Banco Plata Collections Board

Weekly Executive Dashboard

Mar 30 - Apr 5, 2026 · curated weekly view for MAB and Field Visits in a board-style HTML prototype.

2026-03-30_to_2026-04-05
0 Planning
3 In progress
0 Roll-out
0 On hold
0 Finished
Published
2026-04-09T21:27:00+00:00
Ready to trigger a new pipeline run.
MAB

MAB WhatsApp Optimization

Improve payment response versus a fixed notification strategy

In progress
Last week reward rate
6.8%
Full duration reward rate
9.6%
Last week recovered
$2,946,259
K=2 exit share
66.0%
Full Duration Snapshot
2026-03-18 to 2026-04-05
Group Clients Cure rate Recovered
A 18,164 58.6% $11,374,043
B 18,177 58.1% $10,989,732
B - A: -0.48 pp · p=0.357 · not statistically significant
Last Week Snapshot
2026-03-30 to 2026-04-05
Group Clients Cure rate Recovered
A 5,830 38.7% $2,493,552
B 5,803 37.6% $2,452,132
Key Takeaways
  • Weekly vs full duration: reward rate 6.8% vs 9.6%; recovered $2,946,259 vs $11,230,462.
  • Intensity shift: k=2 moved 64.6% -> 66.0% last week.
  • Event layer still looks diffuse because learning is split across many small segment-hour cells, not one global event pool.
Allocation share k=1 vs k=2
Bandit inclination toward k=2
A/B cumulative cure rate
Reward rate by intensity
FIELD VISITS

Field Visits - Kobra

Improve recovery in selected overdue segments through visits

In progress
Visits
359
Accounts visited
355
Access rate
62.1%
PTP amount
$31,047
Full Duration Snapshot
2026-03-01 to 2026-04-05
Group Clients Cure rate Recovered
A 1,783 7.9% $765,267
K 1,780 10.7% $974,231
K - A: +2.82 pp · p=0.004 · statistically significant
Last Week Snapshot
2026-03-30 to 2026-04-05
Group Clients Cure rate Recovered
A 1,691 2.2% $199,625
K 1,673 3.8% $294,170
Key Takeaways
  • Weekly execution: 359 visits, 355 accounts, 22 agents, 3 states.
  • Weekly vs full duration: located 83.3% vs 81.0%, access 62.1% vs 57.7%.
  • Business signal: PTP 6 for $31,047; K weekly cure 3.8% vs A 2.2%.
Operational funnel vs full test
Contact mix vs full duration
Visit outcome mix
Geography mix
AI IN_COLLECTIONS

AI in Collections

In progress
Description
This project is a process of trial and error on how to improve collections processes with the help of AI agents, LLM skills, etc.
Key Takeaways
  • built an automated weekly_update pipeline
  • met with SWE to clarify LLM usage
  • Irganized a meeting with Misha & Gosha on how to implement LLM & best practices
  • Talked to CDO about meetings on best practices
  • Developed a simple window to get insights on tests for BA/PM