Cognos Analytics 11.1.4 brings exciting and long sought forecasting capabilities to the platform. While Cognos is far from the first analytics tool to offer forecasting, IBM’s implementation is more flexible, more powerful and easier to use than competitors. Let’s take a dive into Cognos Analytics forecasting to understand why.
Where is forecasting available?
Forecasting is available in line, column and bar visualizations in Dashboards, Stories and Explorations as of 11.1.4. Report Authoring is on the horizon (in fact there’s a ton of cool stuff coming for Report Authoring.) The ability to forecast column and bar charts sets Cognos apart from Tableau and Power BI. Cognos hides the forecasting button until a visualization meets the following criteria:
- Visualization is the proper type (line, column or bar)
- Visualization has a measure in the Y axis
- Visualization has a time element in the X axis
Once this criteria is satisfied the forecast button appears next to the insights button in the top right of the visualization.
In the above example I use the ‘Great outdoors data module’ available in the samples. Using the Sales table I apply Revenue to the Y axis and nest Year and Month on the X axis – the perfect scenario for forecasting.
Single category forecasts
Click the forecast button, turn the feature on and there you have it – your very own Cognos Analytics forecast! You may immediately notice that the slope of the line changed dramatically once a forecast was applied. I first assumed this was a bug until I clicked the yellow ‘!’ next to the forecast button and read these magical words:
That’s right – Cognos automatically reordered the time categories to be in chronological order rather than alphanumeric! I’ve been waiting for this feature for all my life and it’s finally here. Now we just need IBM to automatically apply it to all visualizations, not just forecasts.
In the above example I add ‘Retailer Type’ from the Retailers table to the color attribute of the visualization. Cognos instantly re-applies the forecast to each individual category. This is one advantage Cognos has on Power BI, which currently forecasts a single category only. This makes data discovery on forecasts extremely fast and easy.
Interacting with forecasts
Hovering over an individual point in the forecast will show you the confidence upper bound, forecast and lower bound for that intersection of the chart:
In this example I hover over the projection for ‘Outdoors Shop’ in September, 2018. You can see Cognos’ predictions for the upper and lower bound.
Clicking on the line for ‘Outdoors Shop’ will filter the dashboard by this category, bring the selected category into visual focus and plot the confidence intervals as they evolve in the forecast. Super duper cool.
Forecast configuration options are available in the menu that appears when the forecast button is clicked.
Let’s take a look at what these options do:
- Forecast periods: The number of periods included in the forecast. The default ‘Auto’ will project forward 20%. If you have 10 months of data, auto will generate an additional 2 months of forecast
- Ignore last periods: Ignore last periods is useful for cases where you have incomplete data at the end of your chart – for example, the month of November currently has 6 days worth of actuals. Ignoring this incomplete period creates a more accurate forecast.
- Confidence level: Controls the confidence level displayed on the tool tip and confidence visualization – options are 90%, 95% and 99%.
- Seasonality: Cognos automatically detects seasonal fluctuations in your data and accounts for them in its forecasts – think about retailer revenue during the holidays for example. The default ‘Auto’ setting will build multiple models with different seasonal periods and and select the best one, however you can specify a season period by entering numbers here.
That last option is incredibly powerful and another strong advantage for Cognos vs Power BI and Tableau. The inability to detect and account for seasonal variation when forecasting renders this functionality useless for many industries – retail, hospitality, utilities – and only Cognos has it.
Cognos uses exponential smoothing models to generate these forecasts. I don’t know what the means either but you can read about it here. The important thing to know is that you have access to the forecasting statistical details in the data tray.
I’m a BI/DW guy and this screen might as well be magical incantations to me, but the fact that Cognos provides this level of detail means that I can always find a friendly wizard to explain it. IBM has a nice explanation that I also don’t understand here.
Bar chart forecasts
Switching to a column/bar chart retains the forecast and forecast parameters that were applied to your line chart.
One thing it doesn’t do, which I just discovered in writing this blog post, is automatically sort the time category in chronological order like it does with line charts. I’m not sure if this is purposeful, an oversight or a bug. You can vote for my feature request to rectify this here.
Cognos Analytics forecasting is the real deal
I have to say, I think this feature is extremely well done and I’m certain end users are going to love it. If you need help prepping your environment for self-service see my guide here. Allowing your users to build their own data modules will make forecasting even more powerful, read more about it here.
IBM was late to the market with it but delivered something much better than their competitors. This has been a recurring story in the last year – Cognos has come so far since the 11.0 releases. It’s up to IBM – and to us – to make sure the broader BI community is aware of the good things going on in Cognos Analytics.