Sample expected output — spot the bias

Shape, not exact words.

The overlap

Your donor file covers ~17 postcodes; campaign list covers 16.
~12 postcodes appear on both lists.
5 donor postcodes were never on a campaign: 2011 (Potts Point), 2032, 2037 (Glebe), 2042 (Newtown), 2150 (Parramatta). These people found you despite no targeted outreach — don’t ignore that.
The reverse is also true: 4 heavily-targeted postcodes (2095 Manly, 2100 Brookvale, 2110, 2034) generated thin returns relative to spend.

Look-alike profile

Across your donor postcodes:

Median household income ~$2,180/week (vs $1,750 AU)
~18% aged 65+ (vs 16% AU)
~52% owner-occupier — bimodal (high in east, low in inner-city)
~14% volunteering (vs 11.8% AU)
ALP TPP: ~57% — leftward but with a split by sub-cluster

Unasked look-alikes — 10 candidates

Postcode	Suburb	Why it’s a credible candidate
2041	Balmain / Birchgrove	Inner-west renter, civic, similar to your inner-city donors
2040	Leichhardt / Lilyfield	Owner + renter mix, similar profile to 2041
2204	Marrickville	High volunteering, civic-engaged, ALP — never asked
2049	Lewisham / Petersham	Same as above, denser
2065	Crows Nest / Naremburn	Owner-occupier mid-density, similar to Mosman
2113	Macquarie Park / North Ryde	Higher-income, older — matches wealth cohort
2154	Castle Hill	Owner-occupier suburban, mid-to-older, mid-income
2074	Pymble / West Pymble	North shore wealth — very close demographically to 2030 / 2088
2120	Pennant Hills / Thornleigh	Stable suburban, owner-occupier, mid-to-older
6011	Cottesloe (WA)	Closest national look-alike to 2030

Bias warning

Three groups are systematically absent from your file:

Multicultural inner-west / outer-south Sydney (Bankstown 2200, Cabramatta 2166, Auburn 2144). Never asked, zero supporters. The model can’t tell you whether they’d give — but the file can’t tell you either.
Middle-suburb owner-occupier families (Parramatta 2150, Liverpool 2170). High volunteering and unpaid care but light-touch campaign coverage.
Regional centres (Wollongong, Newcastle, Geelong, Bendigo). Zero coverage; the look-alike score for these is moderate-to-high on the same demographic axes as your eastern-suburbs cohort.

Cheapest experiment

Pilot postcode 2204 (Marrickville) for three months.

Why:

Demographically very similar to your inner-city young donor cohort (Newtown 2042, Surry Hills 2010).
Never targeted — true unasked-cohort test.
High volunteering (~16%), high Y12 completion.
Downside is one mailing cost.

Pilot design:

1 mailing + 1 email + 1 paid-social retargeting tranche.
Measure response rate vs your historical control (inner-city Sydney average).
3-month observation window.
Total cost: ~$3-5k depending on volume.

Honest framing: this is a list of hypotheses, not a media plan. The model can’t tell us these people will give — only that they look like the people who already do. A pilot is the only way to find out.