Advanced Calculated Metrics - Standard Deviation Revenue Per Visit (Omniture)

Avatar

Avatar
Boost 5
Level 2
andrews2719625
Level 2

Likes

5 likes

Total Posts

3 posts

Correct reply

0 solutions
Top badges earned
Boost 5
Boost 3
Boost 1
View profile

Avatar
Boost 5
Level 2
andrews2719625
Level 2

Likes

5 likes

Total Posts

3 posts

Correct reply

0 solutions
Top badges earned
Boost 5
Boost 3
Boost 1
View profile
andrews2719625
Level 2

08-05-2019

I would like to build a calculated metric to get the sample standard deviation of Revenue per visit in Omniture as it is required to calculate statistical confidence. Ideally, this metric would make it so I would not have to pull from data warehouse to calculate confidence intervals and could do this in real time.

Is this possible?

Accepted Solutions (1)

Accepted Solutions (1)

Avatar

Avatar
Validate 1
Level 4
brentrad
Level 4

Likes

54 likes

Total Posts

82 posts

Correct reply

16 solutions
Top badges earned
Validate 1
Contributor
Shape 1
Give Back
Boost 50
View profile

Avatar
Validate 1
Level 4
brentrad
Level 4

Likes

54 likes

Total Posts

82 posts

Correct reply

16 solutions
Top badges earned
Validate 1
Contributor
Shape 1
Give Back
Boost 50
View profile
brentrad
Level 4

09-05-2019

Answers (6)

Answers (6)

Avatar

Avatar
Boost 3
Level 1
joshuaa33980431
Level 1

Likes

3 likes

Total Posts

2 posts

Correct reply

0 solutions
Top badges earned
Boost 3
Boost 1
View profile

Avatar
Boost 3
Level 1
joshuaa33980431
Level 1

Likes

3 likes

Total Posts

2 posts

Correct reply

0 solutions
Top badges earned
Boost 3
Boost 1
View profile
joshuaa33980431
Level 1

06-09-2019

Hey Andrey -

Scenario is evaluating the confidence level of the difference in RPV observed in an A/B test. One of the required metrics in the excel file from Adobe is sum of revenue squared. I'm trying to find a way to do this in workspace without needing to request a DW report every time. This gets time consuming when we are trying to evaluate the RPV for multiple tests. Any thoughts?

Avatar

Avatar
Boost 5
Level 2
andrews2719625
Level 2

Likes

5 likes

Total Posts

3 posts

Correct reply

0 solutions
Top badges earned
Boost 5
Boost 3
Boost 1
View profile

Avatar
Boost 5
Level 2
andrews2719625
Level 2

Likes

5 likes

Total Posts

3 posts

Correct reply

0 solutions
Top badges earned
Boost 5
Boost 3
Boost 1
View profile
andrews2719625
Level 2

09-05-2019

I've gone through all of that documentation in great depth, but haven't had any success using the Std. Dev function to get a sample's std. deviation at a visit level.

When I compare two samples, it either shows the same value or zeros for both.

Avatar

Avatar
Boost 5
Level 2
andrews2719625
Level 2

Likes

5 likes

Total Posts

3 posts

Correct reply

0 solutions
Top badges earned
Boost 5
Boost 3
Boost 1
View profile

Avatar
Boost 5
Level 2
andrews2719625
Level 2

Likes

5 likes

Total Posts

3 posts

Correct reply

0 solutions
Top badges earned
Boost 5
Boost 3
Boost 1
View profile
andrews2719625
Level 2

25-09-2019

Hey Andrey,

Jumping back in here, the easy answer for desire to do this in AA vs. Target UI is that the Adobe Analytics UI is just outright more user friendly for monitoring and analysis purposes, and some basic expectations are not met.

Some examples:

  • In using target, we lose the ease of applying multiple segments and instead either have to build a combined segment or open the "View in analytics" which is again, less user friendly.
    • i.e if I wanted to segment by device and only look at the segment of users that viewed a certain page in their session I would have to build a segment combining those two to directly apply in target
  • Target does not calculate confidence or lift on custom built metrics
    • i.e. If I am curious about revenue net of order level discounts, I lose the ability to see confidence in target, requiring me to either pull from Omnihit manually or submit a DW request.
    • The ability to build STD Dev. Revenue Per Visitor or visit into my analytics workspace would mean I could quickly plug in rates for a sample and calculate confidence without having to go through these processes

These are, to my understanding, features on the road map for AA but I would like to reiterate that their current absence is quite frustrating in A/B test analysis/monitoring and their addition would be greatly appreciated.

Avatar

Avatar
Boost 3
Level 1
joshuaa33980431
Level 1

Likes

3 likes

Total Posts

2 posts

Correct reply

0 solutions
Top badges earned
Boost 3
Boost 1
View profile

Avatar
Boost 3
Level 1
joshuaa33980431
Level 1

Likes

3 likes

Total Posts

2 posts

Correct reply

0 solutions
Top badges earned
Boost 3
Boost 1
View profile
joshuaa33980431
Level 1

06-09-2019

Just want to bump this up. I have the same question and have not been able to find a solution. The idea of having to resort to DW every time seems a bit inefficient.

Avatar

Avatar
Shape 25
MVP
Andrey_Osadchuk
MVP

Likes

538 likes

Total Posts

973 posts

Correct reply

264 solutions
Top badges earned
Shape 25
Bedrock
Coach
Boost 500
Affirm 250
View profile

Avatar
Shape 25
MVP
Andrey_Osadchuk
MVP

Likes

538 likes

Total Posts

973 posts

Correct reply

264 solutions
Top badges earned
Shape 25
Bedrock
Coach
Boost 500
Affirm 250
View profile
Andrey_Osadchuk
MVP

06-09-2019

The question is why do you want to evaluate that in AA rather in the the A/B test UI?

P.S.: Square and Square Root can be built in the Metric Builder.

Avatar

Avatar
Shape 25
MVP
Andrey_Osadchuk
MVP

Likes

538 likes

Total Posts

973 posts

Correct reply

264 solutions
Top badges earned
Shape 25
Bedrock
Coach
Boost 500
Affirm 250
View profile

Avatar
Shape 25
MVP
Andrey_Osadchuk
MVP

Likes

538 likes

Total Posts

973 posts

Correct reply

264 solutions
Top badges earned
Shape 25
Bedrock
Coach
Boost 500
Affirm 250
View profile
Andrey_Osadchuk
MVP

06-09-2019

Revenue is one of the metrics that does not follow normal distribution (aka bell curve). This is why the StdDev of RPV may not make any sense.

Could you share what scenario you are working on? — if you are running an A/B test, Target (or other tools) should have all required metrics in their UI.