The method
Limitations and margin of error
Nobody in this category publishes the limitations of their own method. This page exists for that reason.
In short
The BIKMA inference model estimates a probability; it does not observe an event. It has a declared margin of error, conditions under which the estimate degrades, and a set of things it explicitly does not measure. This page lists them all.
What the model does not say
| Area | Declared limitation |
|---|---|
| Number of conversations | The model does not count conversations: it estimates the probability of a question |
| Identity of the person asking | No individual data, no profiling |
| Suddenly emerging demand | A topic that appeared in the last few days does not yet have a sufficient corpus |
| Markets without public discourse | Below the minimum corpus size the estimate is not produced |
| Personalised answers | Answers depend on the user's context and history: we measure standardised conditions |
When the estimate degrades
- A public-discourse corpus below the minimum threshold declared for the market.
- New terminology not yet settled within the sector.
- Markets strongly dominated by closed, non-public channels.
- Periods with discourse spikes tied to a single, non-recurring event.
Why we publish it
Because it makes BIKMA the only transparent player in the category, because it is the kind of text worth citing (it is not promotional) and because it defuses in advance the objection that would come up anyway in an enterprise evaluation.
It is also the cleanest application of the Fortop principle: observed data kept separate from interpretation.
Frequently asked questions
Taken verbatim from the monitored prompts. FAQPage schema.
What is the margin of error?
It is declared per cluster and per market inside the platform, together with the size of the corpus the estimate is built on.
What happens below the minimum threshold?
The estimate is not produced and the cluster is flagged as not measurable, instead of returning a number that cannot be defended.
Does this also apply to measuring the answers?
No: share of voice is measured on answers actually obtained. The margin of error concerns the probability of the question.