LupsDunningDecide how to charge. Recover what fails.
Autonomous collections decision engine. Standalone or integrated with Lups.
LupsDunning is consulted before you charge: it decides the optimal amount, timing and channel for each customer, and your system executes the charge. If the charge fails, Dunning decides the next action — calibrated retries, omnichannel communication, custom payment plans or AI-agent negotiation — and persists until collecting 100% or until the deadline you set. Fully autonomous.
The real cost of charging blind
It's not just what fails — it's also what gets charged wrong. Charging the wrong amount, at the wrong time, through the wrong channel lowers your success rate even before the first failure. And between 20% and 48% of churn on recurring billing platforms is due to failed payments — not dissatisfied customers. These are people who want to keep paying, but the system didn't know how to charge them.
Without Dunning
With Dunning
How it works
LupsDunning acts as an autonomous decision engine. Before charging, your system asks it how to do it — Dunning returns the optimal amount, timing and channel. Your system executes. If the charge fails, your system asks again — Dunning decides the next action. The cycle repeats, learning from each interaction, until collecting 100%.
Dunning doesn't replace your billing system. It makes it smarter. It integrates in days and operates from before the first charge.
decisión IA
5 phases of intelligent collection
Each phase increases pressure intelligently. Only escalates when the previous one wasn't enough.
Phase 0 — Pre-charge decision
Before the first attempt, Dunning decides the optimal amount, the best timing (day, hour, payroll cycle) and the right channel for each customer. It predicts the probability of success and, if it's too low, recommends postponing. Result: +8–15% first-attempt success rate and fewer charges that cascade into failure.
First-attempt success: +8–15%Phase 1 — Silent & intelligent collection
Retries the charge invisibly to the customer. If they don't have the full amount, it probes how much they have and charges the maximum possible. Learns the best time to charge each person: day, hour and payroll cycle.
Estimated recovery: 38–57%Phase 2 — Decision engine: charge or communicate
After three failed attempts, decides between retrying or sending a message through the optimal channel (SMS, WhatsApp or email). Learns which channel each customer prefers and when they respond.
Additional recovery: +10–15%Phase 3 — Custom payment plan
When the customer can't pay all at once, offers a flexible plan. If an installment fails, Phase 1's silent collection activates for that specific installment without canceling the plan.
Additional recovery: +15–20%Phase 4 — AI conversational agent
For the toughest cases, an AI agent negotiates directly with the customer via WhatsApp. Understands context, handles objections, proposes plans and closes agreements. All within the limits you define.
Additional recovery: +10–15%Why choose Dunning?
Decide before, not just after
The only engine on the market consulted before the charge to decide amount, timing and channel — and after if it fails. It turns collections into a predictive operation, not a reactive one.
Learns from each customer
Builds a unique profile: when they have funds, which channel they respond to. Each charge feeds the model and makes it more accurate.
Protects the relationship
Respects legal contact hours in each Latin American country, frequency limits and the right tone to recover without damaging the experience.
Native to Latin America
Knows the payroll cycles of Chile, Mexico, Colombia, Brazil, Argentina and Peru. Knows when people have funds.
Fully configurable
You define the rules: max installments, minimum amounts, days between retries, channels to use. Dunning adapts to your commercial policy.
Clean API integration
Works as a black box. Your platform reports failed charges; Dunning returns instructions. Without changing your infrastructure. In days.
Multi-tenant, isolated data
Each company has separate data. AI models benefit from anonymized learning across the entire network from day one.
Results backed by data
| System charging blind (no Phase 0) | 70–80% first-attempt success |
| With Phase 0 (pre-charge decision) | 78–92% first-attempt success |
| No dunning (manual retries) | 15–20% |
| Basic automatic retries | 20–30% |
| Dunning Phase 1 | 38–57% |
| Dunning Phases 1+2+3 | 70–80% |
| Full Dunning (Phases 1–4) | Up to 85% |
| Combining Phase 0 + Phases 1–4 | >95% of planned revenue captured |
Sources: Stripe, Recurly, GoCardless, FlexPay, Churnkey — public benchmarks 2024–2025.
Who is Dunning for?
Banks & acquirers
Decision engine for direct debit, auto-pay and recurring charges.
Billing SaaS & vertical ERPs
Add pre- and post-charge intelligence without replacing your stack.
SaaS & subscription platforms
Reduce involuntary churn and recover subscribers lost to payment failures.
Fintech & lending companies
Maximize recovery on overdue installments without collection teams.
Healthcare, education & utilities
Collect respectfully and persistently, maintaining the user relationship.
Marketplaces & B2B platforms
Manage collections from businesses with the same intelligence as individuals.
“Thanks to Lups we set up our auto-replenishment commerce in days. Everything has worked perfectly.”
Francisco, CEO of Tukan
“With Lups, we manage our subscriptions and recurring medication plan billing very efficiently for our customers.”
Julio, CEO of Aiya Te Cuida
Engine pricing on its own
For B2B2C: banks, acquirers, other billing SaaS and vertical ERPs with their own collections.
LupsDunning Engine
Up to 15,000 mandates under management
Plug the engine into your current system. Smart Retry, Standard and Deep recovery, with performance fees only on charges actually recovered.
Engine Enterprise
Above 15,000 mandates
High volume, custom integrations, SLAs, dedicated support. Let's talk through your operation.
Talk to a specialistBy definition, everything coming into the engine has already failed in your system (failure rate = 100%).
Average amount of each charge retried via the engine.
Percentage of failed charges you expect to recover (default 95%).
See breakdown by window
Start charging smart — before and after the attempt
Connect LupsDunning to your system in days. Results from the first planned charge, not from the first failure.
Talk to a specialistEngine Standalone is just the engine. Want the full platform?
If your business charges end customers, the full Lups + Dunning add-on for $99,990/mo (vs $299,990 standalone) is a better fit. 67% cheaper bundle + platform + LupsOne.
See Lups + Dunning