Skip to content

Keep changing metrics as observations instead of overwrites

Store metric values with observation times and source context so later differences remain interpretable.

Guides

Keep changing metrics as observations instead of overwrites — what should the operator record?

A displayed metric can change between two authorized reads, and a table that overwrites the first value loses the basis for comparison. Keep observations separate from stable record identity. Differences between observations describe those returned values, not necessarily activity occurring entirely within the interval or a complete account history.

Practical operations check
Store metric values with observation times and source context so later differences remain interpretable.
Use source_key, metric_name and observed_at as a local composite key, preserving the metric definition and unit alongside the returned value. For example, attach the native metric label and unit to each value so a later rename cannot silently make two noncomparable observations look like one series.
Call the result a difference between observed values and keep negative or unexplained changes visible rather than forcing a monotonic growth story. A negative observed difference may reflect a correction or visibility change; preserve it with uncertainty rather than deleting the observation to force an always-increasing series.
01 / Guides

Define the observation key

Use source_key, metric_name and observed_at as a local composite key, preserving the metric definition and unit alongside the returned value. For example, attach the native metric label and unit to each value so a later rename cannot silently make two noncomparable observations look like one series.

02 / Guides

Preserve unknown values

Distinguish unavailable, absent and numeric zero; do not impute a missing earlier value merely to make a delta computable. A returned zero is an observed numeric value; a missing property cannot safely become that value merely because the downstream chart expects every cell to contain a number.

03 / Guides

Append a new observation

Add the later returned value under its own receipt and coverage state instead of replacing the old cell with a latest label. If two receipts arrive at the same local timestamp precision, retain distinct receipt references rather than letting a timestamp-only key overwrite one observation with the other.

04 / Guides

Check comparability

Confirm that both observations use the same metric meaning and object identity before calculating a local difference; source definition changes invalidate simple deltas. Check whether values describe the same object and scope before subtracting; a whole-account metric and a single-item metric are not interchangeable despite similar labels.

05 / Guides

Label derived differences

Call the result a difference between observed values and keep negative or unexplained changes visible rather than forcing a monotonic growth story. A negative observed difference may reflect a correction or visibility change; preserve it with uncertainty rather than deleting the observation to force an always-increasing series.

Keep the boundaries clear.

  • Store metric values with observation times and source context so later differences remain interpretable.
  • Operational record: Define the observation key. Context badges do not certify access.

Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.

Primary references

Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.

Know what social media shows about you.
Start with a question.

Get started
montage