5.1 Customer Identification and Segmentation
Key Takeaways
- External customers receive the organization's output; internal customers are downstream processes that receive work from the process in scope.
- Segmentation dimensions include demographic, geographic, behavioural, value, product or service line, and channel.
- An averaged requirement across heterogeneous segments describes no real customer and hides bimodal demand.
- The end user is not always the buyer, and neither is always the decision maker; all three may impose different requirements.
- ASQ tags customer identification at the Apply level, so the exam expects you to select the segmentation that resolves a described conflict.
Who is the customer?
The Body of Knowledge asks Black Belts to identify and segment customers and to show how a project will impact both internal and external customers. Three distinctions carry most of the exam weight.
External customers receive the organization's product or service and pay for it directly or indirectly. Internal customers are the downstream processes, departments, and colleagues who receive the output of the process in scope. Every process has at least one internal customer unless it is the last step before delivery.
Buyer, user, and decision maker are frequently different people with different requirements. A hospital purchasing department buys infusion pumps on total cost and service terms; nurses use them and care about alarm behaviour and setup time; the biomedical engineering director approves them on maintainability and integration. A project that optimizes only for the buyer produces a pump nobody wants to use.
Beyond these, a full customer map for a Six Sigma project usually includes:
| Category | Who | Typical requirement |
|---|---|---|
| End user | The person who uses the output | Fitness for purpose, ease of use, reliability |
| Purchaser | The person who pays | Price, total cost of ownership, terms |
| Specifier / influencer | Engineer, clinician, architect, IT | Conformance to a standard or specification |
| Internal downstream process | The next operation | Complete, correct, on-time input |
| Channel partner | Distributor, retailer, installer | Packaging, lead time, documentation |
| Regulator | Agency, notified body, accreditor | Compliance evidence, traceability |
Why segmentation is not optional
A segment is a group of customers whose requirements are similar enough that one specification serves them and different enough from other groups that a single specification would not.
The reason to segment is that averages hide structure. Suppose an insurer measures claim cycle time and finds a mean of 11.4 days with a large standard deviation. Segmented, the picture is different: simple auto claims average 3.1 days, complex bodily-injury claims average 34 days, and there are almost none in between. The distribution is bimodal, the "average customer" does not exist, and any target set on the aggregate mean is wrong for both groups.
The same logic drives the analytical practice of stratification, which is the Measure and Analyze phase expression of segmentation. If you have not segmented customers in Define, you will not know which stratification factors to collect in Measure.
Segmentation dimensions
| Dimension | Examples | When it discriminates |
|---|---|---|
| Demographic / firmographic | Age, income; industry, company size, sector | Requirements differ by who the customer is |
| Geographic | Region, country, urban versus rural, climate | Regulation, logistics, or usage conditions differ |
| Behavioural | Usage frequency, order size, order pattern, channel used | The way the customer transacts drives the process load |
| Value | Revenue, margin, lifetime value, strategic importance | Resource allocation must be differentiated |
| Product or service line | SKU family, service tier, contract type | Different value streams serve them |
| Needs-based | What the customer is trying to accomplish | The most useful and the hardest to obtain |
Two practical tests for whether a segmentation is useful:
- Do the requirements actually differ across segments? If two segments want the same thing to the same tolerance, the split adds cost without insight.
- Can you tell which segment a transaction belongs to at the moment you process it? A segment you cannot identify in the operational data is a marketing construct, not a process input.
Prioritizing segments
Not every segment gets equal weight in the project. Two common approaches:
- Value ranking. Rank segments by revenue, margin, or customer lifetime value. Pareto behaviour is normal: a minority of customers usually generates most of the value.
- Strategic weighting. A small segment may be weighted heavily because it represents a target market, carries reputational risk, or is a reference account.
Record the weighting decision in the charter. It determines whose requirement wins when requirements conflict, and that argument is much easier before the data arrives than after.
Showing the impact on internal and external customers
The Body of Knowledge asks specifically that you show how a project will impact both. A short impact table in the charter does this and prevents the classic sub-optimization failure:
| Customer | Internal or external | Current experience | Expected impact of the project | Risk |
|---|---|---|---|---|
| Assembly cell 2 | Internal | Receives 4% of kits short | Kits complete; pick sequence changes | Operators must learn new kit layout |
| Field service | Internal | Handles rework of miswired units | Fewer callouts | Loses visibility of failure modes unless data is shared |
| Fleet customers | External | 6-day average lead time | 3-day lead time | Batch size reduction may raise unit price |
| Retail customers | External | Unaffected | No change | None |
The row that matters most is usually the internal one with a risk entry, because internal customers absorb the friction of a change without having asked for it.
A worked identification
A Black Belt is chartered to reduce errors in outpatient prescription dispensing.
- External customers: the patient (accuracy, wait time, clear instructions), the prescribing physician (formulary substitution notification), and the insurer (correct coding).
- Internal customers: the pharmacy technician who receives the entered order, the billing team that receives the claim, and the inventory function that receives the consumption signal.
- Segments that matter: new prescriptions versus refills (completely different error profiles), controlled substances versus routine (different verification requirements), and walk-in versus mail-order (different time requirements).
Without that segmentation the team would compute one error rate over everything and lose the fact that almost all serious errors occur in new prescriptions for controlled substances -- a segment that is under 5% of volume.
An insurer finds claim cycle time has a mean of 11.4 days and a large standard deviation. Segmentation shows simple auto claims at 3.1 days, complex bodily-injury claims at 34 days, and almost nothing in between. What does this demonstrate?
A team segments customers by a behavioural attribute that cannot be determined from any field in the transaction system at the time an order is processed. What is the practical problem?
A hospital purchasing department selects infusion pumps on total cost and service terms, but nurses find setup slow and alarms confusing. Which customer-identification principle was violated?