Sorting your topics by verbatim count means prioritising what gets talked about the most, not what costs the most. The Business Impact score combines several sentiment and volume signals to give each topic a score from 0 to 10: the higher the score, the higher the priority.

Step 1 — Each signal is scored out of 100

Five signals can feed the score. Each one is brought onto a single scale where 0 = no issue and 100 = maximum priority.

Signal	How it is scored
Volume	Share of the topic's verbatims out of the period total
Sentiment	`((100 − sentiment) / 100)^1.7 × 100`
NPS	`% detractors × 100 + % passives × 30`
CES	`(average CES / 10) × 100`
Ratings	`((10 − average rating) / 10) × 100`

Every signal points the same way: a high score always flags a problem, never a success.

How promoters, passives and detractors are treated

The NPS signal only counts what is going wrong:

a detractor counts as 100;
a passive counts as 30, roughly 3.3 times less than a detractor;
a promoter does not count at all.

Passives are therefore not ignored, but they weigh considerably less than a detractor.

Why sentiment uses a curve

The 1.7 exponent applied to sentiment is not neutral: it makes a severely negative topic weigh far more than a merely lukewarm one. Two moderately negative topics will not add up to the weight of one genuinely critical topic.

Step 2 — Each signal is weighted

Signals do not carry the same weight. These are the base weights:

Signal	Base weight
Sentiment	45
NPS	35
CES	30
Ratings	30
Volume	25

Three rules apply:

Only signals actually present on the topic are counted. If a survey does not collect CES, the CES signal stays out of the calculation — it is not replaced by a neutral value.
The weights of the retained signals are rescaled to total 100%. Missing signals therefore do not "dilute" the score: their weight is redistributed across the others.
Volume is always included.

These weights are fixed. They do not vary with the number of responses, the size of the survey or the topic being analysed.

Step 3 — Everything is brought back onto a 0–10 scale

Business Impact = ( Σ signal score × its weight ) ÷ 10

This is the weighted average of the scores, divided by 10.

The result is a figure that is comparable across topics, across entities and across surveys: a topic at 7.8 comes before a topic at 4.2, even if the second one shows up more often in the verbatims.

The score is recalculated with every new piece of feedback. A topic with no feedback in the period has no score.

A worked example

Take the topic "Delivery times" on a post-appointment satisfaction survey: 60 verbatims out of the 400 collected in the period, an average sentiment of 35/100, and an NPS split of 45% detractors / 25% passives.

This survey collects neither CES nor ratings. Only three signals are available, so their weights (25 + 45 + 35 = 105) are rescaled to 100%.

Signal	Observed data	Score /100	Weight	Contribution
Volume	60 verbatims out of 400	15.0	23.8%	3.6
Sentiment	35 / 100 on average	48.1	42.9%	20.6
NPS	45% detractors, 25% passives	52.5	33.3%	17.5
Total	CES and ratings: not collected		100%	41.7

41.7 ÷ 10 = 4.2 / 10

The same topic on a different survey

It is the structure of the survey that determines the weight combination. On a survey that includes a rating table, the "Ratings" signal comes back into the calculation, and the combination becomes:

Signal	Applied weight
Volume	25%
Sentiment	45%
Ratings	30%

Here the three weights already total 100 (25 + 45 + 30), so no rescaling takes place.

What this means in practice: the same topic can have a different Business Impact score depending on the survey — not because the underlying sentiment differs, but because the available signals are not the same. This is why it is best to compare topics against each other within a single survey before comparing across surveys.

Key points

The weights are fixed. The platform does not "favour" NPS when volume is low: volume enters the calculation as a score (its share of verbatims), not as a statistical confidence indicator.
There is no minimum verbatim threshold. A low-volume topic that scores very badly can therefore rank high. Read the score alongside the volume shown on the topic.
A topic with no feedback has no score — it is not scored 0.
The score is recalculated automatically with every new piece of feedback, and can be read at every segmentation level available on your data (entity, team, journey type, and so on).