Skip to content

What your recorded results support

Reading the performance recommendations, the sample size every claim must clear, and the list of questions the recorded data cannot answer.

Updated

The recommendations page reads your reporting and attribution and states, at most six times, what your own recorded rows support. Every recommendation shows what it was counted from.

Nothing here predicts and nothing here learns. The rules are fixed and deterministic: no model produces this output, nothing on this page is counted from a test arm, and nothing on this page is ever applied automatically. A recommendation never edits a brand rule, a writing profile or any setting you chose.

The gate is a sample of five. Where fewer than five rows support a comparison, no claim is made at all, rather than a weaker one.

BEFORE YOU START

  • The "view" permission for the workspace.
  • Recorded metrics or conversions. Nothing on this page is derived from anything else, and no benchmark or example fills a gap.

STEPS

  1. Open Recommendations for the workspace, or go to /app/workspaces/{workspace}/recommendations.
  2. Read the note at the top. It states that these are descriptions of recorded data, not predictions.
  3. Read each recommendation's kind: Repeat, Stop, or Test next.
  4. Open the evidence list under a recommendation. It names the basis, the data source, the date range, the sample size, the comparison basis, the limitations, and which platforms were included, excluded and never recorded.
  5. Check the basis label. "Measured" is counted from recorded rows; "Observed" is a recorded process fact rather than a performance measurement; "Absence" is the absence of rows, which is not a measurement of zero.
  6. Read "What this data cannot tell you" at the bottom. It is always shown, and each entry says what would make that question measurable.

WHAT YOU SHOULD SEE

Up to six recommendations, each with its evidence, or a statement that the data does not reach the sample size needed. The page never fills the space with generic advice.

WHAT THIS WILL NOT DO

  • It will not tell you the best time to post. That question is listed among the ones the recorded data cannot answer, with what would be needed to answer it.
  • It will not say which hook, format or offer performed best. Conversions link to a contact and to nothing else, so recorded revenue cannot be traced to what was being promoted.
  • It will not rank a platform nobody reported. A platform excluded from a comparison is named, with the reason, rather than silently dropped.
  • It will not break a tie. Where two platforms or channels are level, no recommendation is produced.
  • It will not offer date controls on the page. The window can be set in the address only.

WORTH KNOWING

The reasoning that produces a "Stop" on paid spend is careful about one thing in particular: recorded zero revenue may mean the spend returned nothing, or may mean a return was never ingested. The recommendation says so and asks you to confirm ingestion is running before treating it as a reason to cut budget.

IF IT DOES NOT WORK

  • "Nothing has been recorded for this workspace yet, so there is nothing to learn from." means no metrics or conversions exist.
  • "Data has been recorded for this period, but none of it reaches the sample size needed to support a recommendation. No claim is made rather than a weak one." means the gate was not cleared.
  • "view permission required." is the permission refusal.
  • "You do not have access to this workspace." means your access does not reach it.

COMMON QUESTIONS

Does this get better the more I use it?

No. The rules do not adapt and nothing is trained. What changes is the amount of recorded data they read.

Why only six?

Because the rules produce at most six, in a fixed order of what is blocking. A longer list would be padding.

Is there an experiments or A/B feature?

There is an Experiment Registry, and it is a different thing from what that question usually means. Social Studio still does not split an audience, assign anybody to an arm or measure a lift for you. What the registry does is record a test you ran yourself and govern what may honestly be concluded from it. A "Test next" recommendation here is a good candidate to register there.

  • Reading your reporting

    What the workspace report totals, why a figure can say "Not reported" instead of zero, and the two ways a figure reaches it.

  • Attributing revenue to channels

    Reading first-touch, last-touch and linear credit over your recorded touchpoints and conversions, why some conversions are unattributed, and why two currencies are never added together.

  • Planning the week with the Social Operator

    How the Social Operator gathers evidence from your own recorded activity, proposes at most six things to do this week, and why it never acts on its own.

  • Recording an experiment

    How to register a test in the Experiment Registry, record what each arm measured, and read the outcome the evidence gate decides — including why some results say nothing at all.