NPS network. how to deal with 1 nps score that visit multiple cities /level of the data

[Hansdewit]

I am starting an analysis for NPS Network (mobile network). I am working for a Telco company.

every day we are sending some nps question to our customer to measure the quality of our Mobile network.  so the focus of the analysis is to improve the Mobile network (locally). the customer give consent to collect 7 days of the network data before the survey . So a customer can be in multiple cities during the 7 days.

Normally with NPS analysis , I have the data of customer level. 

should aggregate the data on for example citie level and collect all the nps scores. but people are not 100% of the time in 1 city, so Should I make a weight.

I have a problem with the idea having 1 nps score in multiple cities.

of course aggregate the data on customer level, and use 'number of cities'.  But then I don't know which cities. 

anyone experience with these kind of  analysis?

[sbxkoenk]

Hello @gsvolba ,

I have no idea how familiar you are with NPS (Net Promoter Score) calculation in Telecommunications industry.

(I have never worked with NPS concept myself by the way)

But this question (of @Hansdewit ) made me think about this Communities Library article of yours :

Efficient “One-Row-per-Subject” Data Mart Construction for Data Mining
Posted 01-23-2022 04:46 PM | by gsvolba (2508 views)
https://communities.sas.com/t5/SAS-Communities-Library/Efficient-One-Row-per-Subject-Data-Mart-Const...

Any idea how to tackle this NPS question with a kind of 'spatial scale'?

Thanks,

Koen

[gsvolba]

(Sorry for the delayed reply! the last two weeks have been too full with projects + part-time-teaching, so I overlooked this as the automatic forwarding did not work)

I do not have particular experience with NPS in multiple locations as described here. But I think some other ideas can be generalized to solve this.

your approach with aggregating sounds feasilbe, some ideas that pop up in my mind

- do you want to end up with ONE value per customer (aggregated over days + cities) or should/could it be a value per city 

- cities could be ordered by weight, e.g. the top 3 cities with the longest duration

- you can also calculate a variability measure/segement : A: customers who only have one city, customers with >= 2 cities with rather equal time there, customer with a dominant city

just some quick thoughts 

happy to discuss further
Gerhard