DMAIC in five steps and one real case. Each step says what to do, shows it on the warehouse floor, links the tool to use and ends with three questions. Start your own project with the charter at the bottom.
Throughout the course
The case: outbound damage at a parcel hub
A hub ships about 18,400 units a week. Customers complain about crushed parcels, and the damage rate is almost twice what was promised. Follow one improvement project through all five steps, with the same numbers you can load as examples in the tools.
D
Step 1 of 5 · Output: A project charter
Define
Agree what problem you are solving, for whom, how you will measure it and what is out of scope.
Key ideas
Voice of the customer → CTQ
Start from what the customer notices, then turn it into something you can measure: the Critical to Quality requirement.
SIPOC
One line of boxes: Suppliers, Inputs, Process (five to seven steps), Outputs, Customers. It sets the borders of the project before anyone argues about details.
Problem statement
What, where, since when, how much, and the gap to the target. No cause and no solution yet: those come later, from data.
Common mistakes
Writing the solution into the problem: "we need more training".
A scope so wide that the project never ends.
A problem without a number, so nobody can say when it is solved.
In the case
Problem
In weeks 36 to 38, 0.57% of outbound units left the hub damaged (104 of 18,400 in week 38) against 0.30% promised to the customer. Credits cost about €1,150 a week.
CTQ
The parcel arrives undamaged: no crushed, torn or wet outer carton.
Scope
In: from the pick face to handover to the driver. Out: linehaul and supplier packaging.
Target
0.30% or lower by week 46, without longer loading times.
Tools for this step
KPI Diagnostic →to check whether the number is clear to everyone
Find out how the process performs today, with data you can trust, before you change anything.
Key ideas
Operational definition
Write exactly what counts, with photos. "Damaged" means different things to different people until it is written down.
Data collection plan
What is recorded, where, by whom, how often and in which format. Record the stage and cause at the moment a problem is found.
Check the measurement
Let two people judge the same twenty cases. If they disagree, fix the definition before you trust the data.
Baseline
The current level as a rate, DPMO and sigma level, over two to four weeks.
Common mistakes
Changing the process while you are still measuring it.
Trusting a number that two people would count differently.
Measuring only at the end of the line, so you never see where it starts.
In the case
Definition
A unit is damaged when the outer carton is crushed, torn or wet enough for the customer to notice, found anywhere from the pick face to driver handover.
Check
Two shift leads judged the same 20 cartons and agreed on 17. Three were "slightly dented": the definition got a photo of what counts and what does not.
Find the few root causes behind most of the problem, and prove them with data rather than opinions.
Key ideas
Pareto
Sort causes by size. Usually two or three carry about 80% of the problem: the vital few.
Stratify
Split the data by stage, shift, hour, product or route. Problems hide in averages.
Go and see
Watch the work where the number is produced, at the time it goes wrong.
5 Whys and the fishbone
Ask why until you reach something in the process you can change. The fishbone checks people, method, machines, material, measurement and environment.
Common or special cause
A control chart tells you whether you are looking at a normal level of the process or at a few unusual days.
Common mistakes
Stopping at a person: "the loader was careless".
Fixing the loudest complaint instead of the largest bar.
Treating an opinion as a root cause without checking it in the data.
In the case
Pareto
Stacking 56%, load securing 14%, handling 11%: three causes carry 81% of the damage.
Where and when
63% is found at loading, 60% on the late shift, most of it in the 11:00 departure wave.
On the floor
At the late wave, loaders build walls fast and put heavy parcels on top to fill gaps. There is no stacking rule at the dock.
Proof
Damage per 1,000 units is 2.4 times higher in the peak loading slots than in the others. The control chart shows no special days: the level comes from the method.
Design, test and prove a countermeasure that removes the root cause.
Key ideas
Options, then choice
Collect several countermeasures and pick by impact and effort. The first idea is rarely the best.
Poka-yoke
Make the error hard to make: layout, markings, fixtures and defaults instead of reminders.
Standard work
Write the new way as a short, visual standard that a new person can follow on day one.
Pilot
Test small, on one dock or one shift, for long enough to see the effect. Plan, do, check, act.
Common mistakes
A reminder at the briefing instead of a change to the work.
Rolling out everywhere before the pilot proves anything.
Ignoring side effects on other numbers, such as loading time.
In the case
Countermeasures
A stacking picture at every dock (heavy low, fragile high, no gaps), floor zones in the vehicle for heavy parcels, load bars after the last wall with a photo, and ten more minutes for the peak loading slots.
Pilot
Docks 3 and 4, two weeks, same definition and same log.
Result
Damage fell from 0.57% to 0.31%. The control chart shows ten days in a row below the old centre line. Loading time rose by two minutes per route: accepted by the sponsor.
Tools for this step
Floor checks →in Damage Control, to see whether the standard holds
Step 5 of 5 · Output: A control plan and a handover
Control
Keep the gain: make the new way the standard, watch it, and hand it over to the process owner.
Key ideas
Standard and training
Put the new way into the SOP and into onboarding, so it does not depend on the project team.
Control plan
What is measured, how often, by whom, and what to do when a signal appears: the reaction plan.
Control chart
Recalculate the limits from the new stable level and watch for signals, not for single bad days.
Close the project
Hand over to the process owner, record the savings and the lessons, and celebrate the team.
Common mistakes
Declaring victory when the pilot ends.
Keeping the old limits on the chart, so no signal ever shows.
No owner after the project team leaves.
In the case
Standard
The stacking standard is in the SOP and the onboarding, with the dock pictures.
Control plan
Daily damage on a p-chart with limits from the pilot weeks. A day above the upper limit: the shift lead checks the same shift and runs 5 Whys within 24 hours. A weekly floor check at the dock.
Result
0.31% at 18,400 units a week: about 47 fewer damaged units and €500 less in credits a week. The outbound manager owns it from week 47.