Optimizing a process is a sequence, not a single fix: map how the work actually runs today, find the bottlenecks, handoffs, and rework, quantify the cost and cycle time of each step, prioritize fixes by impact versus effort, redesign the flow, automate the parts that are now stable, and measure against the baseline. The order is what makes it work — automating before you map just makes a bad process run faster and locks in the waste.
Key takeaways
- Map reality, not the diagram. Document the process as it actually runs, including the workarounds people use, before you change anything.
- Wait time is the real enemy. Work sitting in queues and handoffs almost always dwarfs actual processing time — attack queues first.
- Quantify before you prioritize. Put a cost and a cycle time on every step so you can rank fixes by impact versus effort instead of by who shouts loudest.
- Optimize, then automate. Automating a broken process locks in the waste; simplify the flow first, then automate the stable, repetitive parts.
- Measure against a baseline. Track cycle time, cost per unit, and rework rate before and after, or you can't prove the improvement.
The seven-step playbook
Every durable process improvement follows the same sequence, and skipping a step is the most common reason gains don't stick.
- Map the current process. Walk the workflow end to end with the people who run it. Capture every step, decision point, and handoff — including the undocumented workarounds. The map you want is reality, not the official diagram.
- Find bottlenecks, handoffs, and rework. Mark where work waits, where it crosses a team boundary, and where it loops back for correction. These three patterns account for most of the lost time.
- Quantify cost and cycle time. For each step, record processing time, wait time, and the fully loaded cost of the person doing it. Sum them to get total cycle time and cost per unit of work.
- Prioritize by impact versus effort. Plot each fix on a simple grid: high-impact, low-effort changes go first. Avoid the trap of starting with the most interesting problem instead of the most valuable one.
- Redesign the flow. Remove steps that add no value, collapse redundant approvals, and reduce handoffs. The goal is fewer steps and shorter queues, not a more elaborate process.
- Automate the stable parts. Once the flow is simplified and repeatable, automate the rules-based, high-volume steps — data entry, routing, notifications, reconciliations. Leave judgment-heavy steps to people.
- Measure against the baseline. Track the same three metrics for 4–8 weeks after launch and compare to the pre-change numbers. Keep what moved the metric; revert what didn't.
Common bottlenecks and how to fix them
Most process waste falls into a handful of recognizable patterns, each with a standard fix.
| Bottleneck | Symptom | Typical fix | | --- | --- | --- | | Manual approval | Work sits in an inbox waiting for sign-off | Auto-approve under a threshold; escalate only exceptions | | Team handoff | Tasks queue between departments | Shared queue, clear SLA, single owner per item | | Re-keying data | Same data typed into multiple systems | Integrate systems or sync with low-code automation | | Rework loop | Items bounce back for correction | Fix the root cause upstream; add a validation gate | | Single-person dependency | Flow stalls when one person is out | Document the rule, cross-train, or automate the step |
A useful rule of thumb: wait time is usually 5–10x the processing time. A task that takes 10 minutes of work often takes two days to complete because it spends most of that time in a queue. That is why attacking handoffs and approvals beats speeding up the work itself.
Quantify, then prioritize — a worked example
The point of measuring every step is to make prioritization objective instead of political.
Say an invoice-approval process runs 800 invoices a month across five steps. After mapping and timing each step, the numbers look like this:
- Total cycle time: 3.2 days per invoice, of which only 18 minutes is actual processing.
- The largest single cost: a manual two-level approval that adds 1.4 days of wait time for invoices that are routine.
- Fully loaded cost per invoice: about $9, dominated by re-keying data between the procurement tool and the accounting system.
Two fixes rank highest on impact versus effort: auto-approving invoices under a set threshold (removing 1.4 days of wait for ~70% of volume) and a low-code integration that syncs the two systems so no one re-keys data. Together they cut cycle time to under a day and cost per invoice below $4 — measured against the baseline, not estimated.
The discipline is the same every time: don't redesign the step that annoys people most, redesign the one the numbers say is most expensive.
FAQ
What are the steps to optimize a business process?
Map the current process exactly as it runs today, find the bottlenecks, handoffs, and rework, quantify each step's cost and cycle time, prioritize fixes by impact versus effort, redesign the flow, automate the stable parts, and measure against the baseline. The order matters — automating before you map just makes a bad process run faster.
Should I automate a process before or after optimizing it?
Optimize first, automate second. Automating a broken process locks in the waste and makes it harder to change later. Once you have removed unnecessary steps, fixed handoffs, and stabilized the flow, automation delivers far more value and is cheaper to maintain. Map, simplify, then automate the parts that are now stable and repetitive.
How do I find the bottleneck in a process?
Measure the cycle time and the wait time of every step, then look for where work piles up. The bottleneck is usually a handoff between teams, a manual approval, or a step that depends on one person. Wait time — work sitting in a queue — is almost always larger than processing time, so target queues and handoffs before optimizing the work itself.
How do you measure process improvement?
Compare three metrics against a pre-change baseline: total cycle time end to end, fully loaded cost per unit of work, and the rework or error rate. Capture the baseline for 2 to 4 weeks before you change anything, change one thing at a time, and track the same metrics for 4 to 8 weeks after. A faster process that produces more errors is not an improvement.
Want to put numbers on your own process? Our process optimization work starts by mapping how the work really runs and quantifying where the time and cost go — before anyone redesigns or automates. If you're not sure which process to start with, talk to us and we'll help you find the one with the most slack to recover.