The first polling tool that labels its sample frame, flags mode effects, redacts small cells, and weights on ACS demographics — before you publish.
Email + SMS live today · phone IVR in beta · USPS mail coming H3
If a journalist could ask you "how did you sample?", this is for you.
City and county engagement offices running resident polls on budgets, services, and planning.
501(c)(3)s and community orgs doing constituent research that has to survive external review.
Parent, student, and staff climate surveys with weighting that reflects who actually enrolls.
Program evaluation and patron feedback, with methodology transparent enough to publish.
Per-unit, pass-through pricing. No bundles. No surprise bills.
Fast, cheap, and high-deliverability. Best for residents who have opted in to municipal notifications.
Higher response rates than email for short polls. Reaches residents without desktop access.
Bilingual interactive voice response. Twilio Voice ($0.025/min) + AWS Polly Neural TTS ($0.060/min) at cost, with a transparent Rally markup. Reaches residents who don't use email or SMS.
First-class letter via Lob (black-and-white, single-page, domestic US). Critical for reaching households without reliable internet. Not yet live — shipping H3.
Most polling tools let you publish a number with no context — no sample frame, no redaction, no weighting, no trace of what you did. Rally won't let you do that. Every poll you publish includes the full recipe.
Every poll is stamped with the frame you actually sampled from — probability, convenience, panel, voter file. On a convenience sample, an immovable banner tells readers the results cannot be generalized to the full population. You can't remove it. No one can.
lib/methodology-statement.js · routes/polls.js
Crosstabs with fewer than 5 respondents in any cell are redacted. This protects residents from re-identification — especially in small geographies where a demographic slice can uniquely identify a person.
routes/poll-crosstab.js · routes/poll-crosstab-preview.js
If more than 60% of your responses arrived through one channel, the poll shows a warning; above 80%, an alert. Channel mix matters: email-only residents look different from phone-only residents, and we won't let you hide that.
lib/mode-effect.js
Raking/IPF weighting against American Community Survey 2018–2022 five-year defaults, with a Pew-style 5×-mean cap to keep weights from blowing up on a thin cell. We flag significant divergence between raw and weighted results so you see the correction, not hide it.
lib/weighting.js · lib/poll-analytics.js
When you publish a poll, we capture an immutable snapshot of the sample frame,
target demographics, field dates, channel plan, and weighting inputs — SHA-256
hashed, stored with REVOKE UPDATE at the database level. You can
verify it months later. So can anyone else.
lib/poll-methodology-archive.js · migration 029
Every public poll page carries a methodology statement written in plain English: sample frame, field dates, response count, weighting, and what the number does and doesn't mean. Not buried. Not behind a PDF. On the same page as the result.
lib/methodology-statement.js
Civic nonprofits get the extended free tier: 500 responses a month with the full methodology stack — frame labels, redaction, mode-effect warnings, weighting, methodology snapshot. We verify 501(c)(3) status via ProPublica Nonprofit Explorer.
Apply for the nonprofit tierWe review applications within 48 hours. Approved orgs keep the tier as long as they're in good standing on the ProPublica record.
Read the full Trust page for our security posture, data handling, and what we do and don't claim about certifications.