Let's just focus on MAU for this week's report. Relevant query is
SELECT
SUBSTR(bucket_id, 6, 7) AS month,
COUNT(DISTINCT IF(
product_id = 'a21b31b4-74b8-4a6e-a2a8-28b85d1f8315'
AND merchant_name != 'Billbox'
AND merchant_name != 'Simpl', user_id, NULL)) AS pl_users,
COUNT(DISTINCT IF(merchant_name = 'Billbox', user_id, NULL)) AS billbox_users,
COUNT(DISTINCT IF(
product_id = '541eeb1c-8212-40f7-af04-05990ff00961',
user_id, NULL)) AS pi3_users,
COUNT(DISTINCT user_id) AS total_users
FROM
public.transaction_events
WHERE
event_name = 'TransactionChargedEvent'
AND bucket_id BETWEEN 'date_2024_01_01' AND 'date_2025_08_31'
GROUP BY 1
ORDER BY 1
| Month | PL Users | Billbox Users | PI3 Users | Total Users |
|---|---|---|---|---|
| 2024-01 | 1,638,670 | 251,549 | 23,320 | 1,696,915 |
| 2024-02 | 1,677,891 | 253,272 | 22,343 | 1,735,296 |
| 2024-03 | 1,759,024 | 271,397 | 27,101 | 1,819,797 |
| 2024-04 | 1,816,902 | 268,842 | 28,186 | 1,876,102 |
| 2024-05 | 1,557,680 | 264,085 | 24,172 | 1,609,186 |
| 2024-06 | 1,442,627 | 264,281 | 25,310 | 1,493,922 |
| 2024-07 | 1,381,548 | 262,008 | 22,928 | 1,432,833 |
| 2024-08 | 1,339,141 | 257,472 | 19,996 | 1,388,459 |
| 2024-09 | 1,309,829 | 266,733 | 24,316 | 1,361,489 |
| 2024-10 | 1,347,160 | 266,124 | 26,716 | 1,398,514 |
| 2024-11 | 1,353,713 | 270,544 | 27,454 | 1,405,364 |
| 2024-12 | 1,385,643 | 271,830 | 75,842 | 1,451,483 |
| 2025-01 | 1,385,597 | 243,336 | 57,524 | 1,443,728 |
| 2025-02 | 1,366,777 | 224,554 | 68,834 | 1,435,293 |
| 2025-03 | 1,390,820 | 223,869 | 92,076 | 1,471,725 |
| 2025-04 | 1,350,550 | 217,312 | 72,194 | 1,423,517 |
| 2025-05 | 1,338,442 | 231,519 | 106,479 | 1,431,227 |
| 2025-06 | 1,321,566 | 231,283 | 144,161 | 1,435,622 |
| 2025-07 | 1,341,391 | 236,529 | 179,651 | 1,468,219 |
| 2025-08 | 1,379,142 | 231,203 | 195,791 | 1,513,038 |
PL MAU showed a growth after several months, reaching 5 month high. Still far away from life time max reached in April 2024 (but unit economics is better now). Pi3 MAU has continued sustained increase.
Here are PL delinquencies (DPD35)
And here is Pi3 collection view of delinquencies (DPD35) - slight degradation in July.
Collection team is calling their new collection strategies as Alpha - Carinae which I see is another name of Canopus.
Thanks to Sagar, Shivam, Salman, Harshdeep, Abhishek, Taqee - we are moving really fast on canopus collections. Specific focus is on D300+ collections, and I discussed exact approaches in last week update. Following is daily collection graph of canopous. We have colllected 1.13 cr in first six days of September, compared to 1.09 cr in all of August.
Identity service database was deleted last week. Thanks Vinayak for being persistent and everyone else for doing their part - Shivam, Vikas, Akhil, Krishna and anyone else whom I may have missed.
You can view this talk by Sheekha Verma for a trip to history lane. Sheekha was the leader of our DS team around 2 years ago.
Sveltekit migration has been ramped to 100%. Some merchants are seeing a bit success rates drop, but they are relatively minor and we are trying to debug.
Expect slightly lower delinquencies because of this in coming times.
In CRM, there is a concept of "tracked events". Id of a processed event is put in some DB to guard against duplicate processing (I think). Earlier this was in postgres and that table was becoming big. Now we have moved it to redis, and also reduced ttl, and also stopped processing needless elements. Shoan was pointing out that a date partitioned postgres would be better since then large volume of events only degrades performance and does not need an upgrade. Anyway, we have done this, and later Vikas and team the new owners, will take care.
Last Friday was Yatin's last day in office. He is away to Delhi because of medical emergency in family. We will ask him to come to office once when he is back to Bangalore. We wish Yatin all the best.
Akshay will be managing the payments team from now on. Thanks Akshay for stepping up.
Now direct integrations can also be routed to SIKA via an adapter that Prakash/Ayush have written.
We have implemented in which we route linked SIKA users to orchestrator and non linked SIKA users to split pay service. This is a temporary change till the time we figure out the cause of latency problems in orechestrator.