Agent specialisation analysis

Agent specialization analysis is the study of how concentrated each support or sales agent is on a narrow set of issue types, products, or customer segments rather than handling everything that comes in. If an agent resolves 400 tickets in a month and 320 of them are billing questions, that agent is highly specialised in billing. If the same 400 tickets are spread evenly across eight categories, that agent is a generalist.

Specialisation matters because depth and speed tend to move together. An agent who sees the same problem class every day builds pattern recognition, knows the edge cases, and resolves faster with fewer escalations. A generalist context-switches constantly and pays a tax on every switch. Analysing specialisation tells you whether your routing is matching work to skill, or whether tickets land wherever there is free capacity.

The goal is not to make everyone a narrow specialist. Some specialisation lifts quality and lowers average resolution time, but too much creates single points of failure and coverage gaps. The analysis is about finding the right concentration for each role and spotting where routing is working against you.

The simplest measure is the share of an agent volume that falls in their single largest category. A cleaner approach borrows the Herfindahl index from economics and squares the share of every category, which rewards true concentration and penalises an even spread. Both need clean category tags on every ticket, so the analysis is only as good as your taxonomy.

1. 1

   Tag every ticket with a category

   Assign each resolved ticket to a single issue type, product area, or segment. Inconsistent tagging is the most common reason specialisation analysis produces noise rather than signal.

2. 2

   Count volume per agent per category

   For each agent, build the distribution of tickets across categories over a fixed window. A month is usually long enough to be stable and short enough to stay current.

3. 3

   Compute the concentration index

   Divide the primary category volume by total volume for a quick read, or sum the squared category shares for a Herfindahl style index that captures the full distribution.

4. 4

   Compare against resolution quality

   Overlay each agent index with their resolution time and reopen rate. The pattern you are looking for is whether higher specialisation actually buys faster, cleaner resolutions.

Specialisation does not move on its own. It is the downstream result of how work is routed, how teams are structured, and how agents are trained. A metric tree makes those drivers explicit, so when specialisation drifts you can trace it to the routing rule or training gap that caused it rather than guessing.

KPI Tree lets you model this decomposition and connect each branch to the team and action that influences it. Routing logic sits with operations, training depth sits with team leads, and category taxonomy sits with the knowledge base owner. When each branch has a named owner, a drift in specialisation becomes a specific question for a specific person rather than a chart nobody acts on.

There is no universal target, because the right level of specialisation depends on the breadth of your product and the complexity of each category. The ranges below are a practical starting point for support teams. Treat them as guard rails, not goals, and always read them against coverage and quality.

Improving specialisation is rarely about pushing the index higher. It is about matching concentration to the work so that depth rises without coverage falling. The levers below move the underlying drivers rather than the headline number.

1. 1

   Treating higher as better

   A maximised specialisation index hides coverage risk. An agent at 0.95 is a liability the day they take leave. Read concentration and redundancy together.

2. 2

   Ignoring tag quality

   If categories are inconsistent, the analysis measures your tagging discipline, not your routing. Clean the taxonomy before drawing conclusions.

3. 3

   Averaging across the team

   A team average of 0.5 can hide a mix of narrow specialists and pure generalists. Always look at the per agent distribution, never just the mean.

4. 4

   Optimising in isolation

   Specialisation only matters if it improves outcomes. Pair it with resolution time and reopen rate, or you are optimising a number that does not change the customer experience.