Dynamic Parameters for What if analysis
[Will Thompson] I've just merged a couple of very close scenarios, which all have similar great ideas! Something we're considering for the future, so I'm keen to hear your feedback on this. I like this core scenario of a numeric value that is bound to a slider control, and can then be used as part of some expression.
- In a sales report, I can adjust a '% increase' slider to see what impact a small change in revenue might have had on my sales.
- In a mortgage calculator, I can change the % rate, borrowed amount and term to see projected repayments
- In a quota report I might change a 'Opportunity confidence' control to see the likely bonus I could earn at the end of the year.
Please, keep voting and help us prioritize this!
Original suggestions follow:
So this scenario brings what-if-analysis to the visualization without changing the underlying data source. Let's say I have a table that has numeric values (integers, floating values, etc.). I created a chart (column, line, etc.) visual of that table data. Then instead of using SUM, AVG, COUNT, DISTINCT COUNT, I now can use a math expression applied to the numeric value (say I want to see what my chart looks like if I increase sales by 10%). Then I would see the new column in my table (calculated column with new values). I can now add that new column to the chart and graph that, labeling that as 'if sales increased by 10%' to show how that would compared to the actual numbers I see.
This is something we’re planning to start in the near future. Stay tuned and let us know if you have any feedback in the comments!
Bill Ritzel commented
The date slider is already there - just give us a version that returns a number between two specified values, with defaults of 0 and 100 for percentages, and I believe that it would cover a lot of us.
John Merchant commented
This is great news. I built a pricing model where users can change the price for each product (50-60 products) and see the impact to sales using a filter / disconnected parameter table. The issue is that there are 50-60 parameter tables within the data model, and our sales force mentioned the report filters are not intuitive. A free form input box would be ideal for my situation, where users can enter a number (example value: 1000,2000,3000), and use this number within an expression (units * example value = new sales); but the method described in this post should help as well. I am excited to see how this new feature can help improve the data model and user interface of my report. Thank you!
Alan Strange commented
Brilliant! I work for an insurance company so I'd like to be able to model eg what if premium were x% more, or claims y% less. And what would be great would be to be able to see what would be the effect of an excess of £z (or adding a fixed amount to an existing excess), meaning reducing each claim by a set amount (ideally bound by a minimum value - ie minimum adjusted value of £0).
This is HUGE especially from a competitive perspective. Looking forward to the release!
It would be cool if you could add an additional row for a fact table. For instance, we use power bi for monitoring portfolio exposure, and in order to "what-if we add or remove position X and Y" we need to reload the entire model. Another ting would be to change values of current rows.
Great, great news!! Tableau have this feature since a long time and it is very used by users. It is a little feature but with a strong impact.
Neville de Sousa commented
It would be great if the what-if analysis will be able to handle modifying discrete values.
I work in the logistics industry and look at turnaround time (estimated and actual: arrival to complete load) for ships coming to load at our terminal. As such we look at the overall turnaround for x number of vessels over a particular future date window.
It would be great to be able to modify the estimated arrival and complete load for 1(or multiple) ship(s), which then updates the turnaround metric, to gauge the impact to the overall turnaround. This will have the added benefit of smoothing out load on terminal assets.
Alan Walnoha commented
They have a slider in their latest release, but it's still in its infancy...
So if I understand this thread, a numeric range slicer control is being reviewed for possible future inclusion in Power BI, but there is no What If solution currently available? Am I right?
Duncan Pryde commented
Just as an aside, I've been trying to use the numeric range slicer for exactly this purpose, but run into an issue as described here: https://ideas.powerbi.com/forums/265200-power-bi-ideas/suggestions/18810706-enable-discrete-values-in-numeric-slicer and in more detail here: http://community.powerbi.com/t5/Desktop/Numeric-Range-Slicer/m-p/151282
I can get around it by removing the slider and keeping the entry boxes, which may work for may of the cases here.
In my scenario, we wanted to create an upper and lower bound for what-if analysis, so the range slicer is perfect for that purpose, but being able to enter a single value would be very useful.
What if analysis would close the gap on Tableau and increase adoption
SUBHOLINA DEB commented
WHAT IF ANALYSIS IS VERY MUCH WANTED FEATURE IN POWER BI??
Tien Dung Thoi commented
I like the "% increase" slider idea, please do that.
For many more "conservative" users, a "data entry" box where they can type in 1.123456% would be even better, but that will possibly require the notion of a global variable, a separate thread.
Mathias Thierbach commented
In tabular modelling it is a very common (and useful) patterns to define "Parameter Tables" (see the pattern described here: http://www.daxpatterns.com/parameter-table/). In a way, those "tables" are a neat work-around as they don't technically contain a dimension of the database, but instead are created primarily for the purpose of creating slicers for users to interactively change the behavior of certain measures.
Both, Excel and Power BI, as clients of tabular models already distinguish between two "types" of tables when representing the schema to users, even though both use slightly different semantics.
Excel groups all measures from a given table into a distinct node and keeps them separate from the table's columns. Power BI displays a "measure group" node only when the containing table has no visible columns, otherwise it merges all measures and columns together into the same table node (which makes for a very poor user experience).
Despite the differences, though, both effectively have a notion of "measure (only)" tables vs "dimensional" tables in terms of how the model is displayed to users.
I would like to propose a change where data modelers can mark any table as a parameter table, to be used by the Power BI UI so that it would use a third type of icon (something like a slicer?) to represent those tables and keep them separate from tables that represent a genuine dimension of the data model.
Especially when dealing with more complex models, this would enhance usability enormously, at a very low cost. The use case I'm targeting here is where there is an organisational split between data modelers and analysts/report builders - which should be fairly common in enterprise environments. Where the semantics of available models need to be communicated to consumers.
Absolutely need this functionality in Power BI. Ability to do what-if analyses is crucial to making visualization operationally useful.
we would like to have this possibility. To change one or some parameters to look the impact of some changes.
What will be my charge if some taxe goes for 1 to 1.2 ?
Thank you to add this !
This describes what we need also; thanks to whomever shared. A slider for different scenarios would be a great enhancement.
This is a much needed feature for a true BI application. I view this feature similar to how you could have a linked table in a PowerPivot model, and each user with a copy of the model can have their own set of "what-if" values/parameters that affect the measure calcuations. This doesn't seem to be an option for SSAS Tabular, unless the model only has 1 user that determines the "ad-hoc" sources.
Here's an article that describes this scenario in PowerPivot and what would be nice from a Power BI and SSAS Tabular perspective.
Rajnish Sinha commented
For the existing parameter feature, you have to set the parameter value before hand and it remains constant until you change and refresh your model again. The request here is to create dynamic parameters that you can adjust without the need to refresh the model. Also, the ability to use this parameter value in creating custom columns (calculations), or measures. And the ability to expose this parameter in form of a text box or a range slider so that you can use this on your report and make changes to the value affecting the calculated fields/measures. Please let me know if additional information is required.
Can you please explain how the current parameter implementation doesn't solve this?