Timeline for How to design AB test for mean of means
Current License: CC BY-SA 4.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 20, 2022 at 11:10 | comment | added | kv123 | I have accepted the answer as It gives a lot of information but I wanted to leave a comment explaining in more detail the situation and our objective. I created a new question instead though, as it was getting to big for the comment so please have a look here | |
| Jul 20, 2022 at 8:05 | vote | accept | kv123 | ||
| Jul 18, 2022 at 10:50 | comment | added | kv123 | Yes, we do capture the number of payments anyway but the success(metric) of the campaign will be based on the mean of individual customers' collection rates. So, you are saying that this can be converted to individual customer payments without changing the statistics and the parameters of the test? | |
| Jul 18, 2022 at 10:24 | comment | added | kqr | If you multiple a customer's invidivual payment rate with the number of expected payments, don't you get the number of made payments? | |
| Jul 18, 2022 at 10:05 | comment | added | kv123 | I am bit confused though. In your simple analysis the data points are individual customer payments. However, in the scenario I described above, the data points are customers' individual collection rate calculated for a specific period of time. Is this correct? How does that impact the simple scenario in your answer? | |
| Jul 18, 2022 at 9:09 | comment | added | kqr | @kv123 then you can also do the simple analysis in my answer! | |
| Jul 18, 2022 at 9:05 | comment | added | kv123 | Yes, we can do that | |
| Jul 13, 2022 at 6:17 | comment | added | kqr | @kv123 so then if you have the collection rate for a customer and howany days they have been a customer, you can get out of it how many payments they have been expected to make and how many they made in time? | |
| Jul 11, 2022 at 11:00 | comment | added | kv123 | Customers are obliged to pay at least daily for the next many months. Simple scenario: 365 payments per customer per year. We have 15K customers. One customer can decide to pay in advance though - 1 big payment covering 7 days so they will do only 1 payment. | |
| Jul 11, 2022 at 8:48 | comment | added | kqr | @kv123 do you have information on how many payments are expected for each customer? Then you can go backwards to the simple analysis. Expected payments times collection rate for that customer equals payments made. | |
| Jul 11, 2022 at 8:32 | comment | added | kv123 | Thank you for the answer. It makes a lot of sense for the simple scenario. However, businesses offering financial products do not measure the Collection Rate this way(all the payments of all customers divided by the total expected payments of all customers). Instead, the Total Collection Rate is the average of each customer's individual collection rate(referred to as mean of means). The payment frequency concern that you mentioned is valid but not in our case, as every customer is expected to pay at least daily. For this metric, how sample size and duration change? | |
| Jul 8, 2022 at 9:58 | history | edited | kqr | CC BY-SA 4.0 |
added 648 characters in body
|
| Jul 8, 2022 at 9:29 | history | answered | kqr | CC BY-SA 4.0 |