Election Reporting Methodology
Election and year
Every report belongs to a polling unit, contest and cycle. Presidential 2027 is the default. Changing election or year changes the relevant reporting population and candidate ballot.
Expected reports
One expected report is one eligible polling-unit and contest pair. Presidential coverage is national; governorship coverage is within the state. Other contests require confirmed geography. A candidate's named seat remains visible while boundaries await confirmation. Unknown coverage is explicit, not an invented denominator.
Completeness and verification
Submitted counts use the latest submitted revision per pair. Revisions do not add expected reports. Remaining equals expected minus submitted; completion equals submitted divided by expected. Zero or unknown denominators do not yield a meaningful percentage. Vote totals use the current verified report, not pending reports.
Sheets and ballot accounting
Evidence-sheet counts differ from report counts because one report may have several sheets. Accounting includes registered and accredited voters, issued, unused, spoiled and rejected papers, valid votes and total used papers. Missing candidate votes are not declared zeroes.
Publication
Private sheets require separate publication review. These are council reporting records, not official electoral declarations. Read the Election & Publication Disclaimer.
Geographic aggregation
State, LGA and ward figures are calculated within the selected election and cycle. A national presidential view should not be compared directly with a single senatorial contest as though their expected-report populations were equal. A selection changes the population under review, and all displayed completeness figures must be interpreted within that selection. A zero-unit source ward remains part of the source hierarchy rather than being silently deleted.
Candidate totals and vote share
Candidate votes are associated with ballot entries in a contest, not joined solely by a party acronym. The same party can contest several different seats. Charts that combine parties across a selection describe the selected reporting set, not a contest between candidates in unrelated seats. A candidate's reported vote share uses the available verified vote total in the relevant selection. It is not a prediction of unreported votes or an official declaration of victory.
Revisions and duplicate protection
The submission workflow identifies each report by contest, polling unit and cycle through its contest record. Replaying the same request key returns the existing submission; it does not create another report. A correction must explicitly identify the latest submission it replaces. Review history is retained, and a pending revision does not overwrite the currently verified version. Identical figures and evidence do not justify a new correction merely to increase submission activity.
Time and trends
Trend displays use the recorded review timestamps of current verified reports. They describe the timing of the reporting and review process, not when votes were cast. Corrections can change the current reporting series. Publication dates, event dates and review dates serve different purposes and should not be substituted for one another. Nigeria time is used where a calendar day is shown for operational reporting.
Evidence, limits and corrections
A report may contain more than one evidence sheet, so sheet counts can exceed distinct report counts. A sheet can remain private even when aggregate figures are public. Internal verification means that an authorised reviewer accepted the platform submission under the configured workflow. It does not authenticate every external claim about the election. For a suspected error, provide the contest, polling-unit code, year and supporting source through the appropriate review channel. Do not submit a second unrelated record to conceal or replace an earlier one.