis. This says a cohort is cut by segment, not which segment.
- Not
cohort_type. The type is what opened the cohort; the dimension is how its members are grouped once it is open. - Not
segment. Segment is one of the seven dimensions, and it is also a value in its own right on an account. - Not a filter you apply while reading. A cohort’s dimensions are set when it is created and are part of what it is.