Rupert Lowe’s Restore Britain by 21 Scottish Seats in polling error.
A major multi-level regression and post-stratification (MRP) poll released this week briefly handed Restore Britain an eye-catching result in Scotland — until a calculation error was spotted and fixed within hours.
Restore Britain founder Rupert Lowe highlighted figures from a FindOutNow/Electoral Calculus survey that modelled a tactical-voting scenario. In that version of the data, his party came out ahead of both Reform UK and the Conservatives in 21 Scottish seats. Lowe described the outcome as evidence that Restore was now the largest force on the right in Scotland.
The numbers looked striking.
They also turned out to be wrong.
Analysts examining the underlying tables noticed that the tactical-voting model had adjusted other parties’ shares but left Restore Britain’s constituency-level percentages untouched. In effect, when supporters of Reform, the Conservatives or others said they would switch, their votes were redistributed. Restore’s votes were not. This created a relative advantage that did not reflect the poll’s own respondent data.
Electoral Calculus confirmed the inconsistency after it was raised. Corrected figures were published the same day.
Under the revised tactical model, Restore Britain dropped to zero seats in the “largest on the right” comparison in Scotland. Reform’s position improved substantially, rising to first among the three parties in 44 seats, while the Conservatives held 13. National seat projections also shifted: Reform’s total rose and Labour’s fell.
Restore’s overall national share in the tactical scenario was revised downward from 8% to 6%.
The original, uncorrected numbers had already been widely shared. Lowe’s posts claiming a breakthrough for Restore circulated on X and elsewhere before the update appeared.
Complex MRP polls involve layers of modelling and assumptions. Errors can slip through. What this episode shows is how quickly an anomalous result can travel when it confirms a political narrative — and how important independent scrutiny remains. In this case the pollster corrected the record promptly once the problem was identified. The initial claim, however, had already done the rounds.
