Field Collections • Short Decision Version

Run one matched CDMX 31-60 benchmark, keep Kobra focused on 91-120, and use the result to decide where internal scales next.

The page makes one clean argument: launch CDMX 31-60 with internal and Kobra side by side, run Kobra Pilot 4 as a multi-area 91-120 package with Jalisco as the anchor plus León, Monterrey, and Querétaro, and use the result to decide how broadly late-bucket vendor scale should continue. The evidence suggests 10k-30k debt is the best repeat zone in 31-60, while 30k+ total debt is also a credible late-bucket target in 91-120. If internal wins the CDMX benchmark, the next full internal scale move is Jalisco 31-60.

Launch: CDMX 31-60 with internal + Kobra Pilot 4: multi-area 91-120 with Kobra Anchor: Jalisco Expansion legs: León, Monterrey, Querétaro Next internal scale if benchmark works: Jalisco 31-60 Do not scale: broad CDMX 91-120

Key decision message

Field collections should be scaled as a selective portfolio: internal where dense early-bucket routing can beat vendor economics, and Kobra where the late-bucket case is already proven or where the bank still needs evidence. In late collections, the evidence is strongest when the portfolio is targeted rather than broad, with 30k+ total debt now emerging as the clearest high-value screen to test first.

+3.96 pp CDMX 31-60 cure lift in Pilot 3.

4 boroughs CDMX 31-60 core pack with positive gross field economics.

91-120 Recommended Kobra Pilot 4 bucket across the multi-area package.

10k-30k / 30k+ 10k-30k is the best repeat zone in 31-60; 30k+ is the clearest 91-120 late-bucket opportunity.

1. Decision

Just the gist

What we are saying

Move	Bucket	Where	Why
Launch matched benchmark	31-60	CDMX core borough pack, with internal + Kobra	Not the single strongest pure profit pool, but the best benchmark market: dense routing, close operational oversight from headquarters, and the clearest place to test whether internal can reduce cost-to-collect while preserving recovery effectiveness.
Run Kobra Pilot 4	91-120	Jalisco core + León + Monterrey + Querétaro	Jalisco is the profit anchor; the other cities broaden the read on where late-bucket vendor economics still work, with 30k+ total debt as the clearest screen for where to start.
Scale next if internal wins	31-60	Jalisco	Best current region for the next full internal cell.

2. Pilot Proof

Pilot-level proof first; regional drill-down hidden under each pilot

What the three pilots proved

Gross field economics and final delta should be read separately. Field recovery and field spend come from Kobra closeout. Baseline recovery and final delta come from the delta workbook, so both layers are shown here.

Pilot	Bucket	Control cure rate	Kobra cure rate	Cure uplift	Cure sig.	Field recovery	Field spend	Gross field result	Calls & notifications recovery	Final delta	Pre-Due COR diff	Pre-Due Net Yield diff	Read
Pilot 1	91-120	1.56%	3.73%	+2.16 pp	Sig. at 95%	MXN 357,609	MXN 220,000	MXN 137,609	MXN 115,559	MXN -3,595	—	—	Roughly breakeven after baseline subtraction.
Pilot 2	91-120	1.05%	2.14%	+1.09 pp	Sig. at 95%	MXN 541,144	MXN 400,000	MXN 141,144	MXN 300,495	MXN -141,802	—	—	Broad late-bucket vendor scale is not supported.
Pilot 3	31-60	9.87%	13.54%	+3.67 pp	Sig. at 95%	MXN 580,145	MXN 326,012	MXN 254,133	MXN 731,912	MXN 38,400	+$391,522 Not sig.	+$134,155 Not sig.	Best combined proof for selective rollout; broader portfolio quality metrics also moved in the right direction, but without statistical significance.

Cure significance is shown as a two-proportion test on the Kobra group versus the baseline group used in the cure table. Pilot 3 portfolio metrics are shown as Kobra on-top minus control in the pre-due window.

Cure-rate difference by total debt bandShow tables

Pilot 1 • 91-120 cure by 120 dpd
Total debt band	Control clients	Kobra clients	Control cure rate	Kobra cure rate	Diff	Sig.
5k-10k	319	520	2.19%	4.04%	+1.84 pp	Not sig.
10k-15k	248	223	1.61%	2.69%	+1.08 pp	Not sig.
15k-20k	133	112	0.75%	2.68%	+1.93 pp	Not sig.
20k-30k	193	165	1.55%	3.64%	+2.08 pp	Not sig.
30k+	252	132	0.79%	3.79%	+2.99 pp	Sig. at 95%

Pilot 2 • 91-120 cure by 120 dpd
Total debt band	Control clients	Kobra clients	Control cure rate	Kobra cure rate	Diff	Sig.
5k-10k	1439	1415	0.63%	2.12%	+1.49 pp	Sig. at 99%
10k-15k	597	585	1.01%	1.71%	+0.70 pp	Not sig.
15k-20k	338	319	1.18%	1.88%	+0.70 pp	Not sig.
20k-30k	480	485	1.67%	1.86%	+0.19 pp	Not sig.
30k+	378	414	0.53%	2.17%	+1.64 pp	Sig. at 95%

Pilot 3 • 31-60 cure by 60 dpd
Total debt band	Control clients	Kobra clients	Control cure rate	Kobra cure rate	Diff	Sig.
5k-10k	845	799	11.60%	14.27%	+2.67 pp	Not sig.
10k-15k	319	335	8.46%	15.22%	+6.76 pp	Sig. at 99%
15k-20k	162	172	9.26%	15.12%	+5.86 pp	Not sig.
20k-30k	240	265	7.50%	10.94%	+3.44 pp	Not sig.
30k+	212	207	7.08%	9.18%	+2.10 pp	Not sig.

Read across the three pilots: in 91-120, the strongest statistically supported total-debt band is often 30k+; in 31-60, the strongest clean signal is 10k-15k, with broader directional strength across 10k-30k.

Pilot 1 • 91-120 • regional splitShow details

Region	Field recovery	Field spend	Cure lift	Final delta	Read
Jalisco	MXN 109,730	MXN 42,822	+3.63 pp	MXN 46,850	Best region in Pilot 1.
Ciudad de México	MXN 175,369	MXN 102,976	+2.87 pp	MXN 15,456	Positive, but less efficient than Jalisco.
Estado de México	MXN 45,084	MXN 43,359	-0.98 pp	MXN -26,942	Not attractive.
Nuevo León	MXN 27,426	MXN 30,843	+1.75 pp	MXN -10,671	Cure moved, economics did not.

Pilot 2 • 91-120 • regional splitShow details

Region	Field recovery	Field spend	Cure lift	Final delta	Read
Chihuahua	MXN 70,398	MXN 37,157	+2.26 pp	MXN 51,890	Strongest current vendor-backed case.
Jalisco	MXN 138,244	MXN 82,380	+1.93 pp	MXN 31,300	Best anchor geography for Pilot 4.
Ciudad de México	MXN 192,444	MXN 170,518	+0.51 pp	MXN -80,688	Too weak for wave 1.
Nuevo León	MXN 86,628	MXN 61,388	+1.36 pp	MXN -28,608	Gross-positive, net-negative.
Puebla	MXN 28,650	MXN 28,591	+0.54 pp	MXN -16,466	Not attractive.
Querétaro	MXN 19,374	MXN 13,857	-1.06 pp	MXN -389	Weak actual proof.
Estado de México	MXN 5,406	MXN 5,662	+4.88 pp	MXN -256	Too small to use as scale proof.
Chiapas / Morelos / Quintana Roo	~MXN 0	~MXN 149 each	n/a	Not in delta workbook	Ignore for decision-making.

Pilot 3 • 31-60 • regional splitShow details

Region	Field recovery	Field spend	Cure lift	Final delta	Read
Jalisco	MXN 168,317	MXN 83,887	+4.14 pp	MXN 38,413	Best region to replicate after the benchmark.
Ciudad de México	MXN 250,975	MXN 137,676	+3.96 pp	MXN 420	Best benchmark region, not yet a broad-scale case.
Estado de México	MXN 160,854	MXN 104,449	+2.97 pp	MXN -79,996	Operationally promising, economically weak.

3. Funnel and Targeting

Interactive field funnel plus the debt and limit insight that changes targeting

Operational funnel

This is the tighter funnel, from visit to true direct contact to PTP. Reported contact is intentionally omitted here because it is too loose for the bigger picture.

Visits: 100% 6,536 visits

Localized 86.72%

Access 61.20%

Direct debtor contact 5.34%

PTP 1.67%

Default view shows all pilots and all regions combined.

Revisit read

The raw field feed still supports fast revisits. The cleanest signal is Pilot 3, where 4-7 day revisits materially outperformed first visits on direct contact and PTP.

Pilot 3 first visit: 3.26% direct debtor contact and 1.54% PTP.
Pilot 3 revisit in 4-7 days: 21.26% direct debtor contact and 4.72% PTP.
Implication: revisit discipline should be part of rollout, not a later add-on.

CDMX targeting read

CDMX looks attractive in 31-60, but not in 91-120. The Snowflake cut says the best CDMX field targets are in the 10k-30k zone.

CDMX 31-60 best debt bands: 10k-20k and 20k-30k; avg recovery uplift vs control is about +MXN 368 and +MXN 642.
CDMX 31-60 best limit bands: 10k-20k and 20k-30k; avg recovery uplift vs control is about +MXN 272 and +MXN 392.
CDMX 91-120 read: the better late-bucket bands exist, but the city-level economics still do not justify broad launch.
Rollout implication: use MXN 5k+ as the floor, but prioritize 10k-30k debt and limit inside the CDMX benchmark.

4. Economics

Only the model comparisons and launch tables that change the decision

Why CDMX 31-60 is still worth launching

Benchmark market

CDMX is not the highest-conviction immediate profit pool, but it is the most practical market in which to validate the internal model under close oversight from headquarters.

Attractive core areas

Benito Juárez, Coyoacán, Cuauhtémoc, and Venustiano Carranza were all gross-positive in the planning pack.

Targeted 31-60 proof

CDMX already showed a +3.96 pp cure lift in 31-60, and the best target mix is in the 10k-30k debt and limit zone.

Internal team economics

The current internal model uses one field cell with 10 agents + 1 supervisor. These are the economics that drive the benchmark math.

Cost component	Formula	Monthly amount
One field agent	(MXN 10,000 fixed + MXN 12,000 variable) × 1.20 burden	MXN 26,400
Ten field agents	10 × MXN 26,400	MXN 264,000
One supervisor	MXN 30,000 × 1.20 burden	MXN 36,000
Total internal 10+1 team	MXN 264,000 + MXN 36,000	MXN 300,000

The internal model shown here reflects payroll and labor burden; transport, devices, recruitment, QA, insurance, and other operating overhead may add further cost.

CDMX internal benchmark economics

This remains the cleanest internal proof because it explicitly shows all three parts of the delta: field recovery, baseline recovery, and field spend.

Visits / day	Cost / visit	Field recovery	Calls & notifications recovery	Spend	Final delta
8	MXN 170.45	MXN 398,606	MXN 158,552	MXN 300,000	MXN -59,946
10	MXN 136.36	MXN 498,257	MXN 198,189	MXN 300,000	MXN 68
12	MXN 113.64	MXN 597,909	MXN 237,827	MXN 300,000	MXN 60,081

If internal only matches Kobra in CDMX 31-60

Internal still wins economically because the cost base is lower.

Visits / day	Field recovery	Calls & notifications recovery	Internal spend	Kobra spend	Internal delta	Kobra delta	Internal advantage
10	MXN 498,257	MXN 198,189	MXN 300,000	MXN 407,000	MXN 68	MXN -106,932	MXN 107,000
12	MXN 597,909	MXN 237,827	MXN 300,000	MXN 488,400	MXN 60,081	MXN -128,318	MXN 188,400

Benchmark success criteria

The benchmark should be treated as successful only if all three conditions are met.

Economic win: final delta is above zero.
Operating win: field productivity is sustained at or above the modeled level used in the benchmark case.
Risk win: conduct outcomes and complaint levels remain acceptable under internal execution.

What to launch internally

This table is meant to be decision-grade: areas, bucket, and monthly scenario range only.

Launch	Bucket	Areas	Scenario	Monthly field recovery	Monthly calls & notifications recovery	Monthly spend	Monthly delta
CDMX benchmark	31-60	Benito Juárez, Coyoacán, Cuauhtémoc, Venustiano Carranza	Negative case: 8 visits/day	MXN 398,606	MXN 158,552	MXN 300,000	MXN -59,946
CDMX benchmark	31-60	Same core pack, internal + Kobra matched	Base case: 10 visits/day	MXN 498,257	MXN 198,189	MXN 300,000	MXN 68
CDMX benchmark	31-60	Same core pack, high productivity	Positive case: 12 visits/day	MXN 597,909	MXN 237,827	MXN 300,000	MXN 60,081
Next full internal cell	31-60	Jalisco cluster	Pilot-backed run-rate	~MXN 1.12M	~MXN 0.58M	~MXN 0.36M	~MXN 0.18M

Bonus idea: concentrated internal late-bucket launch in the top 3 states

As of 2026-04-21, the current 91-120 dpd, P_CURRENT, 30k+ total debt pool is concentrated enough in a few states to treat as a separate internal thought-starter. The recovery columns below use the observed Kobra 30k+ per-client field recovery from Pilots 1-2 as a simple gross-recovery proxy.

State	Clients	Cost @ 10 visits/day	Cost @ 12 visits/day	Total debt	Median total debt	Potential field recovery using Pilot 1 30k+ benchmark	Potential field recovery using Pilot 2 30k+ benchmark
Ciudad de México	250	MXN 34,090	MXN 28,410	MXN 11,136,023	MXN 40,992	MXN 92,110	MXN 88,753
Estado de México	176	MXN 23,995	MXN 19,999	MXN 7,878,165	MXN 39,750	MXN 64,845	MXN 62,482
Jalisco	131	MXN 17,863	MXN 14,887	MXN 5,786,022	MXN 41,057	MXN 48,266	MXN 46,506
Total top 3	557	MXN 75,948	MXN 63,296	MXN 24,800,210	MXN 39,750 to 41,057	MXN 205,221	MXN 197,741

This is a first-pass one-visit cost view only. It is useful as a concentration screen, not yet as a full launch case. The recovery columns are gross field-recovery proxies only; they do not subtract baseline calls-and-notifications recovery, and a real launch decision would still need density, address quality, municipality clustering, and revisit assumptions.

What to launch with Kobra

Pilot 4 should stay in 91-120, but it can cover several areas. The cleanest way to frame it is as one multi-area package with Jalisco as the economic anchor and León, Monterrey, and Querétaro as additional test zones.

Launch	Bucket	Areas	Scenario	Monthly field recovery	Monthly calls & notifications recovery	Monthly spend	Monthly delta
Continue	91-120	Chihuahua / Juárez pockets	Current proven continuation	~MXN 0.76M	~MXN 0.09M	~MXN 0.29M	~MXN 0.38M
Pilot 4 anchor	91-120	Guadalajara + Zapopan, optional San Pedro Tlaquepaque	Pilot-backed range	~MXN 0.46M to ~MXN 0.80M	~MXN 0.07M to ~MXN 0.09M	~MXN 0.27M to ~MXN 0.34M	~MXN 0.10M to ~MXN 0.39M
Pilot 4 test zone	91-120	León	Proxy-positive	~MXN 0.10M	~MXN 0.02M	~MXN 0.06M	~MXN 0.02M
Pilot 4 test zone	91-120	Monterrey	Proxy-positive	~MXN 0.34M	~MXN 0.18M	~MXN 0.11M	~MXN 0.05M
Pilot 4 test zone	91-120	Querétaro	Weak proof / controlled test	~MXN 0.19M	~MXN 0.05M	~MXN 0.14M	~MXN 0.00M
Pilot 4 package total	91-120	Jalisco + León + Monterrey + Querétaro	Package view	~MXN 1.09M to ~MXN 1.43M	~MXN 0.32M to ~MXN 0.34M	~MXN 0.58M to ~MXN 0.65M	~MXN 0.17M to ~MXN 0.46M

5. Improvement Levers

Only the levers that change economics or conduct risk materially

What should be part of rollout

Address enrichment and prioritization should be prerequisites; they directly improve route quality and reduce wasted visits.
Fast revisits should be built into operations from day one, especially in 31-60.
Debt and limit targeting should prioritize 10k-30k, not just the very top balances.
Image sharing should stay in rollout because it strengthens QA and complaint handling.
Recent contact suppression should be tested so field is used where remote channels are less likely to cure the account anyway.

6. Next Steps

What to do now, and what happens if the internal benchmark works

Immediate actions

Launch CDMX 31-60 benchmark with internal and Kobra in matched carve-outs under the same targeting and reporting rules.
Run Kobra Pilot 4 as a 91-120 package: Jalisco core first, then León, Monterrey, and Querétaro as smaller expansion legs.
Continue Chihuahua 91-120 separately as the cleanest already-proven continuation case.
Apply rollout levers immediately, not after the benchmark.

If CDMX internal works

Step 1: expand within the reserve CDMX boroughs before opening a second city.
Step 2: open the next full internal cell in Jalisco 31-60.
Step 3: only then look at adjacent Estado de México dense metros.
Rule: do not treat benchmark success in 31-60 as automatic approval to scale 91-120 internally.

Method note: cure is based on cure_report, not payment > 0. Pilot 3 uses order_dpd_mature <= 60. Pilots 1-2 use order_dpd_mature <= 120. Field recovery and field spend come from Kobra closeout. Baseline recovery and final delta come from the delta workbook where available.