Managing a contact centre or technical support team is genuinely difficult. You’re balancing three moving parts simultaneously: making sure you have enough agents to handle incoming volume, ensuring those agents have the right skills for the work they’re receiving, and verifying that their performance is meeting the standards your business has set.
Most leaders focus on the first two. The third — systematically understanding where your team actually sits against those standards — gets less rigorous attention than it deserves. Bell curve analysis is one of the most practical tools for doing this properly, and it’s underused in support operations relative to how useful it is.
This post covers how to build a performance distribution for your team, how to read what it’s telling you, and — critically — what to do about what you find.
What a Bell Curve Is and Why It Matters
A bell curve is a histogram — a graph that shows how values are distributed across a population. In a contact centre context, you’re plotting individual agent performance on one axis against the count of agents achieving that performance on the other.
The horizontal axis (X-axis) shows a range of performance values — number of emails responded to per day, tickets resolved per week, or whatever output metric you’re measuring. The vertical axis (Y-axis) shows how many agents fall at each point on that range.
In a healthy, well-managed team, you expect a normal distribution: a cluster of agents performing around the target, a roughly equal spread above and below, and relatively few agents at the extremes in either direction. This is the classic bell shape.
The shape of your distribution tells you things about your team that a simple average cannot. An average can look fine while hiding a deeply uneven distribution — a small group of high performers masking a larger group who are struggling. The bell curve makes the full picture visible.
How to Build Your Distribution
Step 1: Choose the right metric
The metric you choose needs to be measurable, consistent across agents, and directly relevant to the outcomes you care about. Output metrics — emails handled, tickets resolved, calls completed — are the most common starting point. If you’re also measuring quality or CSAT, you can build separate distributions for each and compare them. A team with a healthy output distribution and a skewed quality distribution has a very different problem from a team where both are misaligned.
Step 2: Gather enough data
This is where most informal performance analysis goes wrong. Do not judge your team’s performance on a single month of data.
You need at minimum three to six months of performance data to get a meaningful distribution. You also need to account for anything that would distort individual data points: time away from the role on special projects, extended leave, periods of reduced volume, system outages that affected productivity. If an agent spent two weeks of the measured period on a training assignment, their output number for that period doesn’t represent their actual performance capability.
On sample size: you need at least 20 agents to get a distribution that’s statistically meaningful. Below that, you’re looking at a handful of data points and calling it a pattern. The more agents you have, and the longer the measurement period, the more reliable the picture. If you’re managing a team of 12, you can still do this exercise, but treat it as directional rather than definitive.
Step 3: Calculate and plot
Take each agent’s performance data, average it across the measurement period, and plot the distribution. Your target performance value — the metric you’ve set as the expected standard — should be the reference point you’re measuring against.
If your email response target is 40 per day and your distribution peaks at 38, your team is clustered slightly below target but broadly performing. If the peak is at 25, you have a systemic problem. If you have two peaks — one at 20 and one at 45 — you have a bimodal distribution, which means you effectively have two teams operating under the same management structure with fundamentally different capability or motivation levels.
Reading What the Distribution Tells You
The peak
Where the bulk of your team clusters is your operational centre of gravity. Ideally this is close to your target. If it’s significantly below, the gap between expected and actual performance is a management issue — it might be training, tools, workload design, or target-setting — but it needs to be addressed.
The spread
A narrow distribution (agents clustered tightly together) suggests a consistent team where performance is relatively homogeneous. A wide distribution suggests high variance — some agents significantly outperforming, others significantly underperforming, with the average masking both. Wide distributions often indicate that best practices aren’t being shared, that training is inconsistent, or that agent capability varies significantly and skill-based routing isn’t accounting for those differences.
The skew
If the distribution is skewed left (more agents below target than above), you have a performance problem that needs active intervention. If it’s skewed right (more agents above target than below), you either have an excellent team or — worth checking — a target that’s set too low.
The double hump
A bimodal distribution — two peaks rather than one — is particularly interesting and often misread as a problem with the analysis. It’s not. It means your team naturally divides into two groups with different performance patterns. This happens with call volume by time of day (two peaks: morning and afternoon), with agent cohorts that have very different tenure levels, or with teams that have been merged and haven’t yet converged on common practices. Identify which group is which before deciding what to do.
What to Do With What You Find
Compare your high and low performers
The most practical value of the bell curve is what it enables you to do next: sit with your top-quartile performers and understand what they’re doing differently. How do they structure their time? How do they handle the transitions between interactions? What do they do during wrap time that others don’t?
Then sit with your bottom-quartile performers and understand what’s getting in their way. Is it knowledge gaps that training can address? Tool problems that slow them down? Workload design issues that make it genuinely harder for some agents than others? Or are there individuals who — after support and coaching — are still not meeting the standard?
The bell curve gives you the targeting. The individual conversations give you the answer.
Use it to set meaningful targets
One of the reasons AHT targets get set incorrectly is that they’re often set without reference to what the team is actually capable of. A target derived from your distribution — benchmarked against your top quartile’s actual performance — is grounded in operational reality rather than aspiration. It also gives you a defensible basis for both the target and the development plan for agents who aren’t meeting it.
Use it for staffing and capacity planning
The distribution is also useful input for workforce planning. If your team’s average output is below target, your staffing model built on that target is wrong — you’ll be understaffed for actual demand. Understanding where your team actually performs, rather than where you expect them to perform, makes your Erlang C calculations and shift planning more accurate.
Skew the curve deliberately
Your goal, over time, is to move the peak of your distribution toward your target — and then to raise the target. This happens through: developing your bottom performers through coaching and training; understanding and sharing what your top performers do differently; removing process or tool friction that drags performance down; and making deliberate resourcing decisions when individual performance consistently sits outside the range that coaching can address.
A well-managed team’s bell curve should shift rightward over time. If it isn’t — if the distribution stays in the same place month after month despite management effort — that’s a signal to look harder at whether the target is right, whether the tools are adequate, and whether your coaching approach is actually changing behaviour.
A Few Practical Cautions
Don’t confuse output with quality. A high-output agent who’s generating repeat contacts because issues aren’t being resolved on the first call isn’t a high performer — they’re a hidden cost. Wherever possible, build distributions for both output and quality metrics and compare them. The agents who are high on both are your real benchmark.
Don’t judge on a single period. Performance varies. An agent who has one difficult month against a backdrop of consistent good performance is different from an agent who consistently underperforms. The distribution is most useful as a rolling view, not a snapshot.
Don’t use the distribution to create competition. Pinning a bell curve on the break room wall and letting agents see exactly where everyone sits creates anxiety and gaming rather than improvement. The distribution is a management tool — use it to inform your coaching conversations, not to rank agents publicly.
Context matters. An agent handling complex escalations will have lower output than one handling straightforward enquiries. If your distribution mixes different interaction types without controlling for complexity, you’re comparing things that aren’t comparable. Segment your analysis by queue or interaction type if your team handles significantly different work.
The Honest Summary
Bell curve analysis isn’t complicated. What makes it powerful is the discipline of doing it consistently, with enough data, and with the commitment to act on what you find.
The distribution won’t tell you why your low performers are struggling. It won’t tell you how to develop your team or what to change in your processes. What it will do is cut through the noise of anecdote and impression and show you — clearly, quantitatively — where your team actually is. That’s the starting point for every meaningful performance conversation.
And if you look at your distribution and don’t like what you see? That’s valuable. A problem you can see is a problem you can address.
Related reading:
- KPIs and the Importance of Measurements
- Erlang C and Contact Centre Scheduling
- Navigating Skill-Based Routing and Scheduling
Hutch Morzaria is a Director-level CX and Support Leadership professional with 19 years of experience building global support organizations across SaaS, Fintech, and enterprise technology. He has hired dozens of support and CX leaders across his career and holds ITIL Expert certification across V3 and V4.
