Predictive Routing

Predictive routing uses machine learning to match incoming leads and new deals to the sales representative most likely to close them successfully. Rather than relying on simple round-robin assignment or static territory rules alone, SalesOS analyzes each opportunity's characteristics and each rep's strengths to produce an optimal match. The result is higher win rates, faster response times, and more balanced workloads across your team.

What Is Predictive Routing

Traditional lead assignment methods distribute work evenly but blindly. Round-robin ignores that a rep who specializes in enterprise healthcare deals is a better fit for a large hospital system inquiry than a rep whose strength is mid-market technology companies. Territory-based routing accounts for geography but misses expertise alignment.

Predictive routing adds an intelligence layer on top of your existing rules. It considers:

• The characteristics of the incoming lead or deal (industry, company size, deal value, product interest, source).
• The profiles of available reps (past win rates by segment, product expertise, language skills, current capacity).
• Historical outcomes (which rep-to-opportunity pairings have produced the best results).

The system then recommends an assignment -- or, if configured for auto-routing, makes the assignment directly -- with an explanation of why that rep is the best fit.

How Matching Works

The predictive routing engine evaluates several dimensions when scoring potential rep-to-opportunity matches:

Skills and Expertise

Each rep has a profile that includes expertise tags (see Rep Profiles below). The routing model compares the opportunity's characteristics against these tags:

• Industry expertise -- A lead from a financial services company scores higher for reps who have closed deals in financial services.
• Product expertise -- If the lead expressed interest in your analytics module, reps certified on that module are preferred.
• Technical depth -- Complex enterprise deals requiring technical selling are routed to reps with solutions engineering backgrounds.
• Language -- If the contact's preferred language or account region requires a specific language, reps with that language skill are prioritized.

Territory Alignment

Predictive routing respects your territory definitions as a foundational layer:

• A lead from a geography assigned to a specific team will be routed within that team.
• Named accounts are respected -- if an account is already owned by a rep, new leads from that account go to the owner.
• Territory rules act as hard constraints; the predictive model optimizes within those bounds.

Capacity and Workload

A rep who is the best fit on paper but is already juggling 40 active deals will not deliver a great experience. The routing engine accounts for:

• Current open deal count relative to the rep's configured maximum.
• Pipeline value already under management.
• Upcoming PTO or out-of-office status (synced from calendar).
• Recent assignment velocity -- If a rep received five new leads today, additional leads may route elsewhere to prevent overwhelm.

Historical Win Rate

The model learns from past outcomes:

• For each opportunity characteristic (industry, deal size bracket, source channel), it tracks which reps have historically won similar deals.
• Reps with above-average win rates for a given segment receive a higher match score.
• Recency matters -- recent performance is weighted more heavily than results from 18 months ago.

Deal Characteristics

Specific attributes of the opportunity influence routing:

• Estimated deal value -- High-value enterprise deals may require senior reps or dedicated enterprise teams.
• Source channel -- Inbound demo requests, outbound prospected leads, partner referrals, and event leads may each route to specialized teams.
• Urgency signals -- If the lead indicated a near-term purchase timeline, routing prioritizes reps with immediate availability.
• Competitive mentions -- If the lead mentioned evaluating a specific competitor, reps with competitive displacement experience are favored.

Routing Model Training

The predictive routing model improves over time by learning from your organization's outcomes.

Initial Training

When you first enable predictive routing, the model trains on your historical data:

• Past 12 months of closed-won and closed-lost deals.
• The rep assigned to each deal and the deal's characteristics.
• Time to first touch after assignment.
• Deal cycle length and conversion rates by rep-segment pairing.

The initial model requires at least 100 closed deals with recorded outcomes to produce reliable predictions. If you have fewer, SalesOS uses a rules-based fallback until sufficient data accumulates.

Continuous Learning

After the initial training:

• Every deal outcome (won, lost, disqualified) feeds back into the model.
• The model retrains weekly on a rolling 12-month window to adapt to changing rep skills and market conditions.
• Seasonal patterns (Q4 budget flush, summer slowdowns) are captured in the training data.

Model Performance Metrics

Track how well the routing model performs on the Routing Analytics dashboard:

• Model accuracy -- What percentage of the model's top recommendation resulted in the best outcome versus a random assignment.
• Win rate lift -- The win rate improvement for predictively routed deals compared to manually routed deals during the same period.
• Match score distribution -- A histogram showing how often the model produces high-confidence matches versus low-confidence suggestions.

Real-Time vs Batch Routing

SalesOS supports two routing modes that can be used independently or together:

Real-Time Routing

Leads and deals are routed immediately as they enter the system:

• A new inbound lead from the website is scored and routed within seconds.
• The assigned rep receives an immediate notification with the lead details and match explanation.
• Best for high-intent inbound leads where response time correlates with conversion rate.

Batch Routing

Leads are collected and routed in scheduled batches:

• Useful for outbound prospecting lists, event lead imports, or partner referral batches.
• Configure batch frequency: hourly, twice daily, or daily.
• Batch routing considers the full set of leads in the batch and optimizes across all reps simultaneously, producing a globally optimal assignment rather than greedy one-at-a-time assignments.
• Results are published as a routing queue that managers can review before confirming.

Choosing a Mode

Routing Rules Configuration

Navigate to Settings > Routing Rules to configure how predictive routing operates for your organization.

Rule Priority

Rules are evaluated in priority order. Higher-priority rules act as overrides:

1. Named account rules -- If a lead matches a named account already owned by a rep, route to that rep regardless of other factors.
2. Territory hard rules -- Geographic or segment-based assignments that must be respected.
3. Team rules -- Route certain lead types to specific teams (e.g., enterprise leads to the enterprise team).
4. Predictive optimization -- Within the constraints above, the model selects the best-fit individual rep.

Creating a Rule

1. Click Add Rule on the routing configuration page.
2. Define the trigger conditions (lead source, industry, deal value range, geography, or any custom field).
3. Set the action: route to a specific rep, route to a team (model picks individual), or exclude certain reps.
4. Set priority relative to other rules.
5. Save and activate the rule.

Fallback Behavior

If the model cannot produce a confident recommendation (match score below threshold) or all qualified reps are at capacity, you can configure fallback behavior:

• Round-robin within team -- Distribute evenly among available reps.
• Route to manager -- Assign to the team manager for manual distribution.
• Hold in queue -- Place in an unassigned queue for manual pickup.

Rep Profiles and Expertise Tags

Each rep's profile powers the matching engine. Navigate to Settings > Team > Rep Profiles to manage these.

Profile Fields

Auto-Tagging

SalesOS can automatically infer expertise tags from a rep's closed-won history. If a rep has closed 15 deals in the healthcare industry over the past year, the system suggests adding "Healthcare" to their industry expertise. Managers receive these suggestions quarterly and can approve or dismiss them.

Capacity Management

Capacity settings ensure no rep is overwhelmed and that routing accounts for real-world constraints.

Configuring Capacity

For each rep, set:

• Maximum concurrent deals -- The total number of active deals a rep can manage (e.g., 25 for mid-market, 8 for enterprise).
• Maximum new assignments per week -- A cap on how many new leads or deals can be assigned in a rolling seven-day window.
• Current load -- Automatically calculated from the rep's active pipeline. Displayed as a percentage of maximum.

Capacity Thresholds

Temporary Unavailability

Reps can mark themselves as temporarily unavailable (vacation, sick leave, focused on closing existing pipeline). During this period:

• No new leads are routed to them.
• Their existing pipeline remains assigned.
• When they return, their capacity recalculates and they re-enter the routing pool.

Routing Suggestions vs Auto-Assignment

SalesOS supports two operational modes for how routing decisions are applied:

Suggestion Mode

The model recommends an assignment, but a human confirms it:

• New leads appear in a Routing Queue with the model's recommendation and match score.
• Managers or ops leaders review the queue and approve, reject, or modify assignments.
• Best for organizations new to predictive routing who want to build trust in the model.
• Also useful for high-value enterprise leads where the assignment decision warrants human judgment.

Auto-Assignment Mode

The model makes the assignment directly without human intervention:

• Leads are assigned and the rep is notified immediately.
• Reduces time to first touch since there is no queue wait.
• Best for high-volume inbound leads where speed matters most.
• Requires confidence that the model and rules are well-calibrated.

You can use different modes for different lead types. For example, auto-assign inbound SMB leads but use suggestion mode for enterprise leads over $100K.

Override and Manual Assignment

Even with predictive routing enabled, managers and ops leaders retain full control:

• Manual reassignment -- Open any lead or deal and change the owner. The system logs the override and the reason (if provided).
• Bulk reassignment -- Select multiple leads from the queue or list view and assign them to a specific rep. Useful when territory changes occur or a rep departs.
• Override with feedback -- When you override the model's suggestion, SalesOS asks for a reason (optional). This feedback is used to improve future recommendations.
• Lock assignment -- Mark a lead as "locked" so automated re-routing workflows do not move it. Useful for strategic accounts with deliberate rep assignments.

Routing Analytics

Navigate to Analytics > Routing to monitor the effectiveness of your routing strategy.

Key Metrics

Dashboards

The routing analytics page includes:

• Match score distribution -- A histogram showing how confident the model is across all recent assignments.
• Win rate lift chart -- Compares win rate for predictively routed deals vs manually assigned deals over time.
• Rep performance by match tier -- Shows each rep's win rate for deals where they were a "high match" vs "low match," validating that the model is identifying true affinities.
• Time to first touch trend -- Tracks whether routing speed is improving as you refine rules and enable auto-assignment.
• Workload balance -- A bar chart showing current deal load per rep relative to their capacity cap.

A/B Testing Routing Strategies

To validate that predictive routing outperforms your current approach, SalesOS supports controlled experiments:

Setting Up a Test

1. Navigate to Settings > Routing > Experiments.
2. Click Create Experiment.
3. Define the control group (e.g., round-robin assignment) and the treatment group (predictive routing).
4. Set the traffic split (e.g., 50/50 or 80/20).
5. Define the primary success metric (win rate, time to first touch, deal velocity).
6. Set the experiment duration (minimum recommended: 4 weeks or 200 leads per group).

Interpreting Results

After the experiment completes:

• SalesOS displays statistical significance for the observed difference between groups.
• Review secondary metrics (average deal size, customer satisfaction scores) to ensure the winning strategy does not come with trade-offs.
• If predictive routing wins, promote it to 100% traffic with one click.

Progressive Rollout

For risk-averse organizations, use a progressive rollout:

1. Start with 10% predictive routing, 90% existing method.
2. If metrics are positive after two weeks, increase to 50/50.
3. If still positive, move to 90% predictive, 10% control.
4. Eventually disable the control once confidence is high.

Integration with Other Features

Predictive routing integrates with several other SalesOS capabilities:

• Lead scoring -- Leads with higher scores can be prioritized for faster routing or reserved for top-performing reps.
• Territory management -- Territory boundaries are respected as hard constraints within which the model optimizes.
• Notification system -- Reps receive real-time notifications when a new lead is assigned, including the match rationale.
• Activity tracking -- Time to first touch is measured automatically when the rep logs their first call, email, or meeting after assignment.
• Coaching insights -- If a rep consistently underperforms on a segment they were predicted to excel in, the coaching module can surface this as a development area.

Best Practices

• Keep rep profiles current -- The model is only as good as the data it has about each rep. Review expertise tags quarterly and update them as reps develop new skills or shift focus areas.
• Start with suggestion mode -- When first enabling predictive routing, use suggestion mode for at least 4-6 weeks. This lets you validate the model's recommendations before trusting it with auto-assignment.
• Set realistic capacity caps -- Overly generous caps lead to rep burnout and poor responsiveness. Track time to first touch by load percentage to find the right maximum for each role type.
• Provide override feedback -- When you change the model's suggestion, take a moment to note why. This feedback loop accelerates model improvement and helps SalesOS learn your organization's unwritten rules.
• Monitor win rate lift continuously -- The routing analytics dashboard shows whether predictive routing is delivering value. If the lift plateaus or reverses, review whether rep profiles need updating or if the market has shifted.
• Use A/B testing for major changes -- Before changing territory structures, capacity limits, or routing rules, run an experiment. Data-driven decisions prevent costly mistakes in assignment strategy.
• Balance speed and quality -- For high-intent inbound leads, time to first touch is critical -- favor auto-assignment. For complex enterprise opportunities, a few hours of review time is worthwhile to ensure the right rep is assigned.
• Align with comp plans -- Ensure that routing rules and territory boundaries are consistent with how reps are compensated. Routing leads to a rep who will not receive quota credit creates friction and gaming behavior.
• Review unassigned queues daily -- If leads are falling into the fallback queue because no rep matches, it indicates a gap in your team's coverage. Hire, train, or adjust rules to fill that gap.
• Communicate the system to your team -- Reps should understand how routing works and what drives the model's decisions. Transparency builds trust and encourages reps to keep their profiles accurate.