Direct answer: freeze an evidence window for the niche already in operation, check whether execution was consistent enough to interpret, identify which audience signals repeat, and choose one branch: Continue the current angle, Narrow it around a clearer repeated interest, or run one Adjacent test while holding other variables steady.
- Continue: the promise is clear, delivery was consistent and evidence broadly matches it.
- Narrow: one sustainable sub-interest repeats across several independent records.
- Adjacent test: a connected question has evidence, while the current identity stays recognisable.
- Repair first: execution gaps make the present window too weak to interpret.
This decision tree starts after a niche has operating evidence. It does not choose a first niche, list supposedly profitable niches, design a full rebrand, or jump to an entirely unrelated identity. If there is not yet a selected niche and a defined test, begin with the niche selection guide.
A pivot is not a reaction to one quiet post or one enthusiastic comment. The Pivot decision tree slows the decision down long enough to separate an unclear premise from inconsistent execution, a narrow pocket of interest, or a plausible adjacent question.
Pass the evidence eligibility gate
Write the current niche statement in one sentence: audience interest, creator perspective, repeatable value, and boundary. Record when it started, which surfaces carried it, how often it appeared, and what was held consistent. A vague label such as “lifestyle” is not enough to evaluate.
Confirm the creator actually delivered the planned sample. Missing weeks, changing formats, inconsistent labels, or several simultaneous offers can make the evidence hard to read. When execution was incomplete, the first branch is often “repair the test,” not “replace the niche.”
Exclude extraordinary events and note them rather than silently averaging them in. Keep raw records, dates, and definitions. The eligibility gate does not demand perfect data; it demands enough context to know what the records describe.
Freeze a comparable evidence window
Choose an evidence window before reviewing outcomes. It could be a defined number of weeks or content cycles appropriate to the creator's cadence. Record start and end dates, included surfaces, content count, formats, and any planned absence. Do not extend the window only because the result feels disappointing.
Use comparable records: published items, audience questions, saves or replies where available, completion observations, creator workload notes, and offer enquiries. Define every signal in plain language. A view, reply, purchase, and repeated question describe different behaviours and should not be merged into one score.
Small windows remain noisy. Show raw counts alongside any rate and preserve no-response periods. The tree guides the next test; it does not turn a small sample into a market conclusion.
Separate execution signals from niche signals
Execution signals include inconsistent publishing, weak descriptions, missing assets, poor navigation, delayed replies, or quality defects. Niche signals include repeated audience language around one sub-interest, recurring confusion about the promise, or engagement concentrated on a specific angle.
Ask whether someone had a fair chance to understand and experience the current niche. If the presentation changed constantly, evidence about the niche is entangled with execution. Repair the controllable issue and run another comparable window before choosing a directional branch.
Use the niche validation scorecard as the source record when it already contains the current premise and evidence. This tree consumes that evidence; it should not duplicate or overwrite the scorecard.
Branch one: Continue the current angle
Continue when the niche statement remains clear, execution met the planned standard, repeated signals broadly match the promise, workload is sustainable, and no narrower interest clearly dominates. “Continue” means preserve the premise for another defined window, not stop learning.
Write what will remain constant and which single operational improvement, if any, will occur. If five changes accompany the continuation, the next review cannot distinguish normal variation from the effect of those changes.
Set the next review date and evidence question. A continuation without a date becomes inertia. The decision log should explain why the evidence did not support narrowing or an adjacent test.
Branch two: Narrow around repeated evidence
Narrow when a coherent sub-interest repeats across several records, the creator can sustain it, and the broader statement hides the value people actually recognise. Rewrite the niche sentence with a more specific audience job and boundary. Keep the connection to the existing identity visible.
Do not narrow around one high-value event or one loud respondent. Look for repetition across different items or checkpoints. Note contrary evidence as well as supporting evidence so the decision is not built from selective examples.
Create a transition window that compares the original and narrowed statement using equivalent formats where practical. Narrowing is a testable refinement, not a permanent declaration. Define what would cause a return to the broader angle.
Branch three: run one Adjacent test
An adjacent angle shares the current audience job, creator strength, or content capability while changing one meaningful dimension. It is not an unrelated niche selected because it appears popular. Write the connection explicitly: “Current evidence shows X; adjacent question tests Y because Z repeated.”
Create a test card with hypothesis, held constants, changed variable, content sample, evidence window, stop conditions, creator workload, and review date. Keep format, publishing cadence, and measurement definitions stable where possible. Change one main angle so the result remains interpretable.
Label the work as a test in internal planning. Do not remove the current niche identity before the evidence window ends. A bounded adjacent experiment preserves continuity and makes reversal straightforward.
Use public trend data as context, not verdict
Google Trends can compare exact terms or broader topics, but those choices are not interchangeable. Record spelling, geography, time range, category, and whether the query is a term or topic. A visible trend can inform wording research; it does not establish subscriber demand for an offer.
Compare related phrases only when they represent the audience language under investigation. Do not use an unrelated high-volume term to justify a full identity change. Public search behaviour and private subscriber behaviour are different evidence sources.
If Trends has insufficient data, record that state. Do not substitute a nearby query without disclosing it. The tree can proceed from direct creator evidence when public context is unavailable.
Write the decision record before producing more content
Record: current niche statement; evidence window; execution status; repeated supporting signals; contradictory signals; creator workload; selected branch; rationale; held constants; changed variable; stop condition; owner; and next review. Include links to the underlying records.
Turn the selected branch into a bounded content plan through the content strategy planning guide. This tree decides direction only. It does not create the calendar, select exact formats, or allocate production capacity.
Do not rewrite old results to match the new label. Preserve the prior niche statement and version. Version history makes it possible to see whether the pivot clarified the premise or simply renamed it.
Add an evidence table with one row per signal. Columns should include date, source, content or event ID, exact observation, signal definition, supporting or contradictory status, execution note, and reviewer. Grouping happens after the observations are recorded. This order reduces the temptation to keep only records that support the preferred branch.
Include a “cannot interpret” state for records affected by broken links, missing publication, incorrect labels, or a changed format. Excluding a record needs a written reason. The raw record remains available even when it is not used in the directional decision.
Review the next evidence window
At the review date, confirm the planned sample was delivered and compare the defined signals. For an adjacent test, evaluate the changed angle against the held-constant baseline. For a narrowed test, check whether audience language and creator workload align more clearly with the refined promise.
The next decision can be continue, narrow, adjacent test, return to prior version, or stop the test. State uncertainty. A mixed result may support another small evidence round instead of a bigger change.
Keep creator preference in the record. An angle can attract attention yet be undesirable or unsustainable to produce. The worksheet does not require a creator to pursue every observed interest.
Use a prewritten review question for each branch. Continue asks whether the existing promise remains clear and supportable. Narrow asks whether the refined wording concentrates repeated interest without losing the core audience job. Adjacent asks whether changing one angle produced a sufficiently different pattern to justify another test.
Record the same execution measures in both the baseline and test window. If cadence or quality changed, disclose the difference rather than attributing the whole result to the niche angle. When comparability breaks, the appropriate conclusion may be “run a cleaner test.”
Set a decision expiry date. Old evidence can inform history but should not automatically authorise a current change after creator capability, audience context, or product structure has shifted. Refresh the evidence window when the operating environment is materially different.
Limitations and sources
Limitations: the tree depends on the quality and comparability of creator-held evidence. It does not estimate market size, select a winning identity, or isolate every external influence. Decisions remain provisional and specific to the recorded window.
The evidence-first method adapts GOV.UK guidance to start by learning user needs. Term comparison caveats come from Google Trends Help. The Continue, Narrow, Adjacent test tree and evidence gate are original SirenCY editorial tools.