Skip to main content

Dynamic Routing

Dynamic routing in Pulse automatically adjusts routing priorities based on real success rate performance. Instead of relying solely on static rule-based routing, dynamic routing monitors how payment partners perform over time and updates routing preferences when performance trends change.

Pulse supports two dynamic routing strategies:

  • Threshold-based management takes a conservative approach, switching only when performance drops below a configured tolerance.
  • Success rate maximization takes an aggressive approach, continuously evaluating all partner options and selecting the best performer.

Both strategies use smart guardrails to ensure that routing changes are confident, safe, and auditable.

StrategyTrigger ConditionAdjustment BehaviorRisk LevelBest For
Threshold-based managementSuccess rate drops below configured tolerance compared to benchmarkSwitches away only when threshold is breached; returns to preferred partner after exile durationLowerStability-first routing with protection against performance drops
Success rate maximizationContinuous evaluation, no threshold gateEvaluates all partners continuously and selects the best performer with fair competitionHigherAggressive success rate optimization with controlled exploration

Dynamic routing builds on top of your base rule-based routing configuration. It observes transaction outcomes over adaptive time windows, scores payment partner performance using a weighted blend of recent and historical data, and recommends routing changes when the evidence supports a switch. Smart guardrails enforce minimum data requirements and safety checks before any change takes effect, and every routing change is recorded in an audit trail.

Threshold-Based Management

Threshold-based management is a conservative strategy well suited for merchants who have a preferred payment partner and want automatic protection against performance degradation without frequent switching. It intervenes only when necessary and returns to the preferred partner when conditions improve.

How It Works

A preferred payment partner handles traffic by default. Pulse continuously monitors that partner's success rate within a configurable time window. If the success rate drops below a configured tolerance compared to a benchmark, Pulse switches traffic to the next-best partner. The monitoring window adapts based on transaction volume so the system gathers enough data before acting. Because the threshold must be breached before any change occurs, this strategy avoids unnecessary switching during normal fluctuations.

Weighted Scoring

Pulse scores each payment partner using a weighted blend of recent (short-term) and historical (long-term) performance. Recent transactions carry a recency bias so the score reflects current conditions, while the historical component prevents overreaction to brief anomalies. When a threshold breach triggers evaluation, the partner with the highest weighted score is selected as the replacement. This balanced approach ensures routing decisions are grounded in both trend and context.

Return to Preferred

When a preferred partner is switched out, it enters a cooling-off period defined by the configured exile duration. After this duration elapses, Pulse re-evaluates the preferred partner's performance. If the preferred partner's success rate has recovered, traffic is restored to it automatically. This return-to-preferred mechanism ensures that temporary degradation does not permanently displace a merchant's preferred routing path.

Configuration

Merchants configure the following parameters for threshold-based management:

  • Preferred partner (preferred_psp) -- the default payment partner that receives traffic under normal conditions.
  • Drop tolerance (srt_drop_tolerance) -- the success rate threshold below which a switch is triggered.
  • Exile duration (preferred_exile_duration) -- how long the preferred partner is bypassed before Pulse checks for recovery.
  • Minimum count (min_count) -- the minimum number of transactions required before the system evaluates performance, ensuring statistical significance.
  • Minimum time span (min_time_span) -- the minimum monitoring window duration.

Success Rate Maximization

Success rate maximization is an aggressive strategy well suited for merchants who are willing to accept more frequent partner switching in exchange for continuously optimized success rates. It continuously evaluates all partner options and selects the best performer.

How It Works

Unlike threshold-based management, success rate maximization does not designate a single preferred partner. Instead, Pulse continuously evaluates all available payment partners and routes each transaction to the partner with the highest real-time weighted score. There is no threshold gate -- the system always selects the current best performer. This dynamic selection approach responds rapidly to changing partner performance and is designed to maximize overall success rates across all traffic.

Fair Competition

To maintain statistically meaningful performance data, Pulse ensures every payment partner receives a minimum share of traffic. Without this safeguard, a partner that temporarily underperforms could be starved of transactions and never have the opportunity to demonstrate recovery. By allocating a baseline volume to each partner, the system produces reliable score comparisons and prevents a single dominant partner from monopolizing traffic indefinitely.

Weighted Scoring

Success rate maximization uses the same weighted scoring model as threshold-based management, blending recent and historical performance. A rolling evaluation window defines the time frame for score calculation. Recent transactions carry greater weight so the score tracks current conditions. Transaction volume is normalized to ensure partners with different traffic levels are compared fairly. Because there is no threshold requirement, the highest-scoring partner is always selected regardless of the margin.

Configuration

Merchants configure the following parameters for success rate maximization:

  • Switching type (switching_type) -- set to maximise_srt_fair_volume to enable this strategy.
  • Minimum count (min_count) -- the minimum number of transactions per partner required before scores are considered valid.
  • Minimum time span (min_time_span) -- the minimum scoring window duration.

Smart Guardrails

Smart guardrails are the safety layer that governs both dynamic routing strategies. They ensure routing changes are backed by sufficient evidence and do not introduce unintended risk.

Adaptive Time Windows

The monitoring and scoring window adjusts automatically based on transaction volume. During high-traffic periods, Pulse uses shorter windows (as low as 15 minutes) to detect and respond to performance shifts quickly. During low-traffic periods, the window extends (up to 180 minutes) to accumulate enough transactions for a statistically meaningful evaluation. The current window state is persisted per payment option so each payment mode and partner combination is evaluated at an appropriate cadence for its volume.

Safety Checks

Multiple built-in protections prevent premature or erroneous routing changes:

  • Minimum transaction count -- no routing decision changes until the configured minimum number of final-outcome transactions is reached within the current window.
  • Minimum time span -- enforces a floor on how quickly consecutive changes can occur.
  • Pending transaction exclusion -- transactions that have not reached a final status are excluded from score calculations to avoid skewed data.
  • Duplicate change protection -- a hash of each routing change payload prevents the same change from being processed more than once.
  • Supported payment modes -- dynamic routing is restricted to supported modes (such as Cards, UPI, and Netbanking) to avoid applying performance-based logic where it is not applicable.

Configuration

Merchants and operations teams can configure the following guardrail parameters:

  • Minimum time span (min_time_span) -- the adaptive window floor and ceiling (15 to 180 minutes).
  • Minimum count (min_count) -- the minimum transaction count required before a routing change is evaluated.
  • Pending exclusion window -- the duration within which pending transactions are excluded from scoring.
  • Provider exclusions -- specific payment partners that should be excluded from dynamic routing evaluation.

Audit Trails

Every routing decision is logged with full transparency. The audit record captures which payment partner was selected, why the change was triggered, and what the weighted scores were at decision time. This change history is available for review and ensures complete accountability. Audit trails support both operational troubleshooting and compliance by providing a clear record of every automated routing adjustment.

Both strategies can be combined with Fallback & Recovery to handle individual payment failures, while dynamic routing handles broader performance trends at the routing-configuration level.