IPRN and CDR Reconciliation: Making Settlements Auditable

IPRN and wholesale voice settlement rarely fail on price alone—they fail on how minutes become money. The operational spine is settlement reconciliation: the daily, weekly, and month-end process that aligns SIP and SS7 CDRs, mediation outputs, and rate-deck versions against signed terms. Done well, it compresses DSO, lowers dispute rates, and stabilizes partner churn. Done poorly, it traps cash in arguments about billable seconds and CLI integrity. This memo details how to make settlement reconciliation auditable and scalable.

Where reconciliation breaks in IPRN, SIP and SS7 flows

Most settlement leakage starts before the invoice exists. It starts when two parties believe they are describing the same call, but their records were produced by different switches, timestamps, mediation rules, and commercial assumptions. In wholesale voice, that gap is not an engineering footnote. It becomes trapped cash, write-off pressure, and a partner relationship that consumes senior time every close cycle.

CDR schema drift is the first break point. One upstream may expose P-Charging-Vector and a stable Call-ID. Another may expose only trunk labels, disconnect causes, and a proprietary session reference. On SS7 legs, CIC, OPC, and DPC can identify a circuit path without giving a deterministic call join across the full route. The commercial consequence is familiar: one side can prove a payable call; the other can prove traffic existed; neither can prove they are pricing the same event.

Timing mismatches add systematic variance. A counterparty may anchor billable start to 200 OK, while another uses ANM or suppresses early-media cases. Per-second charging, 30/6 increments, and minimum-charge rules can all be correct under different annexes. Timezone handling is less visible but just as material. A one-hour DST error around a rate change can move a full traffic band into the wrong price period.

Rate-deck ambiguity is the third recurring source of leakage. Mid-cycle rerates, overlapping effective dates, and currency conversion timing can create double-applied changes or missed changes. When the rate annex says “effective immediately” without a timestamp, finance teams inherit a preventable dispute. In volatile destinations, the difference between rate version A and rate version B can exceed the gross margin on the route.

Identity inconsistencies then obscure destination logic. CLI may be altered by privacy masks, transit gateways, or presentation rules. DID mapping tables for IPRN ranges may not be versioned, which means the same called number can be explained under two different destination buckets after the fact. Policy dependencies make this worse. MNP dip results and portability database snapshots can differ at time-of-call and time-of-bill, changing rating, jurisdiction, and tax treatment.

A Tier-2 MNO, EMEA, ~22M subscribers, found this pattern in a high-margin inbound wholesale route set. The price book was not the main problem. The dispute file showed mismatched answer anchors, partial CDRs after failover, and unversioned IPRN mapping tables. Once those were fixed, the operator did not materially change the route price. It changed the cash profile, because fewer invoices entered dispute and fewer balances rolled into the next close.

From CDR correlation to payable invoice: a workflow that survives audit

A reconciliation workflow should begin with correlation, not with billing. The invoice is an output of evidence alignment. If the evidence is thin, finance teams negotiate from summary totals. If the evidence is structured, they can isolate exceptions, clear the undisputed balance, and reserve only the disputed portion.

The first control is normalization. Upstream SIP, SS7, and IPRN CDRs need a canonical schema with consistent timestamps, E.164 CLI and called-number formatting, trunk labels, points of interconnect, answer event, release cause, and currency. Each leg should receive a stable correlation key. The key does not need to be identical across all technologies, but it must be reproducible from source data and preserved through mediation.

Deterministic joining then links commercial legs. SIP legs can usually join on Call-ID plus P-Charging-Vector where the path preserves both. SS7 legs often require ANM timestamp, CIC, OPC, DPC, and duration windows. IPRN ingress may require DID-to-destination tables and a versioned mapping snapshot. Probabilistic matching can support investigations, but payable settlement should not depend on opaque matching rules when the contract requires auditable records.

Enrichment should occur at bill time using historical snapshots. Rate deck, FX curve, MNP status, tax jurisdiction, number category, and fraud flags should be attached by effective timestamp. This prevents a common failure mode: the billing system reruns a prior month using today’s lookup tables and silently changes the result. The same principle applies to routing changes after a partner migration. Historic traffic should be priced against the commercial and numbering state that existed when the call occurred.

Pre-bill controls make the workflow operational. A test-call harness should cover high-yield IPRN ranges, volatile destinations, and recently rerated bands. Mediation minutes should be compared with SBC counters per trunk, with a documented tolerance. For many wholesale routes, ±0.5% is a reasonable trigger for investigation, not automatic rejection. The goal is to catch structural breaks before the invoice is issued, while avoiding a dispute queue filled with immaterial noise.

Exception handling should separate fraud, quality, and data defects. FAS indicators, ACD inflation, PDD suppression, anomalous answer times, one-way audio flags, and partial failover records do not belong in the same bucket as a rate-deck mismatch. Each exception class needs an evidence pack: source CDR extract, counter view, rate digest, correlation map, and the contractual clause that governs the result.

Outbound evidence should be designed for counterparties and auditors. Hashed CDR subsets, header-only files where permitted, immutable rate-deck digests, and a correlation map reduce the scope for later disagreement. They also allow finance to release undisputed balances while route managers handle exceptions. That is the practical distinction between reconciliation as a spreadsheet exercise and reconciliation as a cash-control process.

Contract parameters that decide the P&L

The MSA and rate annex decide more margin than many route managers acknowledge. A low per-minute price with weak evidence terms can underperform a slightly higher price with weekly pre-bill review, narrow dispute windows, and enforceable data requirements. Contract structure determines cash velocity, fraud exposure, and the amount of management attention consumed by the close.

Financial cadence should be explicit. IPRN often justifies weekly settlement or weekly pre-bill files because payout exposure accumulates quickly and fraud patterns move faster than month-end governance. Standard wholesale voice can remain monthly if the evidence exchange is disciplined. Hybrid models are common: weekly pre-bill estimates, monthly invoice, and a true-up tied to agreed tolerances.

Rerate rules require the same precision. The contract should state the rate-deck version, effective timestamp, notice method, and FX source. If WMR 4pm or another index is used, the timestamp belongs in the annex. The same schedule should state how IRSF claims, VAT, and GST treatment are evidenced. Ambiguity here invites opportunistic repricing in volatile months and post-facto tax adjustments after finance has closed the period.

Settlement cadence

Weekly preferred for IPRN; monthly for standard wholesale; hybrid weekly pre-bill with monthly true-up is common

Evidence exchange SLA

T+3 business days for month-end CDRs; dispute submission window of 15–30 days

Dispute rate target by value

≤0.5% wholesale voice; ≤1.0% IPRN given higher variance

CDR acceptance criteria

Canonical timestamps, E.164 CLI and CLD, correlation key, answer and disconnect cause, trunk and POI identifiers

Rerate policy

Only with written rate-deck version and effective timestamp; FX per WMR 4pm or contract index

Billing increments

Define minimum and step, such as 30/6, per destination class; anchor to ANM or 200 OK event

Fraud reserve / holdback

10–30% of IPRN payouts; release on rolling 60–90 day lookback with KPI thresholds

Chargeback window

60–180 days, with IPRN typically 90–120 days, for IRSF and FAS disputes

Error budget

Tolerance of ±0.3% minutes or ±0.2% value auto-accepted; beyond that, route to dispute track

Late payment terms

Interest after T+30; right to net against inbound wholesale if mutual traffic exists

Minimum commit / shortfall

Define by minutes or revenue; shortfall priced at blended prior-month rate

Tax handling

Reverse charge for cross-border where applicable; document number category mapping to tax jurisdiction

Reserves and chargeback windows deserve CFO-level attention. A 10–30% IPRN holdback is not punitive if release is tied to rolling dispute statistics, clean evidence exchange, and fraud KPIs. It becomes a commercial problem only when the release date is arbitrary or when the annex lets one party extend the reserve without evidence. The clean formulation is simple: reserve against measured risk, release against measured performance.

CDR acceptance criteria also need legal force. If a partner omits answer event, disconnect cause, correlation key, or point-of-interconnect identifier, the contract should say whether the record is non-payable, payable at a conservative fallback rate, or subject to manual exception. Incomplete records should not create unlimited negotiation rights. They should follow a known path with a known financial consequence.

Run-rate execution: dispute reduction and cash acceleration

Execution determines whether the contract has value. Pre-bill variance reviews should reconcile mediation totals against switch and SBC counters per trunk, route, and number category. Variances under the error budget should auto-clear. Variances above it should produce a ticket with source files, affected destinations, value at risk, and the proposed settlement treatment. The operating objective is not zero variance. It is fast separation of noise from money.

Evidence exchange needs file discipline. Whether the channel is a portal or SFTP, both sides should mirror folder structures, hash files, lock rate-deck archives, and preserve submission timestamps. Many disputes start with a banal question: which file did finance use? Immutable archives and file digests remove that argument from the close process.

KPI governance should be linked to commercial consequences. Track ACD, ASR, PDD, no-PDD answers, abnormal short-duration patterns, IRSF lists, and chargeback ageing. Partners that cross agreed thresholds should move to higher reserves, lower payout frequency, or restricted destination access. Partners that remain clean should see reserves released faster. This is how fraud control becomes a pricing and cash-management tool rather than a quarterly argument.

Test traffic remains useful because commercial drift often appears before aggregate counters move. Synthetic calls across destination classes, including IPRN high-yield ranges and recently rerated routes, catch mapping errors, rerate failures, and route changes inside the settlement cycle. The cost of test traffic is small compared with one month of misrated premium traffic.

Audit retention should follow the chargeback window plus a margin. Preserve correlation maps, source CDR hashes, rate decks, MNP snapshots, FX curves, tax mappings, and dispute correspondence for at least the chargeback period plus 30 days. For higher-risk IPRN relationships, retain the evidence longer if local tax rules or financial audit policy require it.

A MVNE servicing 12+ tenants in EMEA applied this run-rate model after repeated month-end disputes across voice and IPRN partners. It did not centralize every raw record. It standardized the correlation map, rate-deck digest, and exception queue. Weekly operations calls cleared small variances before invoice issue. Month-end CFO review focused on aged disputes, reserve release, and partner exposure. The result was not a headline margin change. It was less trapped cash and fewer balances rolling beyond the agreed payment date.

The people cadence should be modest and fixed: weekly operations review, month-end finance sign-off, quarterly contract hygiene. The quarterly review should revisit rerate policy, reserve thresholds, error budgets, chargeback ageing, and tax mapping. If these items wait until a dispute, the commercial position is already weaker.

Reconciliation is a design choice, not a spreadsheet chore. Standardized joins, explicit contract parameters, and disciplined evidence exchange reduce disputes to noise and convert wholesale minutes into predictable cash flow. In thin-margin voice markets, that predictability is often the difference between nominal revenue and collected margin.