The canvassing app marked every unanswered door as opposition
PoliticsAfter my first canvassing shift, the dashboard showed forty-two “opponents” on my route. I had spoken to only eleven people. The rest had not answered, but the app converted an empty response into “leans against.”
The field manager said the campaign’s vendor used nonresponse as a prediction so volunteers could prioritize future doors. On the volunteer screen, prediction and conversation result shared the same label.
I asked the team to separate “not reached” from any modeled leaning. The campaign corrected my route, removed predicted labels from canvasser notes, and kept modeling only in aggregate planning reports.
Silence may be useful to a statistical model. At the doorstep, it is not a statement a resident made.