Where our data comes from
Tally360 is an independent provisional parallel tally. This page states exactly which data is genuine, historical, synthetic or still unverified.
Data status
- Counties, constituencies, wards, centres and station-streams (operational register)
IEBC register of voters per polling station, 2022 General Election (712-page PDF). 46,229 station-streams, 22,120,458 registered voters. Every record checked: unique 15-digit codes with leading zeros, each code consistent with its county/constituency/ward/centre, one name and one parent per code, no duplicate or rejected rows, no gaps in stream numbers. Cross-check: Changamwe's 152 stations total 93,561 voters, matching IEBC's 2022 Form 34B exactly.
Genuine, historical - Station locations (coordinates)
None verified yet. Stations without confirmed coordinates are still counted and searchable, and are shown as "Location unverified — not plotted".
Unverified - Boundary shapes on the map
2012 electoral boundaries from HDX; licence unresolved, so not shown on the published site.
Genuine, historical - Live election candidates
The 2027 candidate list has not been gazetted. The public portal shows an awaiting-setup state until it is configured.
Not available yet - Test election candidates and figures
Made-up names and numbers used only in the separate TEST ELECTION and the product demo pages.
Synthetic (test only) - Demo results centre geography below county level
The product demo page uses made-up constituencies, wards and stations. It never feeds the operational ledger.
Synthetic (test only) - Form 34B template
IEBC 2022 presidential Form 34B, Changamwe (001001), retrieved from IEBC's public forms portal. Used to design and test reading. Its last page carries agents' personal details, so it is linked, not hosted.
Genuine, historical - Form 34C template
IEBC's 2022 portal lists the national Form 34C but its link is broken. Real-document 34C reading is pending an authentic copy.
Not available yet - Form 34A template
Not yet retrieved. Real-document 34A reading accuracy is pending a labelled set of genuine photographed forms.
Not available yet
Boundary datasets
A map background from Google does not give permission to use separate boundary shapes
- Dataset
- County and constituency boundaries (2012 electoral boundaries)
- Source
- Humanitarian Data Exchange — Kenya elections (originally IEBC)
- Date
- 2012 (last modified 2018-08-16)
- Licence / permission
- Listed as "Other" — permitted use not confirmed
- Fit with the register
- Joined by official numeric code. 47 of 47 counties and 290 of 292 register constituency codes have a shape; diaspora (048) and prisons (049) have none; one constituency's parent county differs from the 2022 register.
- On the published site
- Hidden until permission is confirmed — results tables remain available
Official form references
- Form 34B · 2022 General Election — President · Changamwe constituency (001 001), Mombasa · Reg. 87(1)(a), 3 pages, scanned · retrieved 2026-10-24
IEBC copy · IEBC forms portal
Any practice forms used for development carry the watermark “SYNTHETIC TEST — NOT AN OFFICIAL RESULT”.
How a photographed form is counted automatically
Policy 2026-10-24.1
- The document is a Form 34A for this election
- The 15-digit station code is printed and fully legible
- If the sender chose a station, the code on the form matches it exactly
- Every candidate's figure and the rejected-ballots figure is legible
- Each figure was read with at least the minimum confidence
- The total valid votes printed on the form is legible and equals the sum of candidate figures
- No signs of alteration, cancellation or overwriting were noticed
- Each figure read with at least 95% certainty
- The station is in the election’s register, votes cast do not exceed registered voters, and no different copy for that station has already been counted
If any check fails, the form goes to a person on the review desk, and whatever couldn’t be read stays marked as unknown — it is never guessed. The original is kept privately as evidence.
Pending: needs a labelled set of real photographed 34A forms to measure error rates at this threshold.