Business Process Audit
Automating a bad process makes it faster and still wrong.
What is process audit?
A process audit measures how a business process actually runs, step by step, showing where time is spent working, where it is spent waiting, and how often work goes backwards, so that decisions about automation are made on measurements rather than impressions.
Most organisations know a process is slow without knowing which step is slow. Asking people does not settle it, because everybody reports how long their step takes when it goes well, and the delay lives between steps rather than inside them.
Measurement usually finds the same thing. The expensive part is waiting rather than working, and waiting is frequently fixed by changing when something happens or who it goes to next, which costs nothing and needs no technology.
The other common finding is rework. Work going backwards is invisible on a process map and doubles the cost of every step it passes through twice, which makes it the most underestimated number in most operations.
Why many of our findings cost nothing
Duara measures business processes in Kenya, usually for organisations deciding whether to automate one.
A good proportion of our findings need no technology at all. Recommending a scheduling change rather than software is a smaller sale, and it is what the measurements frequently support.
Who this is for
- A process is slow and nobody knows why
- You are about to automate something
- Different people describe the process differently
- A previous automation did not deliver what was expected
Industries we serve
- Manufacturing
- Financial services
- Logistics and distribution
- Healthcare administration
- Professional services
- Public sector operations
Problems this solves
These are the reasons process decisions get made on impressions.
- Nobody knows which step is slow. The process is known to be slow as a whole, and improvement effort spreads evenly across steps that are not the problem.
- Waiting mistaken for working. Time between steps is invisible in most reporting, and it is usually the largest single component.
- Rework not counted. Work going backwards passes through steps twice and appears nowhere, which makes it the most underestimated cost in an operation.
- Steps nobody can explain. A check added because a system was unreliable years ago outlives the system, and everybody assumes somebody else knows why it is there.
What we deliver
A measured map of the process as it actually runs, separating work from waiting, counting rework, and identifying which steps are worth automating and which should be removed instead.
A measured map of the process as it really runs
Time split between working and waiting
How often work goes backwards
Steps that exist for reasons that no longer apply
Which steps are worth automating and which are not
Fixes that need no technology at all
Features
- Measured rather than described. Self reported timings capture the good case, and the delay usually lives in the case nobody thinks to describe.
- Waiting separated from working. The distinction changes the answer completely, since waiting is fixed by sequencing rather than by speed.
- Rework counted explicitly. Work passing through a step twice is invisible on a map and doubles what that step costs.
- Obsolete steps identified. Checks added for systems that were replaced years ago survive because nobody remembers they were the reason.
- Remove before automating. Automating first locks in the current shape, including the parts that should not exist at all.
Benefits
- Effort goes to the real bottleneck. Knowing which step is slow prevents improvement spread evenly across steps that were never the problem.
- Some fixes cost nothing. Sequencing and handover changes remove waiting without any software being bought or built.
- Automation aimed correctly. Automating a measured bottleneck delivers what automating an assumed one does not.
- Disagreements settle. Where people describe a process differently, measurement replaces an argument that seniority would otherwise decide.
What gets measured
- Step duration. How long each step actually takes, measured rather than estimated by whoever performs it.
- Waiting time. How long work sits between steps, which is usually the largest component and the least visible.
- Rework frequency. How often work goes backwards, and which step sends it there.
- Handover points. Where work changes hands, since that is where most waiting is created.
- Variation. How differently the process runs on a bad day, since the bad day is what people are complaining about.
- Step justification. Why each step exists, which occasionally establishes that nobody knows.
Where a process leaves no trace, measurement has limits and we say what those are. Observation and sampling can establish a great deal, and they cannot produce the precision that recorded timestamps would.
How the process works, step by step
A process audit usually takes two to four weeks, depending on how long one instance of the process takes to complete.
-
Step 1
Scoping
We agree which process is being measured and where it starts and ends, which is less obvious than it sounds.
-
Step 2
Observation and sampling
Watching the work happen and sampling records, since how it runs differs from how it is described.
-
Step 3
Measurement
Duration, waiting, rework and variation recorded rather than estimated.
-
Step 4
Findings ordered
What to remove, what to simplify and what is genuinely worth automating, in that order.
-
Step 5
Report delivered
A measured map with the no technology fixes identified separately, since those are usually available immediately.
Quality assurance and how we work
The commitments attached to measuring how people actually work.
- We recommend non technical fixes when they are the answer. A scheduling change that removes waiting is a smaller sale than automation and frequently what the measurements support.
- We say remove before automate. Automating first locks in the current shape of a process, including the steps that should not exist.
- We state the limits of the measurement. Where a process leaves no trace, sampling establishes a great deal and cannot produce the precision timestamps would.
- We do not use measurement to assess individuals. The audit measures the process. An exercise people believe is about their performance stops producing honest information.
- We report steps nobody can justify. Including ones somebody senior introduced, since a step whose reason has expired costs the same as any other.
Frequently asked questions about process audit
Why not just ask staff how long steps take?
Because everybody reports the good case, and the delay usually lives between steps rather than inside any of them.
What does measurement usually find?
That the expensive part is waiting rather than working, which changes the fix entirely since waiting is a sequencing problem.
Can waiting be fixed without software?
Frequently yes. Changing when something happens or who it goes to next costs nothing and removes the largest component.
Why does rework matter so much?
Work going backwards passes through steps twice and appears nowhere in reporting, which makes it the most underestimated cost in most operations.
Should we automate before or after measuring?
After. Automating first locks in the current shape of the process, including the parts that should not exist at all.
What if automation is not the answer?
We say so. A scheduling change is a smaller sale than software and it is what the measurements frequently support.
Why do people describe the process differently?
Usually because it genuinely runs differently depending on who is doing it, which is a finding rather than a communication problem.
What about steps nobody can explain?
They are common. A check added because a system was unreliable years ago survives long after the system was replaced.
How is the process observed?
By watching the work happen and sampling records, since how a process runs and how it is documented are different things.
Do staff know they are being observed?
Yes. An audit run covertly produces worse information as well as being the wrong way to treat people.
Is this measuring individual performance?
No, and we say so plainly. An exercise people believe is about them stops producing honest information immediately.
What if the process leaves no records?
Sampling and observation still establish a great deal. We state what the measurement can and cannot support rather than implying precision.
How long does a process audit take?
Two to four weeks, depending mostly on how long a single instance of the process takes to complete.
Can it cover a process across departments?
Yes, and those are usually where the waiting is worst, since handovers between departments create the longest gaps.
Does it help after a failed automation?
Often. A common finding is that the automated step was never the bottleneck, so speeding it up changed nothing overall.
What is the output?
A measured map with findings ordered as remove, simplify, then automate, and the no technology fixes listed separately.
Can our own team act on it?
Yes, and the immediate fixes usually need nobody external, which is part of why they are separated out.
Do you measure the bad days?
Yes, deliberately. Variation matters because the bad day is usually what prompted somebody to ask for the audit.
Who sees the measurements?
Whoever you nominate, and findings describe the process rather than naming individuals.
What does a process audit cost?
Process length and the number of handovers determine it, quoted after scoping.
Glossary of terms
Process measurement terms in plain language.
- Cycle time
- How long a whole process takes from start to finish, including all the waiting.
- Working time
- The portion actually spent doing something, which is usually far smaller than the cycle time.
- Waiting time
- Time work spends sitting between steps, generally the largest and least visible component.
- Rework
- Work sent backwards to be done again, which doubles the cost of every step it revisits.
- Handover
- A point where work changes hands, and where most waiting is created.
- Variation
- The difference between a good day and a bad one, which is what people are usually complaining about.
Related services in AI Audit Services
Talk to us about process audit
Ask three people to describe the same process. If you get three different answers, you do not have a process yet.
Start your project