Collections Workflow Design: Dunning Sequences, Prioritization, and Escalation

Segment first: the risk x value matrix

A single dunning sequence applied to every account is wrong for almost every account. Your best customer with a $200,000 balance and a decade of on-time history should not receive the same day-15 template as a two-order account that scored poorly at underwriting. Before designing any timeline, segment the portfolio on two axes: value (exposure, or better, exposure plus strategic revenue) and risk (your internal risk band or score, incorporating payment history).

A 2x2 gives four workable treatment strategies. High value, low risk: relationship-preserving, personal, mostly manual, a named collector, phone-first, generous with grace on the first incident. High value, high risk: the danger quadrant, intensive and senior, tight timeline, early escalation to management, credit hold used decisively, because this is where material losses actually come from. Low value, low risk: fully automated, polite email sequences, portal links, no human time until late-stage. Low value, high risk: automated but fast, compressed timeline, early hold, early agency referral, because the economics of manual effort never work here.

The segmentation earns its keep in capacity terms: in a typical B2B book, the two low-value quadrants hold 70 to 80% of accounts but 20 to 30% of dollars. Automating them releases the majority of collector hours to the accounts where a skilled human conversation changes outcomes. Re-segment monthly and on events, a risk-band downgrade should move an account's treatment the same week, not at quarter-end.

Timeline design: day by day

With segments defined, design the timeline per segment. Here is a reference sequence for a mid-risk, mid-value B2B account on net 30, adjust the intervals by segment, compressing for high risk, extending grace for strategic low risk.

The single most important property of the timeline is that it executes without anyone deciding to execute it. Each step should fire from the calendar and the account state, with human judgment applied to content and exceptions, not to whether the step happens. The second most important property: seriousness must actually escalate. If day 60 sounds like day 15, you have trained customers that your sequence is decorative.

• Day -7 (pre-due): friendly reminder with invoice copy and payment link; confirms receipt and surfaces disputes while they are cheap
• Days 1-5 past due: soft touch, a polite email noting the missed date, assuming administrative oversight; no phone time yet for most segments
• Day 10: second notice, firmer tone, full statement attached, explicit request for payment date
• Day 15: first call. Goal is not scolding, it is diagnosis: paid-and-in-transit, stuck in approval, disputed, or cash problem, each routes differently. Every call ends with a commitment: an amount and a date
• Day 30: escalation letter or email from the credit manager rather than the collector; states consequences (hold at day X, agency at day Y) specifically and calmly; second call if no contact achieved yet
• Day 45: credit hold per policy, communicated to the customer and to sales before it blocks an order, with the release condition stated
• Day 60: formal demand, a letter stating the full amount, a 10-business-day deadline, and the specific next step if unmet
• Day 90: decision point, agency placement, legal, or a documented payment plan; the worst outcome is drift, because collectability decays fast from here

Channel mix: matching medium to moment

Each channel has a distinct job. Email is cheap, documented, and easy to ignore; it carries the routine sequence and creates the paper trail, but an email thread with no reply after two touches is a signal to switch channels, not to send a third. Phone is expensive and effective: a live conversation surfaces disputes, produces commitments, and reads the situation (an evasive AP contact versus a candid one) in ways no template can. Reserve phone time for high-value accounts, diagnosis moments (the day-15 call), and broken promises.

The customer portal changes the sequence's conversion rate more than its tone: every touch that includes a link where the customer can see the invoice, the statement, and a pay button removes the copy-invoice round trip that adds a week to email-only sequences. Physical letters retain force precisely because they are rare; a printed demand on letterhead at day 60 signals procedural seriousness that a fourth email cannot. Some regulated or construction contexts also require specific written notices, know your industry's requirements.

Also, the channel that matters most is the one you have not mapped: the right person. Selling to a buyer but dunning info@customer.com wastes the whole sequence. During onboarding, capture the AP contact, their preferred channel, and the customer's own payables process (approval steps, payment-run days), then time your touches to land before their runs, not after.

Promise-to-pay discipline

The promise to pay is the atomic unit of collections work, and most teams handle it casually: a collector notes 'says they will pay next week' and moves on. Treated with discipline, promises become both the engine of the workflow and its best telemetry. Capture every promise with four fields: exact amount, exact date, payment method, and who made it. 'Next week' is not a promise; '$18,400 by ACH on Friday the 19th, per Maria in AP' is.

Then track promises as first-class objects with three states: open, kept, broken. Open promises pause the dunning sequence for that balance, that is the customer's reward for committing, and drop a follow-up into the collector's worklist for the promise date. Kept promises inform the account's treatment (promise-keepers earn lighter sequences). Broken promises get the strongest rule in the workflow: same-day follow-up, no exceptions. A broken promise chased the day it breaks says your commitments are tracked; chased two weeks later, it says promises to you are free.

Measure promise-kept rate monthly, kept promises divided by promises due in the period. A healthy B2B book runs above 80%. Below 70%, either collectors are accepting vague non-commitments to end calls, a coaching issue, or customers have learned that promising is a costless deferral tactic, a discipline issue. Second broken promise on the same balance should trigger automatic escalation: manager call, compressed timeline, or hold, regardless of where the calendar sequence stands.

Prioritization: beyond oldest-first

The default sort, oldest balance first, systematically misallocates effort: it sends your best collectors to the most calcified, least collectable balances while fresh, saveable delinquency ages into difficulty. Collectability declines steeply with age (industry experience puts expected recovery on 6-month-old B2B debt at roughly half of 90-day-old debt), so marginal effort is usually worth more in the 15-to-60-day band than in the 120-plus band.

Build the daily worklist from a priority score instead. A weighted score with even crude weights beats any single-dimension sort. Recompute it nightly, cap the daily list at what a collector can genuinely work (25 to 35 meaningful touches), and let everything below the line flow to the automated sequence. The point is not algorithmic sophistication; it is that the day's scarce human effort lands where it changes the most cash.

These are the inputs that feed the score, in rough order of weight.

• Promise follow-ups due today. Always first — the discipline above depends on it.
• Dollars at risk: balance x the account's risk of further slippage. A $60,000 balance in a deteriorating account outranks a $75,000 balance in a stable one.
• Risk score and trend.
• Days past due, weighted toward fresh delinquency rather than calcified.
• Dispute status. Undisputed dollars only — disputed dollars route to the dispute process, not the call queue.
• Strategic flags: accounts under a hold decision, and accounts approaching agency referral, where a last internal attempt is cheap against a 25% contingency fee.

Capacity planning and queue balancing

Collections staffing is usually set by history rather than arithmetic, and the arithmetic is not hard. Start from touch capacity: a full-time collector doing substantive work (calls with diagnosis and commitments, escalation prep, promise follow-up) sustains roughly 25 to 35 meaningful touches a day, 500 to 700 a month, alongside email volume that automation should mostly carry. Then estimate demand: accounts entering delinquency per month x average touches to resolution for your book (typically 3 to 6 in the mid segments). If 900 accounts go past due monthly and average 4 touches, you need about 3,600 touches, five to six collectors of capacity, before automation reduces the human share.

Balance queues by workload, not account count. Four hundred low-value automated-segment accounts are a lighter load than 120 high-value manual-segment accounts. Assign by segment where possible, letting collectors develop segment-appropriate styles (relationship-preserving versus firm-and-fast), and rebalance quarterly using touches-required rather than raw counts. Persistent overload shows up in the data before anyone complains: promise follow-ups slipping past their dates and top-priority accounts going untouched for a week are the two canary metrics.

One design warning: do not let queue balancing silently become account ownership so strong that coverage breaks. Vacations, departures, and volume spikes will happen; the workflow, with its state, history, and next actions, must live in the system, so any collector can execute the next step on any account.

Disputes: stop the clock without stopping the cash

Disputes are where dunning sequences go to die, usually because a dispute anywhere freezes collection everywhere. The design principle: isolate the disputed dollars precisely, and keep everything else moving. A $52,000 invoice with a $4,000 pricing dispute is a $48,000 undisputed balance on normal sequence plus a $4,000 item in a resolution process, not a frozen $52,000.

Operationally: when a dispute surfaces (ideally at the pre-due touch, not day 45), log it same-day with a reason code, the disputed amount, and the owning department, pricing to sales operations, shipment to logistics, quality to QA. Ask the customer to pay the undisputed portion explicitly and immediately; reasonable customers do, and a customer who refuses to pay the undisputed remainder has told you the dispute is a stalling tactic, which is itself valuable information that should harden the sequence.

Give dispute resolution its own SLA (10 business days is a workable default) and its own weekly review, with aged disputes read out by owning department. The dunning clock on disputed dollars pauses while the dispute is genuinely being worked; it resumes, at the escalated step, the day the dispute resolves against the customer. And watch the pattern data: an account that disputes 15% of invoices when your book averages 2% is running an extended-terms program on your patience, and belongs in a firmer segment.

Escalating to third parties

Every workflow needs a defined exit from internal effort, because internal effort past a point is worth less than its cost. The trigger should be state-based, not just age-based: refer when contact is established but commitments repeatedly break (two-plus broken promises with no partial payment), when contact cannot be established at all for 30 days despite multi-channel attempts, or when the customer signals inability rather than unwillingness (then the choice is a documented payment plan versus referral). Age alone, commonly 90 to 120 days past due, is the backstop for accounts that drift without any of those signals.

The cost-benefit is straightforward: contingency agencies typically charge 15 to 30% of what they recover, more on small or aged balances, and recovery rates decay with age at placement. Recovering 60% of a balance at a 25% fee nets 45 cents on the dollar, which beats another quarter of internal chasing that yields 20 cents, and decisively beats the zero that comes from waiting until the customer's other creditors have already collected. Placing at day 95 instead of day 180 is often the single most profitable timing decision in the late-stage workflow. Reserve legal action for large balances with documented, undisputed debt and a solvent counterparty; litigation on an insolvent debtor is a fee, not a strategy.

Select agencies like vendors, not saviors: check licensing and insurance, industry specialization (construction debt collects differently than SaaS receivables), and references; understand their process, because their conduct is your brand with a customer you may want back; and measure them on net-back (recovered minus fees) by age band, not gross recovery claims. Split placements between two agencies for a year and the data will pick your primary.

The KPI set that proves the design works

A designed workflow is falsifiable: it makes claims about what will happen, and the KPI set tests them. Four metrics form the core. Collection effectiveness index (CEI), collected as a share of what was collectable in the period, is the headline execution number; above 85% is strong, and it resists the gaming that plagues DSO. Percent of AR current tells you whether the front of the funnel (pre-due and early-stage work) is doing its job; movement here leads DSO by a quarter.

Right-party contact rate, the share of call attempts that reach someone empowered to act on the account, is the workflow's plumbing metric: a low rate (below roughly 40%) means contact data is stale or call timing is wrong, and no script fixes that. Promise-kept rate, above 80% healthy, measures both collector discipline (are they extracting real commitments?) and customer respect for your process. Between them, these two explain most of the variance in early-stage outcomes.

Around the core, track the diagnostics: aging-bucket roll rates (what fraction of 1-30 migrates to 31-60 each month, the earliest warning of process slippage), dispute cycle time and aged-dispute dollars, touches per resolution by segment (rising touches mean sequences are losing force), agency net-back by age at placement (the evidence for earlier referral), and bad debt as a percent of credit sales as the ultimate outcome. Review the core four weekly on the cash call, the diagnostics monthly, and re-tune one design element at a time, a timeline interval, a segment boundary, a priority weight, so you can attribute the change.