What is a Signal?

min read - Updated on Nov 03, 2026

A Signal is a statistically tested finding in customer feedback: a topic that customers mention more often, or more negatively, in one group of feedback than in the others, or a topic whose mentions changed over time. Kimola looks for Signals in every Report and every Space, and lists them on the Signals page in two lists: Differences and Trends. Each Signal comes with the numbers behind it, quotes in customers' own words and an explanation of why it is shown.

A dataset of a few thousand reviews already holds many groups: the platforms the reviews come from, the products they are about, the countries, the customer types, the answers of each Dimension, and the weeks or months they were written in. Checking every theme in every group by hand isn't practical, and a large share alone tells you little: a theme that is common everywhere is common in every group too. Signals answer two narrower questions for you: which topics stand apart in one group of feedback, and which topics are rising or falling.

Kimola first classifies every piece of feedback with themes, sentiment and Dimensions. It then crosses every topic with every group your data offers (the values of your descriptive columns, such as product, brand, platform or country, and the answers of your Dimensions) and tests each combination. The Signals page says how much was tested, for example "1,866 combinations tested in 1 second", and how many records were read, for example "2,118 of 2,121 records analysed." with "3 duplicates removed." A combination becomes a Signal only when its difference passes a statistical test that is adjusted for group size, so a small group can't stand out by chance. Kimola scores and ranks Signals with deterministic methods: the same data always gives the same Signals in the same order. An AI model writes each card's title and text afterwards, from numbers that are already computed.

Getting Ready

To explore Signals, you need a Kimola account and analyzed customer feedback in a Report or a Space. If you don’t have an account yet, sign up first.

For a Report, sign in and open the report you want to work with from the Reports section. If you haven’t created a report yet, follow one of these guides: Create a Report by Uploading a Custom Dataset, Create a Report by Adding Links

Kimola generates Signals automatically after the Report’s analysis is complete. To explore them, see Signals in Reports.

For a Space, you need a Funnel and a Space with classified feedback. If you haven’t set these up yet, see Create a Funnel and Create a Space. Signals run after every classified file, and the Signals item appears in the Space’s menu once its first Signals are available. See Signals in Spaces.

Differences

A Difference shows where a topic stands apart in one group of feedback. The group is a value of one of your columns or an answer of a Dimension, for example the reviews from one platform, about one product or from one persona. Kimola compares it either with the other values of the same column (the other platforms) or with the rest of the feedback. A Difference can point up or down: a topic that one group rarely mentions is a finding too.

A Difference uses one of two measures:

Mentions is how much a topic is mentioned in a group of reviews: the share of its reviews that mention it. For example, Delivery Delays is mentioned in 18% of the reviews written on the Android app and in 6% of the reviews on other platforms.

Negativity is how negatively a topic is mentioned in a group of reviews: the share of the reviews mentioning it that are negative. For example, Battery Life is mentioned about as often for two phone models, but 64% of those mentions are negative for one model and 22% for the others.

A Trend shows how the mentions of a topic changed over time. Its measure is Change: the share of reviews mentioning the topic in the latest periods against the periods before. Trends are built from the dates of your feedback, grouped into periods such as weeks, and every Trend states the periods it compares, for example "2026-W08–2026-W09 compared with 2026-W04–2026-W07 (by week)".

A Trend can cover all of your feedback or one group of it. Trends help you catch an issue while it grows and see whether a topic fades after you act on it.

Example

For a mobile app, App Crashes may be mentioned in 21% of all reviews in the last two weeks and in 9% in the four weeks before. Another Trend may find the same rise only in the reviews of one app version.

How to Read a Signal

Every Signal is a card, and every card has the same parts. Read it from top to bottom: the numbers first, then the words, then the evidence.

The two shares. The two large percentages on the card are the measure in the group and in what it is compared with, each with its name, for example "18% Android app" and "6% other platforms". The line under them names the column and the two sides, for example "platform: Android app other platforms".

The consistency. A label tells how well the finding holds when Kimola splits the reviews into comparable groups: positive and negative reviews, short and long ones, and the values of your other columns. Consistent "holds the same way across positive and negative, short and long reviews and the other compared columns". Mostly consistent "holds the same way in most of the comparable groups of reviews". Varies "holds overall but not in every comparable group of reviews": open its records to see where it comes from. Holds overall means the finding is confirmed on all the reviews together, while each comparable group has too few mentions to break it down.

The title and the text. A short title sums up the finding, and two or three sentences explain it with its numbers, for example which other group shows the same topic at a different rate.

The quotes. Up to three reviews that carry the finding, in customers' own words. Click Show more >> to read a long one in full.

Why This Is Shown

Click Why this is shown under the quotes to see the evidence behind the card. Not every card has every line:

Compared with: names the comparison, for example "other platform values" or "reviews without this tag". After "strata", it lists what the test keeps equal on both sides, such as sentiment and how many labels a review carries, so the group is compared with reviews of the same kind.

Numbers: gives the counts, for example "64 of 356 reviews, 43 more than the comparison predicts (3.05×)": 64 reviews of the group mention the topic, 43 more than the comparison's rate would predict for a group of that size, which is 3.05 times the expected count.

Consistency: repeats the consistency label with its meaning and how many comparable groups it was checked in, for example "(the same way in 3/3 comparable groups)".

Contrast: shows the same topic's share in the other values of the column, for example "iOS app 7%, Website 5%".

Related tags: lists labels that often appear in the same reviews, with the model or Dimension they come from and their share, for example "Late Courier (Pain Points, 31%)". They give context around the finding. They don't mean that one label causes another.

Listing echo: warns that the topic's words appear in the group's own name, for example a topic about returns in a group named "Returns", so the finding may already be known.

The last line names the label that was tested, with the statistics of its test.

Act on a Signal

Each card ends with its actions. In a Report, Open in Tables opens the records behind the Signal in Tables. In a Space, Open in Content opens the Space's records filtered to the Signal, and Make a task turns it into a Task with its evidence.

Below the actions, you can give the Signal a verdict: Known, New & important, Unimportant or Artifact, with an optional note. A verdict follows the same finding into later runs. See Review Signals.

Hidden Signals

Kimola checks each card before it shows it. A card is marked Not supported by the verbatims when the reviews behind it don't back up the finding, and Volume driven when it comes mostly from the size of a group. Marked cards are hidden unless you choose to show all Signals, so the lists start with the findings that hold up.

At the top of the lists, three filters narrow what you see: Any measure (or one measure), Any consistency (or one consistency level) and All columns (or one column or Dimension).

Signals in Reports and Spaces

Signals read, test and look the same in Reports and in Spaces. What differs is when they are generated and what you can do with them.

In a Report

Kimola generates a Report's Signals after its analysis completes, from the records of the Report. Re-generate generates them again from the current records. See Signals in Reports.

Note

Re-generating permanently deletes the current signals and all the feedback given on them.

When the data gives the tests too little to work with, the page says "No signal has enough evidence in this report". Older Reports may show their Signals in an earlier format.

Tip

More reviews, or more descriptive columns (product, brand, platform, country), give the tests more to work with.

In a Space

A Space's Signals run again after every classified file, over the whole Space. Its groups are the Space's columns plus the file itself, so a topic that stands apart in the latest file shows up as a Signal. The Signals item appears in the Space's menu once the Space has its first Signals. See Signals in Spaces.

In a Space, a card opens its records in Content and offers Make a task, which turns the finding into work for a teammate with its numbers frozen as evidence. See Create a Task.

Note

Signals in Spaces are available in the Enterprise package. They are also available in Kimola On-Premises.

A Signal is not an Insight. A Signal is a tested finding that compares groups or periods. An Insight is a text Kimola writes about a Space's latest feedback after every file. Use Signals to find where a segment or a period stands apart, and Insights to read where a Space stands today.

Was this article helpful?

Tell us about your thoughts and experiences regarding the article.