How exactly does the analysis for continuous stories work? This article will explain the different analytic types that power each sentence in your continuous narrative.

Continuous stories measure trends over a certain period of time. The analytics types for these narratives include performance, progression, averages, totals, streaks, volatility, segments, and predictions.

Let's look at the sample narrative below, which measures Shipments per Month:

Breaking down the analytics types that are driving each bullet, let's begin with the first two bullets:

• "Average Number of Shipments was 208.21 across all 48 months."
• "The minimum value was 46 (2014-Feb) and the maximum was 462 (2017-Dec)."

These sentences use average and range functions to call out the average, maximum, and minimum values across the entire time period you are analyzing. This is high level analysis.

The next bullet covers overall performance of your measure over the entire time period, analyzing the trend over the period. Did shipments increase? Decrease? Did the trend at the end of the period match the overall trend?

• "Number of Shipments increased by 485% over the course of the series and ended with an upward trend, increasing in the final month."

Next, the narrative uses progression analysis to call out the largest increase/decrease based on your measure during the time period, using both a percentage basis and absolute basis:

• "The largest single increase on a percentage basis occurred in 2014-Mar (+241%). However, the largest single increase on an absolute basis occurred in 2017-Sep (+241)."

The next few analytics types for continuous stories are correlation, segments, and trendline.

Correlation analytics will call out any notable correlations between different series in the data. See an example of correlation analytics below:

• "Of the three series, the strongest relationship was between Corporate and Home Office, which had a moderate positive correlation, suggesting that as one (Corporate) increases, the other (Home Office) generally does too, or vice versa."

Segment analytics calls out noteworthy increases/decreases in the time period. You can adjust the settings for segment analytics in the settings module under the Analytics tab. Specifically, you can change the threshold for percent changes that you find noteworthy and want to be mentioned in the narrative. For example, if only changes of 60% or greater are interesting to you, the narrative will not include content about a trough in a time series that featured a 30% decline. See examples of segment analytics bullets below:

• "Number of Shipments experienced cyclicality, repeating each cycle about every 12 months. There was also a pattern of smaller cycles that repeated about every three months."
• "Number of Shipments had a significant positive peak between 2014-Oct (159) and 2015-Feb (64), rising to 318 in 2014-Nov."

The final type of analysis for continuous stories is trendline analysis. Trendlines are determined by how well they fit your data with a certain percentage of confidence. You can adjust the settings for trendline analytics in the settings module under the Analytics tab. Specifically, you can set the minimum confidence level that the narrative should include content related to an overall trend line and prediction. For example, if an overall trend line can be drawn at 90% confidence, but you have set the threshold at 95%, content about the overall trend or predictions will not generate. You can also turn on/off predictive analysis.

Below is an example of trendline analysis:

• "The overall linear trend of the series rose at \$902 per month for an absolute change of \$42,394 over the course of the series. If this trend continued for the next one month, Sales is predicted to be about \$69,598."

If there is no clear linear trend in your data, you may see a sentence as such:

• "A prediction could not be made because Number of Shipments did not have a good linear trend."