The app-store rating barely separates one neobank from another

Comdirect scores 1.30 on Trustpilot and 4.46 in the app stores. Same bank, same customers, same year - a gap of more than three stars.

It is tempting to turn that into a rule: app-store ratings are inflated, Trustpilot tells the truth. The data does not support the rule, and what it supports instead is more useful.

We compared both sources for every neobank in our directory with at least 100 reviews on each side - 50 companies, 2.1 million Trustpilot reviews and 45 million app-store ratings.

The app-store rating barely separates anyone

Start with the spread, because it is the finding everything else hangs on:

Trustpilot uses the whole scale. The app-store score sits inside a third of a star. 23 of the 50 app-store scores fall between 4.0 and 4.6 - a band six-tenths of a star wide holds nearly half the market. On Trustpilot the same band holds 21, and they are scattered among scores that run from 1.3 to 4.9.

A rating that assigns almost everyone the same number cannot tell you who is better. If you are choosing between two neobanks and both show 4.5 in the App Store, you have learned nothing. The comparison that looks most authoritative - millions of ratings, prominently displayed - is the one carrying the least information.

The gap is real, and it grows as Trustpilot falls

Across the 50, app stores score 1.08 stars higher on average, with a median gap of 0.73. But the gap is not spread evenly. Group the banks by their Trustpilot score and it appears almost entirely at the bottom:

Read the last column downwards. The Trustpilot score falls from the top of the scale to the bottom; the app-store score barely moves off 4.4. A bank its customers savage on Trustpilot and a bank they praise look almost identical in the App Store.

Where the two sources disagree, and why

The widest gaps in our set:

The mechanism is not mysterious, though we can only infer it rather than prove it from this data. The two ratings are collected in opposite circumstances.

App ratings are prompted. The dialog appears inside the app, usually after something has gone right - a transfer landed, a payment cleared. The people asked are, by construction, people currently succeeding at using the product. Volume follows: across the 50 companies there are 21 app-store ratings for every Trustpilot review.

Trustpilot is sought out. Nobody opens a review site because their transfer worked. They open it after a frozen account, a failed transfer, a support queue that went nowhere. It measures a different thing: not whether the product works when it works, but what happens when it does not.

Neither is dishonest. They answer different questions, and the industry's habit of quoting whichever is higher obscures that.

A pattern in the reversals, with a caveat

Six of the 50 run the other way, rated higher on Trustpilot than in the app stores. Four of those six carry a business, small-business or freelancer tag - against roughly a third of the rest. Those are the accounts where a customer is likelier to write a considered review than to rate an app, so the direction makes sense. But six is a small number, and we would rather flag the pattern than build a theory on it.

How to actually use the two numbers

Use the app-store score for the app. It is a decent read on whether the thing is pleasant to use day to day, and a genuinely bad app does show up - the floor in our set is 3.23.

Use Trustpilot for the failure case. It is where account freezes, KYC deadlocks and support collapses surface first.

Treat a wide gap as information, not as noise. A company at 1.4 on Trustpilot and 4.4 in the App Store is not a contradiction to resolve. It usually describes a product that works smoothly until it does not, and a support function that struggles when it does.

That is why we publish both on every profile rather than averaging them into a single score. Averages hide exactly the disagreement worth seeing. The full directory - every neobank, both sources, per-country breakdowns - is at our neobank directory.