Conversation channel analysis

Conversation channel analysis is the practice of comparing how support conversations perform across each channel a customer can reach you on. A channel is the medium of contact, for example email, live chat, phone, in-app messaging, social media or a self-service help centre. The analysis looks at how volume splits across those channels and how each one performs on speed, quality and cost.

The core output is a channel mix: the share of total volume each channel carries. If you handle 10,000 conversations in a month and 4,500 of them arrive by email, email carries a 45 percent share. That share on its own is descriptive. The analysis becomes useful when you place it next to performance, because a channel can be high volume and slow, or low volume and expensive.

Channel analysis matters because each channel has a different cost to serve and a different customer expectation. Phone is high touch and expensive per contact. Self-service is cheap but only works for simple, well-documented issues. Live chat sits in between. Moving the right conversations to the right channel lowers cost without hurting the experience, but only if you can see the trade-offs clearly.

Channel analysis is not a single number. It is a small set of measures calculated for every channel so they can be compared side by side. Start with channel share, then layer the performance measures that matter for your team.

1. 1

   Channel share

   Conversations on the channel divided by total conversations across all channels. This tells you where demand concentrates and which channels carry the operational load.

2. 2

   Resolution rate per channel

   The share of conversations on each channel that reach a confirmed resolution. A channel with high volume but low resolution is pushing work downstream rather than closing it.

3. 3

   Average resolution time per channel

   How long conversations take to close on each channel. Compare this against the average resolution time for the whole team to spot the slow channels.

4. 4

   Cost per conversation per channel

   Fully loaded cost to handle one conversation on a channel, including agent time, tooling and overhead. This is what makes channel strategy a financial decision rather than a preference.

5. 5

   Satisfaction per channel

   The customer satisfaction score recorded against conversations on each channel. A cheap channel that frustrates customers is a false economy.

With these measures in place you can read the channel mix properly. A channel is healthy when its share, resolution rate and satisfaction are all in balance against its cost. The goal is not to make every channel identical. It is to route each type of issue to the channel that resolves it well at the lowest sensible cost.

A metric tree turns channel analysis from a static report into a diagnostic. It decomposes total conversation volume into channels, then breaks each channel into the levers that determine its cost and quality. This lets you see not just which channel is underperforming, but why.

The first level splits total volume into the channels you operate. Each channel then decomposes into the things that drive its performance: the volume routed to it, the rate at which it resolves issues, the time it takes, and the cost it incurs. A self-service channel decomposes differently from a phone channel, because deflection and article coverage drive one while staffing and handle time drive the other.

This structure makes the cause and effect visible. If cost per conversation is rising, the tree shows whether expensive phone volume is growing, whether resolution rates are slipping so issues bounce between channels, or whether self-service is failing to deflect the simple questions it should absorb.

Channel benchmarks vary with product complexity and customer base, but the relative cost and speed of each channel follow a consistent pattern. Use these ranges to sense-check your own mix rather than as hard targets.

A common healthy pattern is a large self-service share absorbing simple questions, a moderate chat and email share handling the bulk of human conversations, and a small phone share reserved for the issues that genuinely need it. If phone share is high and self-service share is low, the analysis is telling you that simple issues are landing on your most expensive channel.

Improving channel performance is about moving each type of issue to the channel that resolves it well at the lowest sensible cost, then making each channel better at the job it is suited to. The aim is a deliberate mix, not an accidental one.

The metric tree approach starts by finding the channel with the widest gap between its current cost-quality balance and where it could be. If phone volume is high and resolution on cheaper channels is low, fixing routing and self-service will move the headline more than squeezing any single channel.

KPI Tree lets you model this by connecting each channel branch to the team that owns it. The knowledge team owns self-service coverage and deflection. Support operations owns routing and staffing. The product team owns the friction that generates contact in the first place. When the metric moves, the push reaches the accountable owner of the branch that moved, so the right team acts rather than the whole organisation reading the same dashboard.

1. 1

   Counting messages instead of conversations

   A channel that uses many short messages, like chat, looks busier than it is when you count messages. Always normalise to conversations so the mix reflects real demand rather than the chattiness of a medium.

2. 2

   Comparing channels on speed alone

   Chat resolves in minutes and email in hours, but that does not make chat better. The two carry different issue types. Compare each channel against the work it is suited to, not against the fastest channel.

3. 3

   Ignoring cross-channel handoffs

   When a customer starts on chat and ends on phone, the conversation touched two channels. If you do not track that handoff, both channels record a partial story and your resolution rates are overstated.

4. 4

   Optimising cost without watching satisfaction

   Pushing volume to the cheapest channel lowers cost on paper but raises repeat contacts when the channel cannot resolve the issue. Net cost can rise even as cost per conversation falls.

5. 5

   Treating the mix as fixed

   Channel demand shifts with product changes, seasonality and customer growth. A mix that was right last quarter can be wrong now. Re-read the analysis regularly rather than setting routing once and forgetting it.