Where do my donors cluster?

You are a fundraising analyst. Tell me where my donors physically cluster in Australia. Produce two outputs in this directory:

• answer.md — state-level breakdown, top-10 by count vs top-10 by rate, donor-belt and donor-desert descriptions, and an operational “what to do with this” section keyed to the three buckets (both-lists / count-only / rate-only).
• answer.html — two interactive maps side-by-side: top by donor count vs top by donor rate per 1,000 adults. The whole point of this example is the lists aren’t the same. The page also carries a small “What to do” actions card so a non-analyst viewer can read the strategic takeaway without leaving the map.

Inputs

• sample_donors.csv — donor file (postcode, gift_amount, frequency, …).

Data sources (no auth required)

POA boundary geometry — ABS ArcGIS REST:

https://geo.abs.gov.au/arcgis/rest/services/ASGS2021/POA/FeatureServer/0/query?where=poa_code_2021%20IN%20(<list>)&outFields=poa_code_2021,poa_name_2021&outSR=4326&f=geojson

Adult population per POA (the denominator for the rate calculation) — ABS Census 2021 G01 via the Data API:

https://data.api.abs.gov.au/rest/data/ABS,C21_G01_POA,1.0.0/...?format=jsondata

If SDMX is unmanageable, fall back to general knowledge of postcode population sizes and disclose at the top of answer.md.

The analysis (write to answer.md)

1. State-level breakdown — % of donors per state vs each state’s AU population share. Over/under-index?

2. Capital city vs regional — for each state, % metro vs % rest. Compare against ABS urban/regional split.

3. Top 10 postcodes by donor COUNT. Name the suburb(s). Flag the cohort (high-income / mid-income / wealth-belt / regional / inner-urban / outer-suburb).

4. Top 10 postcodes by donor RATE (donors per 1,000 adult population). Compare to the count list — if the lists are different, name the story (e.g. “2042 Newtown appears on rate but not count — lower-population postcode, high per-capita traction”).

5. The donor belt — describe high-density postcodes as a single geographic story (e.g. “eastern Sydney waterfront + Mosman / Cremorne on the north shore + an inner-city renter stripe”). If there isn’t one cohesive belt, say so honestly.

6. The donor desert — same for postcodes where I have zero or near-zero supporters.

7. One chart to commission based on what you found. Tell me what to ask the design team for.

8. What to do with this — three operational plays, one per bucket. The whole point of the count-vs-rate split is that each bucket implies a different move. Don’t give generic advice — name the specific postcodes and the specific moves.

   • Both-lists postcodes (your true core): protect retention. Don’t change what’s working. Flag any misfit inside the bucket (e.g. one renter postcode sitting in an otherwise wealth-belt list) for separate stewardship treatment, because the same plan won’t fit both.
   • Count-only postcodes (volume, average rate): test differentiated creative. Big-population postcodes where rate is average usually means the message is tuned to a different cohort. Name an A/B test (“test creative variant X in 2010 / 2031 against the rate-only benchmark of 2042”).
   • Rate-only postcodes (high traction, small population): don’t try to grow volume — these postcodes are capped on size. Two moves: (a) lookalike acquisition in similar postcodes you haven’t acquired in yet (name the candidates), and (b) lift average gift in the warm donors you already have.
   • Plus, name any single postcode where the count/rate delta is sharp enough to warrant a specific reallocation (e.g. a high-rate, mid-count postcode that’s under-resourced under a count-led plan).
   • Out-of-scope: the donor desert is its own strategic question (geographic expansion), not something the count/rate delta can answer — flag it as a separate decision, don’t merge it into the bucket plays.

After the chart recommendation and the actions section, add a short “How to read this” block (~80 words) aimed at someone seeing this analysis for the first time. Explain:

• The left map ranks postcodes by raw donor count — volume.
• The right map ranks by donors per 1,000 adults — density per capita.
• Purple-dashed postcodes appear in both top-10 lists.
• The story is the delta between the panels: count-only postcodes are where you have volume, rate-only postcodes are where your brand has unusually strong traction relative to who lives there.

Length: 600-800 words plus the explainer (raised from 500-700 to accommodate the actions section — don’t pad to hit the cap).

The map (write to answer.html)

Two side-by-side MapLibre maps in a CSS Grid, each ~50% of the viewport width, sharing the same legend and basemap. The point of the layout is the visual delta between count and rate.

Required spec

• Library: MapLibre GL JS via CDN. Two Map instances, one per panel.
• Basemap (both panels): CARTO light_all raster, no key.
• Layout: CSS Grid, two columns, each panel 50% width, 100vh tall. Each panel has a small h2 title.
  • Left panel: “Top 10 by donor count”.
  • Right panel: “Top 10 by donor rate per 1,000 adults”.
• Geometry: fetch the polygons for both top-10 lists (union of postcode IDs) from ABS ArcGIS POA. Embed inline.
• Layer per panel:
  • fill — colour the panel’s top-10 postcodes by their relevant metric (count or rate). Use a sequential ramp:
    • Count panel: blues #deebf7 → #9ecae1 → #3182bd (3-step quantile).
    • Rate panel: oranges #fee6ce → #fdae6b → #e6550d (3-step quantile).
  • line — postcode outlines #1a1a1a width 0.5, opacity 0.4.
  • Postcodes that appear on both lists get a distinctive dashed purple outline (#7b2cbf, dash-array [4, 3], width 2.0) drawn over the fill in both panels. This makes “same postcode in both” visually unambiguous.
• Initial view: sync both maps to the union bbox of all top-10 postcodes from both lists. Add map.on('move', ...) so panning or zooming one panel mirrors the other.
• Hover popup showing postcode, suburb, count rank, rate rank, donor count, donors per 1,000 adults.
• Shared legend at the bottom-spanning-both-panels:
  • Count colour ramp (3 swatches) with the breakpoint values.
  • Rate colour ramp (3 swatches) with the breakpoint values.
  • Dashed purple swatch labelled “On both lists”.
• Title bar across the top: “Donor clusters — count vs rate”. Subtitle: “Same data, two denominators. The delta is the story.”
• Explainer panel between the two map panels (or above the shared legend on narrow viewports) — small card titled “How to read this map”:

  Left = top postcodes by raw donor count (volume). Right = top by donors per 1,000 adults (density per capita). Purple-dashed outlines mark postcodes on both lists. The delta between the panels — postcodes that appear in only one — is the story. ~320px wide, semi-transparent white, subtle shadow, dismissible.

• Actions card — a second small card titled “What to do”, also ~320px and dismissible, sitting alongside (or below) the explainer. Three rows keyed to the legend colours:
  • Both lists (purple dashed) — list the specific postcodes and one short play (e.g. “protect retention; flag misfits”).
  • Count-only (blue) — list the specific postcodes and one short play (e.g. “A/B test differentiated creative against rate-only benchmarks”).
  • Rate-only (orange) — list the specific postcodes and one short play (e.g. “lookalike acquisition in similar postcodes; lift average gift in warm donors”). Optionally append one line for the sharpest single-postcode call (e.g. “2028 Double Bay: #7 count, #1 rate — rebalance major-gift time here”). Be terse — one line per row, no preamble.

Constraints

• Self-contained HTML, no build step.
• File size <600 KB.
• Works in Safari and Chrome on macOS.
• Panel maps must stay in sync when panned / zoomed.

When you’re done

After both files are written, run open answer.html so the map opens in the user’s default browser straight away.

Rules

• No web tools beyond the ABS / CARTO URLs above.
• The story is the delta. If the count list and rate list are identical, say so explicitly — that’s also a finding.
• Describe geography in words. “Eastern Sydney waterfront from Bondi to Watsons Bay” beats “postcodes 2026, 2030, 2027”.
• Be terse. 600-800 words in answer.md.
• Actions must be concrete. Name the postcodes, name the move. “A/B test creative in 2010 / 2031 against 2042 Newtown as benchmark” beats “consider audience segmentation”.

Disclaimer

Mapulus provides this prompt as educational starter content. The analysis, predictions, and recommendations produced by running it come from a third-party LLM operating on your data — Mapulus is not responsible for the accuracy, completeness, or fitness-for-purpose of any output. Treat the results as hypotheses worth testing, not advice worth acting on without your own validation.