On this page

A useful comparison needs a trail that another reader can follow: which firms you selected, which record you counted, the dates you used and the calculation you made. Saving two percentages without those details leaves too much to guess.

Here is a small comparison worksheet you can reuse. It is an editorial method for organising research, not a rating system or a recommendation to choose a particular bank.

Choose one measure for both firms

Start by deciding what you want to compare. In this exercise, use outcomes in a defined set of published final Ombudsman decisions. Do not put one firm's internal complaint percentage in the neighbouring column.

The FCA expressly distinguishes firm-reported complaints from complaints referred to the Ombudsman and says the datasets should not be compared directly. FCA explanation

Use Bank Conduct Index's comparison page as a place to begin, then check its published method. Record the precise legal firm names shown, rather than assuming that a familiar banking brand identifies one company.

Freeze the comparison settings

Write down these settings before copying either firm's result:

SettingYour research note
Firm identityExact displayed firm name and record URL
PopulationPublished final decisions, with the relevant sector recorded
PeriodStart and end dates, including any uncertainty about date boundaries
Outcome definitionWhat counts in the numerator and denominator
CoverageMissing, unclassified or excluded records identified by the source
ObservationDate you retrieved the figures

The index's method describes a banking, credit and mortgages headline scope and distinguishes outcomes recorded from the Ombudsman's classification from its separate identification of partial upholds. Read those details before choosing a calculation; do not invent a treatment for partial or missing outcomes. Index methodology

Recalculate a small example

Fictional decision set A has 8 upheld out of 40, or 20%. Set B has 30 out of 100, or 30%. Their difference is 10 percentage points. Bars share a 0 to 100 percent scale.
Two invented decision sets, with counts retained and both bars drawn on the same scale. The comparison is arithmetic, not a ranking of real firms or a prediction. Awesome Websites; original explanatory graphic · Original site graphic Enlarge image

The following figures are entirely fictional. Assume the two sets already have the same documented scope and outcome definition.

Fictional setDecisions in the denominatorUpheld decisionsCalculationResult
A4088 ÷ 40 × 10020%
B1003030 ÷ 100 × 10030%

The difference is 10 percentage points. This table describes those two invented sets. It does not establish why their outcomes differ, how either firm handles all complaints, or what would happen in another case.

For real figures, retain the counts alongside the percentages. If the page only supplies a rounded percentage, record that limitation instead of reverse-engineering an exact numerator.

Make the result recoverable

Save the source URLs, the date of access and your completed worksheet. If a source offers an export, keep the original export separately from your calculations.

When you revisit the comparison, make a new dated observation. A changed total should prompt a check of the source and filters before it becomes a claim about a firm's performance. Your finished comparison should make it possible to reproduce the arithmetic and see what the arithmetic cannot answer.