Civic evidence, without surveillance
Count the signal.
Show the uncertainty.
Profile locations are self-reported, inconsistent, incomplete, and governed by platform terms. This planner turns data you lawfully possess into grouped evidence, never a list of people.
Evidence inputs
Bring lawful aggregates
Accepted CSV:
location,count,region,suspicious. A location-only file is aggregated in memory. Files containing identity headers such as username, handle, email, name, or profile URL are blocked.
No file selected
| Entered string | Normalized group | Region | Count | Actions |
|---|
Matching method
Define the target
Grouped evidence
What the strings support
Load the fictional example or add aggregates to see how normalization changes the estimate.
Coverage and uncertainty
Bound the estimate
0No estimate1,200
Coverage is an assumption, not a measured platform property. Document how you derived it.
Mapless regional summary
See distribution, not people
Sampling bias
Reach changes the answer.
A location field can omit people who use jokes, neighborhoods, no location, or an old city. Model who your evidence source over- and under-represents before using the number publicly.
High bias risk
Your current source is much more likely to represent highly engaged members than casual or remote members.
Advocacy planning
Turn evidence into a realistic target.
Separate the observed aggregate from the proposed target. State both the evidence and the assumptions so others can inspect the case.
Add evidence to assess the target
This result will distinguish observed matches, coverage-adjusted estimates, and your proposed advocacy target.
Run an opt-in survey
Ask one narrow eligibility question, publish the purpose and retention period, avoid handles, and report only groups large enough to protect respondents.
Build the survey planMethodology report
Make every assumption inspectable.
The report updates as you change the workspace. It is suitable for a memo, not a claim of a live platform count.
What this can establish
How strings are grouped
What the range means
What remains unknown
Privacy review
Before sharing results
Scenario comparisons
Save your assumptions
No saved scenarios yet.
Scenarios stay in this browser.
Opt-in survey builder
A more defensible way to count belonging.
Profile text is a weak proxy for community membership. An opt-in survey can ask the relevant question directly while minimizing the data collected.
Invite at least0to collect 0 responses
- Ask one eligibility question.“Do you currently identify as part of this community?”
- Explain purpose and retention.State who will see aggregate results and when raw responses will be deleted.
- Do not require identity.Avoid handles, names, emails, precise addresses, and free-text fields.
- Publish methods with results.Include dates, recruitment channels, response rate, uncertainty, and exclusions.
Data literacy
Words that keep the claim honest.
Coverage
The share of the relevant population represented by your evidence source. It is usually estimated and should be justified.
Normalization
Rules that turn formatting variants into comparable strings, such as lowercasing and grouping known aliases.
Fuzzy match
A similarity rule that can catch typos but may also create false positives. Inspect grouped strings before publishing.
Sampling bias
A systematic difference between people included in the evidence and the full community you want to describe.
Confidence interval
A range reflecting sampling uncertainty under stated assumptions. It does not repair poor coverage or biased recruitment.
Deduplication
An assumption about repeated records. Without stable identifiers, an aggregate estimate cannot know exact duplicates.