At the end of every call or chat, the agent picks one or more codes from a list: “Billing – bill amount query”, “Order – delivery delay”, “Technical – no internet”, “Sale – closed”, “Callback scheduled”. This is the disposition, also called a wrap-up code, call tag or reason code. It takes a few seconds and is often treated as admin. In practice it is the main source of information an operation has about why customers contact it at all.
What depends on those few seconds
Call recordings hold the full truth, but nobody can listen to 40,000 calls a month. The disposition is the summary that gets counted. Almost every analysis downstream uses it:
- Contact reason analysis. Which issues drive volume, and which are growing. A client deciding whether to fix its billing format or redesign its app needs to know how many calls each causes.
- Repeat contact and resolution. If a customer calls twice in seven days with the same reason code, that is a likely failure to resolve. Wrong codes make repeats invisible or invent false ones.
- Forecasting and routing. Volumes by reason feed staffing plans and decide which skills a queue needs.
- Outbound campaign management. In outbound, dispositions such as “no answer”, “wrong number”, “call back later”, “not interested” and “do not call” decide what the dialler does with each record next. A wrong code can mean a customer who asked not to be called gets called again.
- Billing. Some contracts pay per transaction type or per sale, so codes directly affect invoices.
- Compliance. Complaints, especially in regulated sectors such as banking and insurance, have reporting and timeline obligations. If a complaint is tagged as a “query”, it may never enter the complaint handling process.
How the data goes bad
Consider a telecom process where 22% of 40,000 monthly calls, about 8,800, are tagged “Others”. An analyst samples 300 of them and finds that 40% are actually billing disputes. Extrapolated, billing disputes hidden inside “Others” are around 8.8% of all calls, enough to change which problem the client fixes first. Nobody lied; the code list just made “Others” the easiest option.
The usual causes are predictable:
Too many codes. A list of 150 codes in a dropdown is not used precisely; it is used approximately. Agents learn five or six codes they can find quickly and use those for everything nearby.
Overlapping codes. If “Payment failed”, “Payment not reflected” and “Transaction issue” all exist, two agents will tag the same call differently. Trends then reflect who was on shift, not what customers needed.
Codes that describe the wrong thing. A list that mixes customer reasons (“bill high”), agent actions (“explained bill”) and outcomes (“resolved”) forces agents to choose between dimensions that should be recorded separately.
Time pressure. When after-call work is squeezed to protect handle time, agents pick the first plausible code. A shorter wrap-up looks efficient while making the data less trustworthy.
Incentive pressure. If a “Resolved” disposition counts toward a target, or a “Sale” code toward incentives, codes drift toward what is rewarded. This is where tagging moves from sloppy to misleading.
Designing a code structure that works
A practical design has a few properties.
Hierarchical. Two or three levels, such as category, sub-category and specific reason. “Billing → Bill amount → Roaming charges”. The agent narrows down step by step instead of scrolling a flat list. Reports can roll up to category level when detail is not needed.
Separate dimensions. Record the customer’s reason, the type of contact (query, request or complaint), and the outcome (resolved, escalated, callback, pending with customer) as separate fields. Then “billing complaints that were escalated” is a simple filter rather than a hunt through combined codes.
Mutually exclusive and collectively exhaustive. Each call should fit exactly one reason at each level, and almost every call should fit somewhere. A useful working rule some teams follow is to review the list whenever “Others” rises above a small single-digit share of volume.
Written definitions with examples. Each code needs a one-line definition and a sample scenario, kept in the knowledge base. “Use for: customer disputes a specific charge on the bill. Do not use for: customer wants an explanation of the bill format.”
Multi-reason handling. Customers often raise two issues. Decide whether the system allows multiple tags or a primary reason only, and train consistently. Otherwise half the floor tags the first issue and half the more serious one.
Keeping it accurate over time
Tagging accuracy should be a parameter in quality monitoring: the auditor listening to a call checks whether the disposition matches what happened. Monthly reviews of tag distributions by agent can also spot problems quickly. If one agent tags 35% of calls as “Others” while the team average is 6%, it is a coaching conversation, not a statistical mystery.
The list itself needs governance. New products, policy changes and system issues create new contact reasons. Without an owner who adds, retires and merges codes on a schedule, lists either freeze and lose relevance, or grow every time someone asks for a new code, until they are unusable.
Where speech analytics fits
Speech and text analytics can now classify contacts automatically from transcripts, and many operations use them to check or supplement manual tags. They are good at spotting volume patterns and phrases agents never tagged. They are not a reason to stop caring about the taxonomy: a model still needs a well-defined set of categories to classify into, and someone still has to decide what each category means. A clean disposition structure makes automated classification easier to train and easier to trust.
Treat the disposition as the operation’s memory of every conversation. If that memory is vague, every report built on it will be confidently wrong.
Written by the RP Strativa Team. Call volumes and percentages in the examples are illustrative.
Leave a Reply