Forecasting

I was having a review of the wonderful sales pacing dash and thinking about targets and forecast.

I'm wondering if/how people have gone about forecasting a final position based on prior performances and what they used in formulas.

eg: 

  1. Averaged Total Ticket Count normalised for capacity &/or number of performances. 
  2. Normalising across day of week and matinee/evening
  3. adjusting for relative subs/single (or other segments you use) at a similar point in the sales period

I was keen to play with sames pacing based on an adjusted forecast target.

Can you tell I'm keen for v16 and predictive modeling?

Parents
  • I've been poking around with Forecasting in Analytics now that we're on 15.2.35. Mystic Aquarium is largely seasonal- we have year-round traffic, but July - Labor Day is by far our busiest time. Is there a way to get Analytics to respect this?

    If we're looking at January through July, the projection will naturally be ' this is trending upward'. However, that projection doesn't know 'Mystic's summer is always higher than it's winters' , or 'Mystic's Fall and winter are always less than its summer', so it won't be able to build that into it's modeling. Given that, how can I use the forecasting in a useful manner?

    For instance, right now, even when looking at all of our Admission attendance data, it's projecting that mid-September and October will be steady at ~2.5k per day. Last year, we averaged less than half of that for our September/October months. Is there a usable way for us to leverage forecasting if the models can't be told 'summer high, winter low'?

  • It should be able to adapt to this, as long as you are feeding the entire year, or multiple years of data in the input, not just the July-Labor day. Whats the setup of the chart you made? because i think the way your describing, the model is assuming a linear relationship, which is the wrong choice. I know the forecasting has multiple models.

Reply Children
  • In the widget I linked, I fed it all of our Attendance data in Tesssitura

    When I zoom the widget and look at the Forecasting, it shows this:

    The Forecast value has almost no slope (2.45 to 2.47), regardless of the Perf Date.

  • I was looking at this earlier today, similar chart but I had mine by month.  It was doing the same thing.  I tried changing the settings to the different model options without much change.  I tried feeding it a lower bound equal to our worst month in history, and still nothing.  Mine is very flat but it's a problem in the opposite direction in that I expect us to do better than it is saying in Oct, Nov, Dec as we have done over the last many years.  I also tried cutting out the pandemic months with a filter and that didn't help either.  
    I am very interested in using this technology, but need to figure out how to make it work better for us first.