Resources

Dunning Agency vs Software: Which Model Fits You

When failed payments start costing real money, two buying models present themselves. You can hire a dunning agency — people who run recovery for you as a service — or you can buy dunning software and run recovery on a system. Both models recover payments. They differ in who does the thinking, how the bill grows, how much control you keep, and how fast you can start. This guide lays out both models honestly, including the stages where the human-led option is simply the better buy.

The two models, defined plainly

The agency model: human-led service

A dunning agency takes recovery off your desk. Typically, the agency reviews your billing setup, configures or runs retry and outreach sequences, writes or adapts customer communications, monitors results, and iterates based on what its team sees across your account and others. The work is led by people, supported by their own internal tools. You are buying judgement, labour, and accountability to a service relationship: calls, reports, recommendations, and a team that answers when recovery numbers move.

That shape has real advantages. It also means the agency's process is partly a black box by design — you are paying for their method, not licensing it — and your recovery quality depends on how much attention your account actually receives relative to their larger clients.

The software model: self-serve system

Dunning software gives you the machine and lets you run it. A dedicated tool connects to your billing platform, reads decline codes, runs retry schedules, sends branded emails and other outreach, hosts card-update pages, and reports what it recovered. Setup is typically self-serve and fast, the rules are visible, and the output is yours to inspect down to the individual payment. Our guide to dunning software explains the category mechanics, and the roundup of the best dunning software compares current options.

The trade is symmetrical: software does not think for you. It executes a well-built playbook consistently and never gets tired, but if your situation falls outside the playbook — an unusual billing model, a delicate customer relationship, a one-off crisis — there is no human on the other side whose job is to improvise.

How the costs are shaped

Exact prices vary by provider and should always be quoted for your own numbers, but the shapes are consistent enough to plan around.

Cost element	Agency model (generic shape)	Software model (generic shape)
Core fee	Retainer-style recurring fee, a percentage of recovered revenue, or a blend of both	Flat subscription, MRR-tiered subscription, or a percentage of recovered revenue, depending on the tool
How the bill grows	Retainers step up with scope or volume; percentage fees grow automatically with recovered revenue	Flat fees stay fixed; tiered fees step up with your size; percentage fees grow with recovered revenue
Setup cost	Onboarding time with their team; sometimes a setup fee	Your own setup time; usually minutes to hours for self-serve tools
Hidden cost	Your time managing the relationship and approving changes	Your time reviewing reports and tuning sequences

Run the arithmetic on your own recoverable base before comparing headlines. Use the recovery calculator to size what failed payments cost you per month, then ask of any quote: what does this model charge me at my current volume, and what does it charge me if recovery works and volume doubles? Percentage-of-recovery pricing, whether charged by an agency or a tool, means your success raises your bill in direct proportion. Flat pricing means it does not. Neither is automatically better — a percentage fee with no retainer can be the lowest-risk way to start — but you should choose the shape with your eyes open, not discover it in month six.

Where humans genuinely win

• High-value, low-volume accounts. If you have a small number of subscribers each worth a great deal, a failed payment deserves a person: a considered email, a phone call, knowledge of the relationship. Software sequences are built for scale, and scale is not your situation.
• Complex or negotiated billing. Custom contracts, manual invoicing, mid-cycle changes, and one-off arrangements break the assumptions automated systems are built on. Humans route around exceptions; software flags or fumbles them.
• No internal owner. Software still needs someone to connect it, read its reports, and act on what it surfaces. If nobody on your team will own recovery at all — not an hour a month, literally nobody — an agency that owns the outcome end-to-end is the honest choice.
• A recovery mess you cannot diagnose. If your billing history is tangled across migrations and legacy plans, an experienced team's audit may be worth more than any tool, at least as a first engagement to clean up and explain what you have.

Where software genuinely wins

• Volume of similar failures. Hundreds or thousands of consumer renewals failing for the usual reasons — insufficient funds, expired cards, generic soft declines — is a classification and timing problem. Systems route decline codes instantly, retry on schedule at 3 a.m., and never let a backlog sit because someone was on holiday.
• Speed to start. Self-serve software can be connected and working a backlog the same day. Agency engagements start with discovery, contracting, and configuration. If failures are accumulating now, the calendar is a feature.
• Visibility and control. With software, every rule is inspectable: which declines get retried, when, with what message, in your brand's voice. You can change a sequence today. Agency changes move at the speed of the relationship.
• Predictable cost at growth. A flat-fee tool costs the same in your best month as your worst. For businesses whose recovered revenue is growing, that predictability compounds in your favour.
• Consistency of the unglamorous basics. Never retrying a hard decline, always asking for one action per message, always working the new failure while the sequence is fresh — systems are simply more reliable at rules that must hold ten thousand times.

Fit by stage

Early stage: low failure volume

If failed payments cost you a few hundred dollars a month, neither model may be worth paying for yet. Native billing retries plus a decent card-update email cover the basics. If you do buy, low-commitment shapes — a self-serve trial or a percentage-only arrangement — keep the risk matched to the prize. Spend your attention on product and onboarding, where early churn is actually decided.

Growth stage: failure volume is now a line item

This is software's home ground for most consumer and prosumer subscriptions: enough volume that automation pays, failures dominated by the standard families, and a team member who can own an hour of review a month. An agency still fits if accounts are high-value or billing is unusually complex. Decide by failure shape, not company age.

Scale stage: recovery is a department-sized problem

At large volume, the question becomes build, buy, or staff. Many scaled operators run software as the engine and keep human judgement for the exceptions the software surfaces — the hybrid the two camps pretend not to sell. Agencies at this stage compete on specialist expertise and overflow capacity rather than basic execution.

Questions to ask either model before signing

1. How do you treat hard declines? The only correct answer is that they are never retried. Any hedging here disqualifies the provider.
2. What does my bill look like at double my current recovered revenue? Get the number, not the shape description.
3. Who writes to my customers, in whose voice, and can I see and approve every message? Your subscribers should never be surprised by who is asking for their card.
4. How is recovered revenue defined and reported? You want recovery attributed at the payment level, separable into retries, card updates, and save offers, against a baseline you agree on — not a blended success story.
5. What happens to my backlog on day one, and my data if I leave? Backfill scope and exit terms are where good relationships stay good.

The bottom line

Choose an agency when your failures are few, valuable, tangled, or ownerless — when judgement per account is the scarce resource. Choose software when your failures are numerous, similar, and time-sensitive — when consistent execution is the scarce resource. Price the shapes, not the pitches: model each option at today's volume and at double, insist on decline-aware basics from both, and keep the customer relationship in your brand's voice either way. The models are tools for different failure shapes, and the honest answer to which fits is the shape of your own decline report.

Frequently asked questions

What is the main difference between a dunning agency and dunning software?

An agency is a human-led service: a team configures, runs, and iterates your payment recovery for you, and you buy their judgement and labour. Dunning software is a self-serve system that connects to your billing platform and runs decline-aware retries, outreach, and card-update flows automatically, with your team owning setup and review. The core trade is judgement and hands-off ownership versus speed, visibility, and predictable cost.

Which is cheaper, an agency or software?

It depends on your volume and on the pricing shape. Agencies commonly charge a retainer, a percentage of recovered revenue, or a blend. Software may be flat, tiered by your size, or percentage-based. Compare totals at your current recovered revenue and at double that figure, including your own time, rather than comparing headline fees.

When is an agency genuinely the better choice?

When accounts are high-value and few, billing is custom or negotiated, nobody internally can own recovery at all, or your billing history is too tangled to diagnose without an expert audit. In those situations, human judgement per account is worth more than automated consistency, and a good agency earns its fee.

Can I use both an agency and software?

Yes, and at scale many operators do: software runs the high-volume standard sequences, while humans handle exceptions, high-value accounts, and unusual cases the system surfaces. If you combine them, define clearly which system owns each failure and each customer message, so subscribers never receive duplicate outreach.

What should I check before choosing either model?

Ask how hard declines are treated, what you will pay if recovered revenue doubles, who writes customer messages and in whose voice, how recovered revenue is defined and reported, and what happens to your existing backlog and your data if you leave. The answers reveal more than any feature list.

Related reading: Gravy alternative