I didn't get to attend the last meeting, but I'm interested in meeting with others to help me ramp up my Tableau skills (I'm pretty much brand new to it). I have a lot of economic information I want to report on, particularly for forecasting machine tool sales. I am also using Tableau for my company's sales data.
One item I would be interested in is using historical trends in my company's sales data to forecast the current year. For example, I know the historic pattern by which we book advertising for the upcoming/current year. We track the day the order was booked and when that order will be delivered (e.g. the January 2015 issue of one of our magazines or a banner for February on of our websites). I've created a chart that compares years with the running total of advertising booked based on how many days prior to the end of the year that amount was booked. Sorry, it's hard for me to explain this in words. Hopefully the picture below helps.
In Excel, I used to use the percent of the year total booked as of a given date to forecast the current year total projection. But, it was hard to do this for anything of the total company since there were too many other variables to manipulate (by brand, by product, by sales person, etc.). If I could get this set up in Tableau, then I could simply filter filter on the various dimensions to see where each one currently stands relative to previous years as well as projected total compared to budget. Another interesting item to be added to the viz is a list of the top 10 or 20 customers that advertised in previous years but we don't have a schedule for the current year yet.
Russel responded with the following via email:
In general Tableau does do forecasting:
http://onlinehelp.tableausoftware.com/current/pro/online/mac/en-us/forecasting.html
However I do not believe it is very robust, so a lot of people probably leverage R/SAS/SPSS to build their models. It doesn't sound like you have a very complex model so Tableau might fit your needs.
To be able to filter by the dimensions you are describing you'll need to create a more granular model. For example: If I create a Monthly Forecast, I can show numbers via. Months/Quarter/or Year, however I lack the ability to do Days (I lack the granularity in my model, I cannot drill down). The trick is to figure out the most granular level of data you have and build the model from there, that way you can roll up to all the views you want. So maybe I create a Hourly forecast, I can roll-up hours to Days->Weeks->Months->etc. Depending on the amount of data you have, that could be a wise decision or a bad decision (how much confidence do you place in your data, as you get more granular the amount of data in each forecast is smaller).
