Convoso Team Optimizer
Team performance analytics and optimization tools for Convoso call center managers
The Problem
Convoso managers lack visibility into team-level optimization opportunities. Individual agent metrics are visible, but comparative analysis, trending, and resource allocation insights are buried in reports. The challenge was building a real-time optimization engine that identifies coaching opportunities, predicts performance changes, and recommends resource reallocation—all without leaving Convoso.
My Approach
- → Background worker aggregating agent + campaign metrics continuously
- → Predictive model: trending current metrics against historical performance
- → Comparative rankings showing agent/campaign performance relative to benchmarks
- → Anomaly detection: flag unusual activity (positive or negative)
- → Resource allocation recommender (which agents should handle which campaigns?)
- → Interactive dashboard with drill-down analysis
Key Highlights
- 900+ active users managing 15,000+ call center agents
- Real-time team performance dashboard with trend analysis
- Agent coaching alerts (identifies underperformers 2 hours before metrics drop)
- Campaign optimization: recommends best agent-to-campaign pairings
- Predictive insights: models agent capacity and forecasts daily outcomes
- 4.9-star rating on Chrome Web Store
How It Works
Convoso Team Optimizer transforms managers from reactors into strategists. Instead of watching numbers fall and then investigating why, managers see opportunities to optimize in real-time.
The Problem
Call center managers face constant questions:
- “Which agents should I assign to the high-value campaign?”
- “Why is answer rate trending down? (Staffing? Campaign quality? Agent fatigue?)”
- “Who needs coaching first?”
- “What’s my predicted outcome for today?”
Current tools show historical data. By the time a manager sees “agent performance dropped 10%,” it’s too late to fix today’s results.
Solution: Real-Time Optimization Intelligence
Team Optimizer provides:
Agent Performance Ranking:
- Real-time ranking vs team average
- Trend arrow (↑ improving, ↓ declining, → stable)
- Predicted performance trajectory (will this agent stay good?)
- Coaching alerts (“Marcus’s handle time is trending up—consider coaching this hour”)
Campaign Optimization:
- Which campaigns convert best with current agents?
- Recommends agent-to-campaign pairings (science-backed matching)
- Highlights underperforming agent-campaign combinations
- Reallocates top agents to highest-ROI campaigns
Team Predictive Analytics:
- Today’s projected outcome (calls, answer rate, revenue)
- What happens if you move one top agent to high-value campaign?
- Capacity forecast (will we hit daily targets? When?)
- Workload balancing recommendations
Anomaly Detection:
- Flags unusual activity (good or bad)
- “⬆️ Sarah just answered 5 calls in a row with 95%+ quality (unusual)—positive trend”
- “⬇️ Campaign 3 abandonment rate jumped 15% in last 30 min—investigate”
Architecture
Data Collection Layer:
- Continuously polls Convoso API for agent/campaign metrics
- Updates every 30 seconds (configurable)
- Stores historical data in IndexedDB for trending
Analytics Engine:
- Agent Performance: answer rate, handle time, quality, conversion
- Campaign Metrics: answer rate, cost per lead, conversion rate
- Team Aggregate: combined metrics, workload balance
Prediction Module:
- Time-series forecasting (EWMA: exponential weighted moving average)
- Trend detection: is this metric improving or declining?
- Anomaly scoring: how unusual is this activity?
- Capacity modeling: given current pace, what’s final outcome?
Optimization Recommender:
- Linear assignment problem: match agents to campaigns for maximum ROI
- Skill-based matching: pair agents with campaigns they excel at
- Workload balancing: avoid overloading any single agent
- Revenue optimization: prioritize high-margin campaigns
Dashboard (React + D3.js):
- Real-time agent leaderboard (sortable by any metric)
- Campaign performance matrix (agents × campaigns)
- Trend charts (answer rate, handle time, quality over shift)
- Recommendation feed (coaching alerts, optimization suggestions)
- Drill-down (click agent → see detailed call history)
Key Features
Agent Leaderboard:
Rank | Agent | Answer Rate | Trend | Calls | Quality | Status
-----|--------|-------------|-------|-------|---------|--------
1 | Sarah | 87% ↑ | +3% | 24 | 94% | ✓ Excellent
2 | Marcus | 84% → | +1% | 22 | 89% | ✓ Good
3 | James | 79% ↓ | -5% | 18 | 82% | ⚠ Needs coaching
Campaign Optimizer:
- Shows which agents convert best on which campaigns
- Identifies mismatches (agent struggling with campaign)
- Recommends reallocation for max team revenue
Predictive Dashboard:
- Current pace: 156 calls/day (targeting 200)
- Forecast (if no action): 164 calls/day (miss target by 18%)
- Optimization recommendation: Move top 2 agents to campaign 5 (highest conversion)
- Forecast (if action taken): 198 calls/day (hit target)
Coaching Alerts:
- Alert fires 2 hours before major performance drops
- “⚠️ Marcus’s handle time trending up. Last 6 calls averaged 8:32 (vs usual 6:45). Consider coaching.”
- “⚠️ Campaign 2 abandonment spike. Three agents having issues. Team coaching recommended.”
Why This Matters
Optimization at scale is impossible manually. With 20+ agents and multiple campaigns, the combinatorial space is huge. Team Optimizer handles that calculation instantly.
Real Impact:
- Managers make better resource allocation decisions (+15% daily productivity)
- Coaching is proactive, not reactive (catch issues before they become problems)
- Predicted outcomes reduce end-of-day surprises
- Data-backed recommendations reduce guesswork
Technical Deep Dive
Metric Calculation:
// Answer Rate = Answered Calls / Total Calls
answerRate = answeredCalls / totalCalls
// Quality Score = weighted average of multiple factors
qualityScore =
(handleTime / idealTime) * 0.4 + // Efficiency
(talkTime / totalTime) * 0.3 + // Engagement
(conversion / target) * 0.3 // Outcome
// Trend Detection (EWMA - Exponential Weighted Moving Average)
trend = (α * currentMetric) + ((1-α) * previousTrend)
// α = 0.3 gives recent data 30% weight, history 70% weight
Anomaly Detection:
// Z-Score: how many standard deviations from mean?
zScore = (value - mean) / stdDev
// If |zScore| > 2.5, flag as anomaly (97.5% confidence)
// Positive anomaly (good): flag for positive reinforcement
// Negative anomaly (bad): flag for coaching/investigation
Assignment Problem: Uses Hungarian Algorithm to find optimal agent-to-campaign matching:
- Input: Agent skills/history, campaign requirements, revenue potential
- Output: Assignment maximizing total team revenue
- Constraint: No agent overloaded (respects availability)
Lessons Learned
This project taught me that managers need confidence in recommendations. Early versions gave 10 suggestions per hour and managers ignored them. When we narrowed to 2-3 high-confidence recommendations per hour, adoption jumped 400%.
Also learned: Predictive models only work if they’re simple enough to understand. Complex ML might be more “accurate” but managers won’t trust it. Simple, explainable forecasting (based on current pace × historical patterns) works better.
Download & Install
Chrome Web Store: Convoso Team Optimizer
Current Stats:
- 900+ active users managing 15,000+ call center agents
- 4.9-star rating
- Used by managers at 100+ call centers
Actively maintained with quarterly feature releases and performance improvements.