The decision
Undecided voters outnumber either major political camp in SPI’s six-city sample. SPI’s observed digital-advertising data also showed little political-advertiser activity at the time of analysis. That is a reason to test messages now—not evidence that attention is uncontested, that inventory is cheaper, or that any channel will produce a higher conversion rate.
Recommended sequence: Use September to test a small set of concrete messages with the intended audience. Scale only the versions that produce evidence of a stronger response.
Three facts that change the brief
- The undecided pool is the largest single group. In the linked six-city sample, respondents who had not expressed support for a candidate outnumbered either side’s existing support base.
- Political advertisers were rarely observed in SPI’s monitored digital-ad stream. This describes SPI’s observation window only; it is not a complete measure of voter attention, available advertising inventory, cost or campaign effectiveness.
- The main digital starting points are clear. Among undecided respondents with the relevant linked stream, 97% used Facebook; LINE and YouTube each exceeded four in five. Facebook is measured in the social-data stream, while LINE and YouTube are measured in the app-data stream, so the linked samples differ slightly. These are observed-use coverage figures—not guaranteed advertising reach.
What the evidence says to test
- Lead with a tangible household benefit. SPI tested 12 policy messages with more than 1,000 respondents. The strongest four concerned night-time childcare, healthcare-fee relief for young children, denture subsidies and youth employment. Everyday benefits generally ranked above broad vision statements in that tested set. This is a starting hypothesis, not validation of any campaign’s own copy.
- Write every proposition as beneficiary + benefit + proof. The leading messages named who would benefit, made the benefit concrete and included something that could be verified. Missing one of those elements was associated with lower relative performance within the test.
-
Move from platform choice to audience context. Facebook, LINE and YouTube are entry points, not the strategy. The next decision is which content environment, audience definition and time window deserves a controlled test:
- On YouTube, SPI’s same-sample viewing behaviour can distinguish the news and non-news channels actually used by undecided respondents. Any channel or creator partnership still requires an independent suitability review; observed viewing is not an endorsement.
- On Facebook and LINE, the linked sample can describe aggregate demographic and time-of-day differences between undecided voters and declared bases. In this sample, the undecided profile tilted toward women under 50, and almost 80% of observed Facebook use among undecided respondents occurred from midday through late evening. These are inputs to test design, not person-level targeting lists or performance guarantees.
- Within the LINE ecosystem, observed use can surface additional environments—including LINE-operated games—for a buyer to assess as test inventory. Coverage alone does not establish suitability or effectiveness.
A paid SPI briefing can provide the aggregate 15-channel map, audience-planning parameters and time-of-day configuration, with evidence limits attached to every recommendation.
New Taipei example: positioning around the incumbent
- Lee Shu-chuan (李四川): “Recognise existing municipal achievements and continue them, with specific improvements” produced the strongest directional response. Criticism need not be the opening frame.
- Su Chiao-hui (蘇巧慧): Criticising the incumbent offered only a limited opening; more than half showed no clear response to any Hou-focused frame. The main proposition should return quickly to her own everyday-policy offer.
Bottom line
The largest pool in SPI’s sample had not selected a candidate, and its linked behaviour identifies a practical set of digital environments for controlled message testing. The evidence does not establish that September media is cheaper, that attention is uncontested, or that one platform will convert better.
The commercially defensible decision is narrower and more useful: test a small number of specific, evidenced propositions among the intended aggregate audience before committing a larger budget. That replaces assumptions about “what undecided voters want” with a measurable comparison.