Field visits: where to launch, where to stop, and what Pilot 4 should test

The corrected pilot close-out sheets show that Kobra generated real recovery and real cure uplift, but the quality of that result varies sharply by geography and bucket. The best scale pools remain Jalisco 31–60, selected Estado de México 31–60, and Chihuahua 91–120. The strongest CDMX answer is a targeted internal test, not citywide rollout.

Recommended borough pack for the first 10+1 internal test month: Benito Juárez, Coyoacán, Cuauhtémoc, and Venustiano Carranza. Combined 31–60 debt pool: MXN 26.69m. This creates a cleaner proof design than citywide CDMX.

• That borough selection improves the citywide CDMX average.
• That the internal team can sustain at least 10 visits/day.
• That higher-balance prioritization lifts incremental recovery enough to move beyond breakeven.

• 8 visits/day: not enough for CDMX internal test economics.
• 10 visits/day: roughly breakeven on the conservative benchmark.
• 12 visits/day: positive internal case.
• Operational implication: routing discipline is as important as the pilot geography.

The regional opportunity file shows very large debt pools in 61–90, 121–180, 181–240, and especially 241+. The right interpretation is not “scale now”; it is opportunity sizing with explicit uncertainty. The next logical adjacency is 61–90 in the same dense metros where 31–60 works. The next logical vendor experiment beyond Pilot 4 is 121–180 in selected metros. 241+ is large, but should be treated as a separate workstream because restructuring / legal strategy starts to matter materially there.

• 61–90: closest to the proven 31–60 internal case, so treat it as an internal adjacency bucket after the first CDMX / Jalisco internal test proves productivity.
• 121–180 and 181–240: late enough to justify a field experiment, but only in tight metro clusters and preferably through Kobra first.
• 241+: the debt pool is huge, but this is not a clean next field-collections decision. It should be paired with restructuring / legal segmentation, not rolled into the next field launch by default.

• Finish apartment / interior enrichment and use it in dispatch gating.
• Use borough-level confidence rules before assigning visits.

• Operationalize approved photo-sharing for field verification.
• Attach photo QA to supervisor review and complaint handling.

• Stop or shrink borough slices that remain negative after 150–200 visits.
• Stop low-performing agents after 40 visits if localization / direct contact / PTP fail threshold.

• One account-level table from assignment → visit → payment → cure.
• Keep PTP and later payment must be explicit fields.

• Same-day callback after useful third-party contact.
• Revisit within 48–72 hours, not weeks later.

• Launch only with 10+1, not 5+1.
• Track productivity daily by route and by municipality.

This is the clearest next Kobra question still worth paying for. It keeps the vendor in a late-stage bucket where field behavior matters, but removes broad-state dilution.