PowerPivotPro is Coming to Atlanta

March 20 - 22, 2018

Registration for 2018 Public Training is now open!


**Use the discount code “3ORMORE” when signing up 3 or more people.

MARCH 20 - 21

Foundations: Power Pivot & Power BI

Instructor: Austin Senseman

Super charge your analytics and reporting skills with Microsoft’s dynamic duo. Designed to handle huge volumes of data, these tools will transform the way you work! Two Days in our class and you are EMPOWERED!

  • Learn Microsoft’s secret weapon behind Power Pivot & Power BI: DAX
  • Taught by Austin Senseman – PowerPivotPro Partner
  • You don’t need to be an IT professional – most of our students come from an Excel background
Atlanta Public Training Classes
Atlanta Public Training Classes

MARCH 20 - 21

Level Up Series: Advanced DAX

Instructor: Ryan Sullivan

The Advanced DAX Course was such a hit in the first half of 2017 that we’ve expanded the course to 2 days!


  • This advanced DAX training class is taught completely in Power BI Desktop
  • Students are encouraged to take our Foundations course and have hands on experience with the DAX language
  • Taught by Ryan Sullivan – Principal Consultant
  • Class material drawn from usage of Advanced DAX applications while consulting with hundreds of international firms


Level Up Series: Power Query for Excel & Power BI

Instructor: Ryan Bergstrom

The second class in the series is our Level Up Series is Power Query for Excel & Power BI.

  • Students are encouraged to take our Foundations course and have hands on experience with Power Query in Excel or Power BI Desktop
  • Taught by Ryan Bergstrom – Principal Consultant and Trainer
  • We will cover common to unique business challenges made easy with Power Query’s data wrangling capabilities
  • Intermediate to Advanced Level Power Query best practices distilled into easy to understand patterns to apply to your most common business challenges
Atlanta Public Training Classes
PowerPivotPro Logo

Moving Average in Power Pivot

The Blue Line Smooths Out Random Fluctuations, Tells a Less “Over-Reactive” Trend

I realized recently that this topic has never been covered before, in its most straightforward form, on this site!  Actually, it was the subject of a guest post by the esteemed David Churchward, and also by the equally-esteemed Kasper de Jonge, but neither of those posts benefited from the v2 functions available to us today).

To illustrate what we can do with state-of-the-art Power Pivot formulas, let’s start with this simple model:


Simple Power Pivot Model

And a simple pivot:


That [Units Sold] measure is the jagged red line in the chart at the top of the post, and its formula is very simple:

[Units Sold] = SUM(Sales[QtySold])

And we want a version of [Units Sold] that is “smoothed” over a 3-month period.

Moving Sum

Let’s start with a formula that is a sum of the most recent 3 months (including the current one):

[3 Month Moving Sum Units Sold] =
CALCULATE([Units Sold],
                        LASTDATE(Calendar[Date]),-3, Month

And see what that looks like:

Power Pivot Moving Sum

Moving 3-Month Sum Reflects the Current Month and the Prior Two Months

Moving Average – First Attempt

OK, but that number is bigger than a single month and doesn’t match the scale of our real-world business, so we wouldn’t want to chart that – we want the average version of that.

It’s a 3-month moving sum, so to get the average, we could just divide by 3:

[3 Month Avg Divide 3] =
=[3 Month Moving Sum Units Sold] / 3

Which looks like:

Power Pivot Moving Average with a Drawback

3 Month Moving Avg Via Divide by 3 Has a Drawback

Those first two months, since they are the first two months in our calendar, are summing up less than 3 months’ worth of sales, but still dividing by 3.  So it “unfairly” drives down their average.

Moving Average – Corrected

We can account for this by changing our denominator to use a similar logic to the numerator:

[3 Month Moving Avg Corrected]=

[3 Month Moving Sum Units Sold] /

In English:  “take the 3 month sum measure we already have and divide it by the number of distinct (unique) months we have over that same 3 month period.”


Power Pivot Moving Average Fixed

This Calc is More “Fair” to the Months at the Beginning


There are a number of variations on this approach – daily/weekly/quarterly versions, correcting for calendars that extend beyond the range of dates where you have sales, adapting it to custom calendars via the Greatest Formula in the World, etc., but I will wait and see what people ask about in the comments before digging into any of those.

  Subscribe to PowerPivotPro!


Rob Collie

One of the founding engineers behind Power Pivot during his 14-year career at Microsoft, and creator of the world’s first cloud Power Pivot service, Rob is one of the foremost authorities on self-service business intelligence and next-generation spreadsheet technology. 

This Post Has 29 Comments

  1. Is it complicated to control “X Months moving” from Excel?
    Have a cell in Excel where we can put 3, 6, 12 or even 1.

      1. Having a slicer to easily/quickly (re)set the moving average length would GREATLY enhance the ability to see the effect of the ma-length on the graph / analysis…

        1. Also, I have created columns in the PP data model having differing moving-average lengths (c.f., lagged: 1,3, 6 mos; & fwd/backward: 1, 3, 6 mos) for a 1.8M row revenues-ledger table. The 2013 .xlsb file has ballooned from ~220MB to over 400MB. Being able to ‘dynamically’ change the ma-period should COMPRESS ALL of this to just a FEW columns — (as well as make it MUCH easier to investigate / select the effects of the ma-length)…

      2. to may it control ‘X months moving’ , how do i use slicer to replace the number of intervals(-3)? Please advise. Thanks a lot

        1. Just use the Search box at the top of the page Billy. “Moving Average Slicer” should do the trick 🙂

  2. I had a different problem: what to do if don’t know exact length of the period? Imagine the situation of some technology where you need to keep things in certain conditions over some period of time and it varies. Simply you have two dates START and END. In Excel it is not difficult using SUMPRODUCT or some kind of array stuff but it’s slow and COPY/PASTE routine is required each time you update. How to do it in DAX then when you keep process parameters in separate table? First thing, setting a relationship led me nowhere (in my opinion it is variation of MANY to MANY when you have parallel processes). I tried to apply USERELATIONSHIP but failed. The solution I found is simple but you need cumulative data. The formula of SUM over the period from A to B looks like: value(B)-value(A), AVERAGE: (value(B)-value(A))/(B-A). The funniest part is you don’t need any relationship at all, just take LOOKUPVALUE to find value(B) and value (A). Maybe this solution is obvious to everyone but for me was not but I found it refreshing.
    Greetings from Poland.
    Forgive me my English – I do not use it very often

  3. Hi I Used the Fomula Calculated Field 1:=Calculate([Sellout Product Units],Datesinperiod([Date],lastdate([Date]),-3,Month)).. the Date feild in e.g.- 5/1/2013 12:00:00 AM, i got an error stating DatesinPeriod Function are only accepting date column reference as a first argument

  4. Hi!

    I would love some help to do the “correcting for calendars that extend beyond the range of dates where you have sales”, as you write in the variations above.

  5. I am struggling to do this for 7 day moving average:


    Could anyone advice me on this please.

    Kind Regards.


  6. Hi, I’ve tried to replicate the 3-month moving average in my file. I followed the instructions above. My moving sum/average only contains the last 3 months (not the whole period with their corresponding data as showed in the first picture above, i.e.I got date only for 2001/09, 2001/10, 2001/11). Could anyone advice what could be the problem

  7. It works! But…..I am working with days – specifically, I need to get a rolling 28 day average of one of my measures. This works fine when I view the result with my date attribute. However, my Calendar dimension also has a discreet, non-continuous attribute called DateMonth (which looks like 1-Jan, 2-Jan etc.)

    I was hoping this would allow me then to take my year attribute and compare the 28 Day Average for the 1st of Jan across multiple years.

    So you could have a line chart where the axis is 1-Jan to 31-Dec (with no reference to the year) and the series category is by year, so a line for each year.

    Hope someone could help 🙂

  8. The last equation provided – Moving Average Corrected – delivers a denominator with a value of 12 (i.e., it is dependent on when my Calendar[year-month] starts. So to have a correct 12 month moving average, my data and calendar need to start at the same time). Any suggestions?

    1. I figured it out:


      What I did was started counting my months starting from the first date of my sales (i.e., 40663 is the last date before I wanted to start counting the months where my sales were).

      Hope this can help someone searching this blog post!

      1. William, you are a life saver. I was attempting to figure out why my denominator held 3 for the oldest month of data I was working with. I appreciate your example and explanation which applied exactly to my situation.

  9. I have a better solution to the final comment in here – one that calculates the number of months to divide by, if your data and calendar do not start at the same time.

    This was the original calc:

    [3 Month Moving Avg Corrected]=

    [3 Month Moving Sum Units Sold] /
    CALCULATE(DISTINCTCOUNT(Calendar[Year Month]),

    And its within the divider that the edit is needed – to correctly work out where to divide by 2 and then 1 (for the first 3 months in your data).

    The below works a treat:

    3 Month Moving Average – Working = [3 Month Moving Sum]/
    CALCULATE ( COUNTROWS ( VALUES ( DimDate[CalendarMonth] ) ), FactSales ),

    Where FactSales is your fact table obvs.

    hope this helps someone else out – as it drove me mad!


    1. This works great, however i adapted it to a 7 day moving average, however the last 7 days are still divivded by 7, making a large drop off, any idea how to fix that?

  10. Hello Experts,

    I have a question.

    What if we need the average of last 30 Trading days or working days.??

    Of Course, we would have a Date column representing trading days.and corresponding sales amount.

    1 Month average with your formula won’t be the same thing. As Exact 30 days’s average is needed here.

    Would you please suggest a formula ??


  11. The difficult thing is the calculation is based on the financial date other than the calendar date, as almost all existing time intelligent date functions cannot be used. Any sample calculation formula based on the financial date?


    1. I got a plenty of info about the calculations based on the customer dates from the above link “Greatest Formula in the world” . It’s very helpful…..

  12. This is a great posting by all. I want to know what if the sales in one of the past three months is zero or no sales at all, then how would the formulas suggested work.
    I want to find out the number of months from current month to last three months having sales. Say I am in May 2014 the last three months will be including May 2014 , Mar 2014, Apr 2014 and May 2014. But Apr 2014 does not have sales. So it should be sum of Mar and May 2014 sales divided by 2 and not 3.
    Assume we have sales for Jun 2014 and Jul 2014.
    For Jun 2014 the average last 3 months should be ( My 2014 + Jun 2014) / 2.
    For Jul 2014 the average last 3 months should be ( May 2014 + Jun 2014 + Jul 2014 ) / 3.

    Any help to achieve this will be greatly appreciated.


    1. Hey there, brilliant solution.

      But now I’m looking for the same addition as CheenuSing mentioned.

      Some series in my data don’t have sales in the last period but others do. CheenuSing, have you found the solution to that? I’d be most grateful, thanks.

  13. Hi Everyone!

    I´m new with all DAX functions and quite struggling with my moving average calculation. I just need the moving average of 3 months back NOT including the current month. I tried using “IF” but never worked out. Can some one help me with it?

  14. So this is great except I have 3 months of the moving average showing into the future that I want to not display on my chart. Since I can’t add a visual filter for date <= today, how do I integrate that into the queries?

  15. when I am trying to replicate this formula, it is only calculating last 3 months of the year correctly for moving average but not for the rest…

Leave a Comment or Question