Normalization
Normalization is the process of scaling values measured on different scales to a single common scale.
What is Normalization?
Upon experience creation, parameters are associated to your metrics, in order to make sure experiences are based upon quantifiable external data. This external data can come in many forms and shapes. In our normalization process we aim to represent the variable external data into readable scores between 0 and 10.
Example
Raw incoming data could be:
A percentage between 100 and 0 (e.g., % of incidents resolved, uptime),
A numeric rating (e.g. from 1 to 5),
A time-based measure (e.g. waiting time in minutes),
A piece of text,
Etc.
While reading al these raw different values in one overview, it might be difficult to grasp what is a good or bad value. Therefore we transform them into a single understandable and common scale.
Normalization Curves
Normalization isn't just about scaling—it's about how we interpret value across a range. Different metrics may require different scaling curves depending on the context or business relevance.
We currently support the following curve types:
Linear
This method maps the input values evenly from worst to best across the 0–10 score range and is best used when each step in the input range has equal importance or weight:
Manual
This method allows for custom mapping of scores to input value ranges. It's especially useful when value importance is non-linear — for example, when certain values have disproportionate business impact (e.g., 95% might still be unacceptable).
Manual curves give you more control, but require a deeper understanding of the metric’s business meaning.
Configuration
During the Draft status of an experience, normalization can be configured for each metric after selecting a parameter. The normalization setting will always start with a linear curve and a default range of 0 to 10 for the worst and best values.
The graph and its axes visualize how incoming data values are converted into normalized scores. The left axis will show us the resulting score, represented by the colored boxes with numbers in it. Left from these boxes and on the horizontal axis, we can see the corresponding incoming data values. Upon making changes to the normalization fields, we can influence the outcoming scores.
The colored vertical axis shows the resulting score, represented by colored blocks ranging from red (0) to green (10).
The horizontal axis and vales left from the vertical axis show the incoming data values from your selected parameter.
By adjusting the normalization settings, you directly influence how raw data is translated into experience scores.
Step-by-Step Configuration
Begin by defining the numeric boundaries for what you consider the worst and best values for the selected parameter. These define the minimum and maximum ends of the scale.
Select between a linear or manual normalization curve, depending on how you want incoming values to be translated into scores.
Linear
If we decide for a linear curve, we can configure it in two ways: via the field named Worst or via the field placed left below the red 0 on the vertical score axis. For the best value, we can fill it in in the field named Best, or left slightly below the 10 score value.
Example:
A survey rating’s values vary between 1 and 5. We set up the question so, that 5 is the most positive outcome. Then we fill in 5 as the best value and 1 as the worst. Upon filling in these values, our graph changes: values in between are evenly scaled:
For more examples refer to Choosing the Right Curve.
Manual
If we decide for a manual curve, we can configure all raw data values for each score. We can still set the top and bottom boundaries via the Worst and Best fields. Besides these values we can now also adapt the values considered for all other scores between these worst and best value. This is done by filling the fields besides the scores shown on the colored vertical axis.
Example:
An uptime % of the wifi connection in an office will vary between 0 and 100%. 100% will be the most positive outcome, but 95% is already a very bad outcome. We fill in 100 in the Best field and 0 in the Worst field. Then we can adapt the most left axis values on what we consider to be average scores:
We consider everything between 100% and 99.5% to be a 9, between 99.5% and 99% to be an 8, everything between 1% and 85% to be a 1 etc.
Choosing the Right Curve
Different types of raw input require different normalization strategies. Below are examples of how to interpret, normalize, and assign value ranges to commonly used metrics.
Percentage (%) Metrics
% of issues not updated in the last 5 days
Curve: Linear
Reason: Each percentage increase represents a steady decline in operational performance.
Value configuration:
Worst: 100 → Score 0
Best: 0 → Score 10
% App Uptime
Curve: Manual
Reason: Uptime is not linearly perceived. 100% is expected, while even 95% may be unacceptable in mission-critical environments.
Value configuration:
100 → Score 10
97% → Score 6
95% → Score 4
90% → Score 1
0 → Score 0
Rating Scales (e.g. 1 to 5 Stars)
CSAT (Customer Satisfaction Score)
Curve: Linear (optional: Manual if you want to amplify high/low scores)
Reason: Ratings generally scale evenly, but you may wish to emphasize low scores more strongly.
Value configuration (Linear):
1 → Score 0
5 → Score 10
Time-Based Metrics (e.g., minutes or hours)
Average waiting time (in minutes)
Curve: Linear or Manual
Reason: If expectations are fixed (e.g., <10 min is OK, >30 min is bad), then manual is better.
Value configuration (Manual):
0 min → Score 10
10 min → Score 7
20 min → Score 5
30+ min → Score 0
Textual Data (after Sentiment Analysis)
Sentiment score from textual employee feedback.
Normalization is standardized using sentiment analysis on the provided text. Extremely negative sentiment is a 0, while very positive sentiment is scored a 10.
Last updated