Direct answer: an OnlyFans Twitter/X post-format test log should record one observable publishing variable, such as text-only versus single-image, across creator-approved matched posts. Pre-write the audience question, topic match, format definitions, fixed copy elements, primary X metric, observation window and decision rule. Alternate or pre-assign variants, save the Post IDs, and mark policy, publication and timing exceptions. Compare only runs with the same metric scope. The log supports a next format choice for this account; it cannot establish a universal benchmark or guarantee reach.
A post-format test log is a lab notebook for public publishing choices. Its value comes from what was decided before the results appeared. Without that record, a creator can always explain the favourite post after the fact.
For a wider test that spans a whole promotional journey, use the promotion test log. This page stops at the X post and its directly available activity evidence.
Write the question as a format comparison
Use: “For [one recurring public topic], how does Format A compare with Format B on [one defined response] over [one window]?” The topic and creator voice stay within the same approved territory.
A valid factor is observable: text-only versus single image, one image versus a short clip, direct statement versus question opening, or short copy versus a creator-defined longer band. Choose only one.
“Casual versus polished” is not ready until the visible differences are specified. It may hide changes in media, copy, styling and production quality.
Freeze both format definitions
Write the required elements, allowed range and exclusion for each variant. For a text-versus-image test, A may be public text with no attachment while B uses one creator-approved image and the same copy structure.
Record the version and approval timestamp. If the creator later adds video to B, begin a new test version rather than expanding the category.
Both variants must pass the same current platform-suitability and creator-approval gate. A restricted or removed post is an exception, not a normal performance row.
Choose matched public topics
Create pairs from the same topic territory and audience question. The words and examples can differ naturally, but the underlying job should match. Comparing a humorous personal observation with a major announcement tests more than format.
Record a topic code, source and asset readiness before assignment. Do not match posts only after seeing their activity.
If a current event changes the context for one post, keep it and mark the event. The review can exclude it under a pre-written rule without erasing the evidence.
Pre-assign the sequence
Alternate A and B, randomize within prepared pairs or use another sequence chosen before publication. Store the assignment list. Publishing all A posts first and B posts later can confuse format with account growth, season or campaign phase.
The log is not a schedule recommendation. It records the creator's existing publish window only so timing differences remain visible.
Never swap a variant after previewing early activity. A production failure should be logged as a failed run and replaced under the stated rerun rule.
Define the primary X metric
Choose a metric the creator can retrieve consistently from the same X surface: impressions or views, defined engagements, earned engagement rate, link clicks or another currently supported field. Copy the interface label and its scope.
X states that view counts are not unique and can include more than one view by the same account. Preserve that limitation rather than translating views into people.
Use one primary metric for the decision and keep secondary fields descriptive. A test that picks whichever metric favours a variant after publication has no stable decision rule.
Fix the observation window
Capture every run after the same elapsed interval, such as a creator-chosen number of hours or days. Store the source time zone, capture time and whether the post remained public for the full window.
If a post was edited, limited, removed or unavailable during the window, mark the affected duration when known. Do not compare its final total as though exposure conditions matched.
Late captures should remain Late snapshot. Do not backdate them or estimate the intended value.
Copy the X post-format test log
Artifact: record Test ID | version | pair ID | topic code | creator-approved asset | variant | assigned order | Post ID | publication time | capture time | observation window | primary metric | metric value | secondary observations | suitability state | exception | valid-run state | reviewer.
| Pair | Variant | Topic job | Views | Defined engagements | Run state |
|---|---|---|---|---|---|
| 1 | A: text-only | Studio lesson | 1,100 | 62 | Valid |
| 1 | B: single image | Studio lesson | 1,260 | 70 | Valid |
| 2 | A: text-only | Routine question | Unknown | 48 | Late capture |
Log activity without merging scopes
Save raw values first. If the X activity surface supplies an earned engagement rate, preserve that value and definition. If the creator calculates a rate, store its numerator, denominator and formula separately.
Do not merge organic and promoted rows, different metric date ranges or public view counts with dashboard impressions unless the platform documentation supports the equivalence.
When a field disappears or changes meaning during the test, close the current metric version. A stable comparison matters more than completing the planned number of rows.
Work through a four-pair example
A creator tests text-only against a single approved image for the public topic “one studio lesson.” Copy length, tone, topic depth and call to action stay inside fixed ranges. Four pairs are pre-assigned B-A, A-B, B-A and A-B.
Pair one records 1,100 views and 62 defined engagements for A, compared with 1,260 and 70 for B. Pair two A is captured late and becomes invalid under the frozen rule. Pair three coincides with an unrelated viral mention and is marked Context exception. Pair four is valid.
The primary review contains pairs one and four. The other rows remain visible. Two valid pairs can support another test cycle or a cautious local preference, not “images always win on X.”
Compare pair direction and totals
For each valid pair, note whether A, B or neither had the stronger primary metric under the decision rule. Then show raw values and, when definitions allow, pooled totals. Pair direction prevents one unusually large post from hiding mixed results.
Keep production effort as a separate field if it affects the choice. Do not turn views, engagements and minutes into one weighted “format score” without a defensible decision model.
State how many valid pairs were reviewed, how many were excluded and why. The denominator of the experiment is part of the finding.
Preserve the pair table even when the final choice is obvious to the creator. The raw sequence shows whether the direction repeated, whether one large event dominated the totals and whether exclusions were balanced across variants. That context is more useful than a standalone percentage.
If one variant systematically takes longer to prepare, report the median creator-entered production minutes with the response evidence. Label estimates and logs separately. The effort difference may guide the next test, but it does not explain X distribution.
Use a bounded decision vocabulary
Choose Keep A for this topic, Keep B for this topic, Use both for different jobs, Repair and rerun, Collect more valid pairs or Close the test. Cite the pair IDs and metric version.
The choice can also favour the lower-effort format when response evidence is practically similar. Write that tradeoff directly instead of calling the formats equal in every respect.
Do not add a reach prediction. The decision governs the next creator-approved format assignment only.
Keep profile outcomes separate
Post activity does not establish what happened after someone visited the profile. If the creator wants to test the next journey step, use the profile conversion test log with its own visit denominator and evidence.
Link a Post ID to that experiment when appropriate, but do not claim that every profile or subscription action came from the test post without supported attribution.
This boundary stops an X format question from expanding into an entire funnel experiment.
Keep community work separate
Replies, conversation selection and community time can influence post context, yet they are a different creator-controlled workflow. Record a major community event as an exception rather than adding reply volume as another format variable.
Use the X community engagement workflow to manage that work deliberately. Do not make community activity a hidden requirement of Format A or B.
If the creator wants to test a reply format, open a new, explicitly scoped test instead of changing this one.
Audit the log
Verify assignments existed before publication, both variants meet their definitions and observation windows match. Trace a valid and excluded row to the X source record.
Check for metric-scope drift, post deletions, promoted activity, late captures, topic imbalance and rules written after results. Recalculate any creator-entered rate.
Write the final statement with account, topic, dates, variants, metric and valid-pair count. That sentence is the true scope of the result.
Lock the completed log after review. Corrections should show the original value, updated value, reason, editor and timestamp so the evidence does not drift with the preferred narrative.
Limitations
Limitations: X distribution changes continuously, public view counts are not unique viewers, activity metrics can update, matched public posts still differ in context and small creator tests are noisy. This log compares one approved post-format variable on one account. It does not plan content pillars, compare channels, automate publishing or provide universal engagement and reach benchmarks.
Freeze definitions, preserve invalid rows, use current X documentation and treat the result as a local next decision.