How should inbound calls be categorized?
Use two separate axes: why the person called, and what happened. Keep each list short enough that a manager can hold it in their head — roughly a dozen reasons and half a dozen outcomes. Categories must be mutually exclusive within an axis, and there has to be an unclassified bucket. If that bucket grows past a small share of calls, the taxonomy is wrong, not the calls.
Most taxonomies fail by mixing two questions
The common mistake is a single list containing entries like new AC install, booked, spam, price shopper and existing customer. Those are four different kinds of fact. A call can be a new install request that booked from an existing customer, and that single call has no correct home in that list.
Split it. Reason answers what the caller wanted. Outcome answers what the business did. Caller type answers who they were. Each is its own field, each is independently reportable, and together they answer questions a merged list cannot.
Short lists beat complete ones
There is a strong pull toward granularity — separate categories for every service line, every equipment type, every scenario. Resist it at the reporting layer. A taxonomy exists so someone can look at a weekly number and decide something. Thirty categories means every one of them is too small to move visibly, and nobody remembers what they mean.
Keep the reportable list tight and push detail into tags or attributes underneath. You can always roll up from detail; you cannot recover a decision from a report nobody reads.
The unclassified bucket is your quality meter
Every classification system needs somewhere to put calls it could not confidently place. Systems without one appear to perform better and actually perform worse, because uncertain calls get forced into the nearest category and quietly corrupt the counts.
Watch the size of that bucket over time. A stable small share is healthy. A rising share usually means one of three things: audio quality dropped, a new service line or campaign started producing calls the taxonomy has no name for, or someone changed the phone tree. All three are worth knowing about, and all three are invisible if uncertain calls are silently absorbed.
Changing the taxonomy is not free
Once you report on categories, the category definitions become part of your history. Renaming or splitting one breaks year-over-year comparison unless the change is applied retroactively to stored calls. That is doable when transcripts are retained and reclassification can be re-run, and impossible when only the label was kept.
This is a good reason to keep raw transcripts alongside labels in your intelligence platform, and a good reason to decide the taxonomy with the people who will be held to it. The categories that survive are the ones a service manager and a marketing manager both agreed to, which is also how marketing reporting stops contradicting operations reporting.
Topics: call categorization · taxonomy · reporting · call reason
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.