Calculation average rate with 0 values

[Anonymous]

Hi everybody,

I need your help.

Our customer needs to calculate an average rate of efficiency on 6 days, without lines = 0.

Example : 

Date	Notes
05/10/2019	8
06/10/2019	5
07/10/2019	0
08/10/2019	15
09/10/2019	0
10/10/2019	13

Average = 6.833...

Expect result = 10.166...

Here is the DAX measure we tried to do, but of course, it's not OK.

Gamme théorique/charge (6jours glissant) % = Calculate(DIVIDE(sum('Heures théoriques gammes'[Temps théorique ACOUS ELYT]);sum('Heures disponibles centre charge'[Heures disponibles (comprend les effectifs réels)]);DATESINPERIOD('Date période'[Date];max('Date période'[Date]);-6;DAY))

Thanks for your help.

Karim

[Anonymous]

When you calculate an average from a sequence of numbers and want to ignore 0's, you have to just sum up all the numbers (0's don't contribute anything) as they are and find out how many non-0's you have in the sequence, then divide by the count. Quickly, a measure that would give you such an average going 5 days back (as you want to have the last 6 days counting the one you're on right now) would be:

// Calendar must be the Date table in your model.
// Slicing can only be done through dimensions.
// All columns in your fact table must be hidden.

// Base measure
[Total Hours] = SUM( FactTable[Hours] )

[6-Day Avg] =
var __totalDays = 6
var __currentDay = lastdate( Calendar[Date] )
var __firstDay =
	dateadd( __currentDay, -(__totalDays - 1), day)
var __periodToAvgOver =
	datesbetween(
		Calendar[Date],
		__firstDay,
		__currentDay
	)
var __sum =
	calculate(
		[Total Hours],
		__periodToAvgOver
	)
var __countOfNonZeroDays =
	sumx(
		__periodToAvgOver,
		1 * ( [Total Hours] > 0 )
	)
var __avg =
	divide( __sum, __countOfNonZeroDays )
return
	__avg

Take the above, think it over and adjust to the problem at hand.

Best

D

[Anonymous]

Hello,

@az38 you're right, my result is false. For every record, I need to do Temps Théorique / Heures disponibles. Then I want to make an average of it on last 6 days, without 0 values. So in this example, I should divide by 4 records. Hope one more time I'm clear 🙂

@Anonymous Thanks for your answer ! Your code seems to be good, I'll validate it with the client, and come back to you ASAP.

Thank you very much guys !

Karim

[Anonymous]

Hi @Anonymous,

I come back to you, because I did a mistake in my explanation of the problem.

You will find below an example :

Date	A	B	A/B (%)	Avg -6 OK	Avg -6 NOT OK
04/10/2019	100	200	50
05/10/2019
06/10/2019
07/10/2019	120	200	60
08/10/2019	140	200	70
09/10/2019	160	200	80
10/10/2019	180	200	90
11/10/2019	150	200	75	70,83	75
12/10/2019

In fact, the client needs to know the avg of the 6 last records <> 0.

On the 11th of october, if I do -6, I should stop on 6th, but as it's empty, I take the 5th, which is empty, so I need to take the 4th, which is OK.

Is it possible to handle it ? Because in fact, all the date functions work on calendar date, without handling if the line corresponds to the wanted filters.

Hope it's clear for you 🙂

Thanks for your help.

Karim

[Anonymous · Accepted solution]

[6LastNon0Entries Avg] =
var __non0EntriesCount = 6
var __currentDay = max( Calendar[Date] )
var __daysWithNonZeros =
	filter(
		all( 'Calendar'[Date] ),
		'Calendar'[Date] <= __currentDay
		&& [Total Hours] > 0
	)
var __6DaysWithNonZeros =
	topn(__non0EntriesCount,
		__daysWithNonZeros,
		'Calendar'[Date],
		desc
	)
var __sum =
	calculate(
		[Total Hours],
		__6DaysWithNonZeros
	)
var __avg =
	divide( __sum, __non0EntriesCount )
var __6DaysWithNonZerosCount =
	countrows( __6DaysWithNonZeros )
var __shouldCalc =
	__6DaysWithNonZerosCount = __non0EntriesCount
return
	if( __shouldCalc, __avg )

Best

D

[Anonymous]

Here's a file with the working measure.

https://1drv.ms/u/s!ApyQEauTSLtOgYMnsGk__jeAJkvMEA?e=5gKPk2

Best
D

[Anonymous]

@Anonymous  thank you very much !

it seems to be the expected result. I'll validate it with the client, and come back to you ASAP.

[Anonymous]

Don't forget to give me Kudos 🙂

Best
D

[Anonymous]

Hi @az38 

Thanks for your answer.

You'll find below client data examples.

Date	Temps théorique	Heures disponibles (comprend les effectifs réels)	Gamme théorique/charge (6jours glissant) %
05/10/2019 00:00			55.71
06/10/2019 00:00			46.73
07/10/2019 00:00	98.98653099999994	195.999998	52.71
08/10/2019 00:00	113.21901100000008	195.999998	56.06
09/10/2019 00:00	121.79969699999995	195.999998	58.52
10/10/2019 00:00	86.29423900000013	195.999998	52.37

But Gamme théorique/charge (6jours glissant) % is not OK.

The client needs to calculate on "6 days before"...

Hope the need is clear 🙂

Thanks. 

[az38]

@Anonymous whats the expected values for your example?

[Anonymous]

@az38 

When you calculate the average of the last column, you should have 36,61, instead of 53,683.

[az38]

Sorry, @Anonymous 

I can not understand how do you get 36% from this data 🤔

[az38]

Hi @Anonymous 

Please, give your data example.

I dont see any hint about any field in your measure.

What is Notes?