DMAIC is the improvement method from Six Sigma: Define, Measure, Analyse, Improve, Control. Textbooks describe each phase in a sentence, which is why it often sounds like common sense in fancy dress. The value shows up only when you run a real problem through it. So here is one, start to finish. The company, the process and every number below are invented for illustration.
Define: say what is wrong, for whom, and how you will know
A telecom client’s billing helpdesk, run by a BPO, receives about 10,000 first-time calls a week. The client complains that too many customers call back about the same issue. The team writes a problem statement that can be tested: “Within seven days of a first call, 18% of customers call again about the same issue, against a client target of 12%.” It also writes the scope (postpaid billing queries only, not network faults), the owner, and the goal (reach 12% in three months). A problem statement without a number and a boundary is the commonest reason these projects drift.
Measure: check that the number is trustworthy
Before accepting 18%, the team asks how it was calculated. The repeat flag is created by matching the customer’s number within seven days. Two problems appear. Customers who call about a different issue are counted as repeats, and customers calling from a second number are missed. Reviewing a sample of 200 flagged repeats by hand, the team finds that about one in ten was a different issue. They fix the rule by matching on the disposition code as well as the number. This is the real work of the phase: confirming the measurement before improving the thing measured.
Suppose, after fixing, the corrected figure still stands at 18%, which means 1,800 repeat calls a week. That is the baseline.
Analyse: find where the repeats come from
The team breaks the 1,800 repeats by the reason on the first call. It sorts the reasons from largest to smallest, a Pareto view. The result: about 40% of repeats, 720 calls a week, trace back to customers who were told their bill was correct but did not understand a pro-rata charge after a plan change. The remaining 60% is a long tail of small causes.
Now a why-chain. Why did they call again? The explanation was unclear. Why? Agents read the charge description from the screen, which says only “adjustment”. Why? The billing system labels it that way and the knowledge base has no worked example. Why? Nobody owns updating the article after the plan-change feature launched. The root cause is not agent skill. It is a missing explanation and a missing owner. Had the team jumped to “retrain the agents”, it would have spent a month on the wrong fix.
Improve: fix the cause, test small
Three changes are proposed: a knowledge article with a two-line worked example of a pro-rata charge, a short SMS to the customer after the call that repeats the explanation, and a request to the client to rename the bill line. The third needs the client’s change process, so it goes on a slower track. The first two run as a pilot on one team of 20 agents for two weeks, with a comparison team untouched.
Say the pilot cuts pro-rata repeats by 60%. Scaled up, 60% of 720 is 432 fewer repeat calls a week. Total repeats fall from 1,800 to 1,368, which is 13.68% of 10,000. Short of the 12% target, but a large step. If each call costs the client about ₹60 to handle, 432 calls is ₹25,920 a week, roughly ₹13.5 lakh a year. The team should say, honestly, that the remaining gap needs the next-largest causes tackled or the bill rename to land.
A caution: a pilot proves less than it appears to. Two weeks may not include a billing cycle peak, and the pilot team may be keener than average. State that when presenting the result.
Control: make the gain survive the project team
Improvements decay. The article gets edited, the SMS template is changed by someone who does not know why it exists, new agents join. The control phase assigns an owner for the article, adds the repeat-call rate for this reason code to the weekly dashboard with an alert threshold, and updates the training material. A review at 30, 60 and 90 days checks whether the rate holds. If it creeps back, the team knows early and knows where to look.
What the phases really guard against
Each phase blocks one familiar mistake. Define blocks vague goals. Measure blocks acting on a bad number. Analyse blocks solutions chosen before causes are known. Improve blocks big-bang rollouts. Control blocks quiet relapse. You can run a small problem through DMAIC in a week and a large one in a quarter, but skipping a phase tends to show up later as rework.
Figures and names in the examples are illustrative.
Leave a Reply