Payment Routing KPIs: Auth Rate, Failover, Cost, and Recovery

Most routing teams report a blended authorisation rate to leadership. That number — one percentage across all acquirers, currencies, and transaction types — is the metric that receives the most attention and tells you the least about where to act.

A routing operation’s auth rate is not one number. It is a weighted average of several per-acquirer approval rates, each reflecting a different BIN distribution, issuer relationship, and traffic segment. A blended 92% can conceal a primary acquirer running at 94% and a fallback route running at 86% — with a failover policy deciding which transactions land on the weaker path. The same aggregation problem applies to cost: a blended effective rate obscures per-route MDR differences, scheme fee exposure from avoidable cross-border routing, and the interchange yield available if Level 2/3 data is applied consistently.

The operational scorecard for a routing function covers five categories: approval performance, routing reliability, cost efficiency, latency, and recovery. This is the measurement layer for teams who have read what a routing improvement is worth in revenue terms and which levers move the approval rate — the scoreboard for tracking whether those levers are in place and working.

The Short Answer

Five categories. A complete payment routing operations scorecard covers:

• Approval performance — authorisation rate by acquirer, network-decline reason code distribution, soft-decline recovery rate
• Routing reliability — primary acquirer success rate, failover invocation rate, failover success rate, routing rule exception rate
• Cost efficiency — effective rate per route, interchange optimisation yield, cross-border routing share
• Latency — authorisation latency P50/P95 by acquirer, retry-overhead latency
• Recovery — retry win rate by decline code, network token coverage, 3DS step-up conversion rate

Operator-set targets. Scheme monitoring thresholds (VAMP, Mastercard ECP) cover chargeback and fraud ratios — not routing performance or cost efficiency. Vendor-published auth rate uplift figures from PSPs reflect results on their own traffic base and should not be used as routing targets. Most routing KPI targets are judgment calls based on your traffic mix, acquirer contract structure, and volume profile.

The KPI Scorecard

Approval Performance

Authorisation rate by acquirer is the most direct measure of routing policy effectiveness, but it must be read at the per-acquirer level to be actionable. A blended auth rate tells you the weighted average output — not whether a 2 pp decline is attributable to one acquirer relationship degrading, a BIN range that should be on a different acquirer, or a shift in issuer risk appetite in a specific geography. Track the metric daily per acquirer and compare against an established baseline before any aggregation.

The network-decline reason code distribution is the early-warning layer beneath the auth rate. Different ISO-8583 codes indicate different issuer decisions. A spike in code 51 (Insufficient Funds) across a subscription rebill cohort may indicate a retry timing problem, not a fraud signal. A spike in code 05 (Do Not Honour) on new cards in a specific BIN range may indicate an issuer detecting a fraud pattern before your own controls have fired. Track the distribution weekly by issuer region and BIN; investigate any code that shifts more than 20% from its four-week average.

Soft declines — codes indicating the rejection may be reversible — are where intelligent retry logic pays off. The recovery rate differs by code and by retry timing. A blended soft-decline recovery rate that improves can hide a declining win rate on a specific code that drives most of your recoverable volume. Segment this metric by decline code, not in aggregate.

Routing Reliability

Primary acquirer success rate and failover invocation rate tell you whether traffic is flowing where routing policy intends. A step-change increase in failover invocation rate is a leading indicator of primary acquirer instability — it typically appears before auth rate drops, because fallback routes usually carry lower approval rates than the primary. Investigate a failover spike before waiting for the auth rate signal.

Failover success rate is the quality measure for the fallback path. A low failover success rate means the fallback route itself is underperforming — either due to the fallback acquirer’s own approval performance, BIN coverage gaps on the fallback route, or incorrect routing logic sending traffic to the wrong fallback for the BIN range. The multi-acquirer architecture behind failover routing is covered in Multi-Acquirer Routing: When and How to Add a Second PSP. For orchestration vendors that manage failover logic, see Spreedly vs Primer vs Gr4vy: Payment Orchestration Layer Compared.

Routing rule exception rate tracks the health of the ruleset itself. Every transaction that did not follow the intended routing policy — because of fallback, acquirer unavailability, a BIN table gap, a rule override, or a configuration error — represents either unintended cost or a governance gap. In payment orchestration setups, exception events are logged by the orchestration layer. A well-maintained ruleset trends toward zero exceptions. A step-change increase indicates either a ruleset change broke intended logic or an operational configuration issue that will compound if not addressed promptly.

Cost Efficiency

Effective rate by route is the most operationally useful cost metric because it captures the full economics of each routing path, not a blended average. An interchange-plus contract is the necessary prerequisite for per-route decomposition — blended-rate contracts obscure the breakdown between interchange, scheme fees, and acquiring markup, making it impossible to identify which route is contributing cost increases. Itemised statement data from an interchange-plus structure lets you calculate effective rate per acquirer-MCC pair and compare against the contracted rate schedule. Unexplained drift above the contracted rate typically indicates scheme fee category changes, avoidable cross-border routing, or MCC mis-classification. For a walkthrough of how processing statement line items map to cost components, see MDR Stack: Reading a Processing Statement Line by Line.

Interchange optimisation yield measures the basis points recovered versus what would have applied without optimisation. Level 2/3 data submission (purchasing card fields that qualify transactions for lower interchange tiers), correct MIT/CIT classification, and network token qualification all affect which interchange category applies. This yield is a data quality problem as much as a routing problem — a decline in yield without a corresponding shift in card mix usually indicates a data submission regression. The specific mechanism for cost-based routing across domestic schemes in markets like Australia — where the RBA mandates least-cost routing on dual-network debit — is covered in Least-Cost Routing: Sending Payments to the Cheapest Eligible Rail.

Cross-border routing share measures how much traffic is incurring cross-border fees when a domestic routing option existed. Every point of avoidable cross-border routing adds scheme assessment fees. The cause is usually a BIN routing table gap — the BIN was not mapped to the domestic acquirer — or a configuration issue routing a domestic card type through a foreign acquirer path. Track this monthly; a step-change increase is addressable through a BIN routing table audit.

Latency

Authorisation latency at P50 and P95, tracked by acquirer, connects routing performance to checkout UX. The only externally enforced ceiling is the scheme auth timeout — the maximum window before a transaction expires. Internal targets should be set well below that ceiling, leaving budget for processing jitter and retry overhead. P50 is the typical-case latency; P95 is where checkout timeout events actually occur. Tail latency is acquirer-specific and can differ materially between primary and fallback routes.

Retry-overhead latency is the additional time added to the auth path for any transaction requiring a retry attempt. A retry that adds 600ms to the auth path on 8% of transactions may not surface in aggregate latency reporting but will appear in checkout abandonment for users experiencing the delay. Set a retry overhead budget consistent with your checkout session timeout, and alert when P95 retry overhead exceeds it.

Recovery

Retry win rate by decline code is a quality metric for retry logic, not just a volume metric. Hard declines — codes like 05 (Do Not Honour), 41 (Lost Card), 43 (Stolen Card) — indicate definitive issuer rejections. Retrying these immediately does not change the outcome and may increase fraud signals. A non-zero retry win rate on hard decline codes does not indicate good retry performance — it indicates the retry classification logic is not correctly distinguishing hard from soft declines. Segment retry win rate by code and confirm hard-decline codes are routing to a customer communication path rather than a retry queue.

Network token coverage is a prerequisite metric for multi-acquirer routing on card-on-file businesses. Network tokens — Visa Token Service (VTS) and Mastercard MDES — are bound to a merchant ID, not to a specific PSP. This portability means stored credentials can be routed across acquirers without re-tokenisation. PSP tokens are locked to the issuing PSP and constrain which acquirers can process the credential. Low coverage limits multi-acquirer routing optionality on the stored credential base. A coverage drop usually indicates new card-on-file enrollments not tokenising — a missing VTS/MDES integration in a new checkout flow — or a PSP migration that left credentials in PSP-token-only state. For a full treatment of token portability trade-offs, see Network Tokens vs PSP Tokens: When Each Wins.

3DS step-up conversion rate matters where SCA applies or where issuers frequently request a challenge flow. A low step-up conversion rate means a significant share of auth attempts reaching the 3DS2 challenge screen are not completing. This abandonment feeds into auth rate as a non-approval outcome and is sometimes misclassified as an issuer decline rather than a consumer drop-off event. Track by market and issuer, and investigate drops not explained by a known issuer availability event.

Segmentation Discipline

The 15 metrics above have limited diagnostic value as blended totals. Every metric should be segmented before aggregating:

• By acquirer — the primary dimension. Blended auth rate and effective rate hide per-acquirer differences entirely.
• By BIN range — issuer behaviour, decline code patterns, and cross-border fee exposure all vary by BIN.
• By issuer region — auth rate, network-decline patterns, and 3DS challenge rates differ by geography.
• By currency and cross-border status — cross-border transactions carry structural cost and auth-rate differences that inflate blended totals.
• By payment method — cards, A2A rails, and local payment methods have different approval, cost, and latency profiles.
• By MIT vs CIT — merchant-initiated transactions have different SCA rules, decline code distributions, and retry logic than customer-initiated transactions. Blending MIT and CIT auth rates is usually misleading for subscription businesses.
• By retry path — first-attempt approvals and retry-path approvals have different cost and latency profiles; blending them hides retry overhead.

Report approval performance and routing reliability metrics daily to the routing operations team. Report cost efficiency and recovery metrics monthly to finance and leadership. Set escalation thresholds at the segment level — a BIN-level or acquirer-level anomaly is actionable; a blended change may not surface until it has already compounded.

Connecting the Operator KPI Scorecards

The routing scorecard is upstream of the fraud and chargeback scorecards — it does not operate in isolation.

The fraud operations scorecard sits downstream of routing decisions. Routing policies that send high-risk BIN ranges through detection paths with lower model coverage, or use soft-decline recovery to retry transactions without fraud re-scoring, can inflate fraud KPIs. Routing rule exceptions — transactions that did not follow the intended policy — may expose the fraud team to traffic they were not calibrated to handle.

The chargeback operations scorecard is the terminal downstream measure. Auth rate decisions that prioritise approval over fraud filtering create dispute liability. Interchange optimisation yield decisions — correct Level 2/3 data, accurate MIT/CIT classification — also affect the evidence available for dispute representment. Cross-border routing share and effective rate drift both feed into the cost-per-dispute metric tracked in the chargeback scorecard.

Reconcile all three scorecards monthly with risk and payments leadership. The connecting question at that review: which routing policy decisions from the prior period show up as fraud or chargeback anomalies, and what adjustment is required upstream.