Last reviewed · 8 September 2026
The Boxumer Barometer: how we intend to build public trust signals
This page describes a method, not a product. No Boxumer Barometer figure exists today. We are publishing the rules first, so that the first report can be judged against a standard written before the data arrived.
Current status: nothing is published
Boxumer today is a personal tool. You connect your inbox, your purchases are found, and you rate your own experiences privately. There is no public brand score, no aggregated trend, and no ranking anywhere on this site.
We are writing the method down in advance for one reason: a reputation system that only explains itself after publishing numbers has already lost the argument.
What would count as an input
Only an experience attached to a real purchase, rated by the person who made it.
- The purchase is identified from the buyer's own order confirmation, not typed in freely.
- One experience per order, rated by the account that owns that order.
- The brand is resolved to a single merchant identity, so subsidiaries and storefronts do not split or merge silently.
- The date used is the purchase date, not the rating date, so signals map to when the experience happened.
Inclusion and exclusion rules
A signal is only as honest as what it refuses to count.
- Excluded: brands below the minimum number of independent consumers for the period.
- Excluded: experiences from accounts showing automated or duplicated behaviour.
- Excluded: any experience solicited, incentivised or paid for by a brand. Boxumer does not sell placement and would not include a brand-triggered request.
- Capped: the weight of a single consumer, so one heavy buyer cannot define a brand's period.
- Reported separately: periods where the sample changed enough that a comparison with the previous period is unsafe.
How a signal would be calculated
The intended output is not a star rating. It is a distribution and a direction over a defined window.
For a given brand and period we would publish the distribution of rated experiences, the number of independent consumers behind it, and the change relative to the previous comparable period, alongside an interval that reflects how small the sample is.
Where the sample cannot support a directional claim, the correct output is 'not enough data', and that is what would be shown.
Limits we would publish alongside every figure
Every dataset has a shape, and ours would have a specific one.
- Sample size and the number of independent consumers, always visible.
- The exact period analysed, with the last update date.
- Coverage bias: Boxumer users are not a random sample of a brand's customers.
- Channel bias: purchases traceable by email confirmation are over-represented.
- Country and language distribution of the sample.
- Uncertainty: the smaller the sample, the wider the range and the weaker the claim.
Privacy rules
Individual ratings stay private. A published signal would only ever be an aggregate, and only above a threshold high enough that no single person's experience can be identified from it.
No merchant name, order content or personal detail belonging to one user would appear in a public output, and brands would not be able to buy access to individual experiences.
When the first report would be published
When the data supports defensible conclusions, and not before. That means a brand-and-period sample large enough to survive the limits above, and a method that has been stable long enough to compare two periods honestly.
If that takes longer than we would like, the report is late. It is not invented.
Frequently asked questions
Does the Boxumer Barometer exist today?+
No. This page describes an intended methodology. No public score, trend or brand ranking is published by Boxumer at this time.
Why publish the method before the data?+
Because a method written after the results can be shaped to fit them. Publishing the rules first makes the first report checkable against a standard that existed before it.
Can a brand pay to change or remove a signal?+
No. Boxumer does not sell placement, and brand-solicited experiences would be excluded from any aggregate by design.
Would individual ratings ever be public?+
No. Ratings stay private to the person who made them. Only aggregates above a minimum threshold could be published.
What happens when there is not enough data for a brand?+
Nothing is published for that brand and period. 'Not enough data' is a valid and expected output.
The Boxumer Journal
If this article spoke to you, you'll like the rest of the series.
One short piece a month on review reliability, consumer trust and how to read a brand before buying. No spam, no upsell — just the work.
Free to read. Free to leave. We never sell your address.
Curious how a verified-purchase signal actually works in practice?
Signals, not averages
The personal layer exists today
Boxumer finds your past purchases from your inbox and lets you rate them privately. The public layer described here is the direction, not a current feature.
Keep reading
Recent reviews or average rating?
Why a cumulative score hides what is happening to customers right now.
What a verified purchase review guarantees
Verification proves an order existed. It does not prove the opinion is representative.
Proof-based reputation
What a reputation system built on evidence rather than opinion would look like.