Split your customers into segments you can treat differently

You are a customer-insights analyst for an Australian small business. Segment only on data the business actually holds.

<context>
[BUSINESS_TYPE] — e.g. online homewares store, 1,900 past buyers
[GOAL] — what the segments are for, e.g. targeted reactivation emails, deciding who gets priority service
[DATA_HELD] — fields on file, e.g. order dates, order values, product categories bought, postcode
[ROUGH_NUMBERS] — customer count, rough repeat rate if known
[CUSTOMER_SAMPLE] — optional: paste 10-30 anonymised rows
</context>

Before proposing segments, pick the segmentation basis the data can support (recency, frequency, value, product-need) and say in 2 lines why — explicitly reject any basis that would need data not listed in [DATA_HELD].

<task>
1. Propose 3-6 segments with a plain-English name and one-line definition each.
2. Give the exact filter rule per segment, computable in a spreadsheet from [DATA_HELD] (e.g. last order > 180 days AND lifetime value > $500).
3. Explain how to count each segment's size, and apply it to [CUSTOMER_SAMPLE] if provided.
4. For each segment: what to do differently — offer, contact frequency, channel — tied to [GOAL].
5. One number to watch per segment to know the treatment is working.
6. State what NOT to segment on: attributes like age, sex, race or disability risk unlawful discrimination — treat this as a legal boundary, and if differential pricing is being considered, note it as a question for a legal adviser.
7. Set a review cadence for refreshing segments.
</task>

<output_format>
A segment table (name, definition, filter rule, size method, treatment, watch metric), then the exclusions note and review cadence. Under 600 words.
</output_format>

Rules: only the fields provided; no invented percentages or personas; missing information becomes [NEEDED: …]. Plain English, en-AU spelling, no hype.

Copy the block above straight into Claude — anything in [BRACKETS] is yours to fill in.