I am working with Tableau Pulse metrics and trying to understand how the Expected Range and Forecast Confidence Range are calculated behind the scenes. In my Pulse metric trend chart.
As shown in the attached screenshot, on May 10, 2026, the metric value is 1.19M and Tableau Pulse mentions that it is within the expected range of 1.08M to 1.49M considering seasonal patterns. Similarly, the forecast for May 2026 is displayed along with a forecast confidence range. In this example, I have applied a Month-to-Date (MTD) filter as well
When I hover over the information icon, it says:
“The range that the metric value is expected to fall in, calculated based on the 200 data points…”
I would like to understand technically:
· How exactly is the Expected Range calculated in Tableau Pulse?
· What statistical or forecasting model is being used behind the scenes?
· Can this same logic be replicated in Tableau Desktop using calculated fields or forecasting features?
I have attached a screenshot for reference. Any documentation or technical explanation would be very helpful.
Thanks in advance!
#Tableau Pulse
Hi, I got the below response from the support team regarding my above-mentioned query on Tableau Pulse and how the expected range and forecasting work.
Sharing this here in case someone else has a similar query and finds this information helpful.
>>>What forecasting model does Tableau Pulse use?
Tableau Pulse does not use ARIMA, Prophet, ETS, LSTM, or other off-the-shelf forecasting libraries.
It uses a custom in-house statistical library built entirely on NumPy.
Forecasting and seasonality are handled using Ridge Regression with polynomial and Fourier (sine/cosine) basis encodings.
Trend detection uses Piecewise Linear Regression with a greedy AICc-based breakpoint search.
LOWESS is used for local smoothing.
>>>How is the Expected Range calculated?
The Expected Range is calculated as:
Trend ± 1.5 × Standard Deviation
This does not correspond to a traditional fixed confidence interval (such as 95%), since time-series data is not assumed to be normally distributed.
>>>How do date filters impact the forecast and expected range?
Tableau Pulse dynamically re-queries historical data relative to the selected date context.
The look-back window is:
Up to 200 periods for WTD, MTD, and QTD metrics
Up to 50 periods for YTD metrics
Depending on the selected time range, Pulse queries lower-grain historical data:
Monthly for YTD
Weekly for QTD
Daily for MTD/WTD
>>>What factors influence forecast confidence?
Forecast confidence is influenced by:
Data quality and completeness
Variance in the data
Recency of available data
Forecast horizon length
For example:
Forecasting near the end of a month generally results in higher confidence because fewer future periods remain.
Forecasting at the beginning of a month requires more historical data and may reduce confidence.
If Pulse determines confidence is too low, forecasts may be suppressed from the overview chart. However, they may still be accessible through the Ask experience on the insights exploration page.
>>>Why are historical expected ranges not displayed?
Expected ranges are computed dynamically at query time based on the current or selected date context.
Tableau Pulse does not store historical model snapshots, which is why historical expected ranges are not available.
Please note that the time-based forecasting feature is currently available only with Tableau+ licensing.
KB links-
https://help.tableau.com/current/pro/desktop/en-us/forecast_how_it_works.htm
https://help.tableau.com/current/online/en-us/pulse_insights_platform_insight_types.htm
