Business Process Improvement for Small Business: One Process, Fixed by Hand First

Step 1: pick one process that moves money

Every company that sells work runs some version of the same chain: inquiry, quote, contract, delivery, invoice, payment. Each link is a process, and each hand-off between links is a place where work can sit with nobody acting on it. Process improvement in a small company should start on this chain, because a delay here is a delay in cash, and a miss here is lost revenue rather than lost tidiness.

Choose the link where the pain is already felt and an example from last month comes to mind without effort: the estimate that never got a follow-up, the job finished two weeks before it was billed, the invoice nobody chased. Resist choosing three. One process improved and measured teaches the method; three changed at once make it impossible to tell which change did what.

Leave internal tidiness — the shared drive, the meeting format — for a later round. The first round needs a result that shows up in the bank balance.

Step 2: write down how the process actually works today

Take one real, recent example and follow it through the records from start to finish: the call log, the email thread, the quote, the job ticket, the invoice, the payment. Write each step as a line — what happened, who did it, in which tool, and the date and time. Then do the same for two or three more examples, including at least one that went badly.

This is a plain-paper version of what Lean practitioners call value stream mapping, which the U.S. Environmental Protection Agency describes as a process-mapping method “for understanding the sequence of activities and information flows used to produce a product or deliver a service.” The full method adds standard icons, timelines and a designed future state. For a small company, the useful part is the discipline: the map is drawn from what happened, not from how the process is supposed to work.

Expect the written version to disagree with the version in everyone’s head: a step everyone assumed was automatic turns out to depend on one person remembering. Those disagreements are the findings. Write both versions down.

Step 3: find where work waits, and who owns each step

With the steps on paper, add two columns. In the first, write how long the work sat before that step started — the gap between the timestamps. In the second, write the name of the person who owns the step: not a department, not “the office,” a person.

Work time and wait time are different things. Writing a quote may take forty minutes; the finished quote may then sit four days before anyone sends it. An improvement aimed at the forty minutes leaves the four days untouched. The long waits are worth hunting for at the hand-offs, where one person finishes and the next is supposed to start, because that is where nobody is watching the clock.

• What event starts this step — and does the next person find out on their own, or only when someone remembers to tell them?
• Who owns it by name, and who covers when that person is out?
• What rule decides it — or does it depend on asking the owner?
• What does the person need to do the step, and how often does it arrive incomplete?
• What happens when the step is skipped — does anyone notice, and how soon?

The step with the longest wait and the blankest owner column is the first thing to fix.

Step 4: fix ownership and rules before tools

What the map reveals is often not a technology problem. A step with no owner needs an owner. A step that waits for the owner’s yes needs a written rule about what no longer needs that yes. A step that arrives incomplete needs a checklist at the step before it. A step done three ways by three people needs one agreed way, written down and tried for two weeks. None of these costs anything but a decision.

The order matters. Automating a step before fixing it makes the broken version faster and harder to change. Michael Hammer made the point in Harvard Business Review in 1990, writing that companies “tend to use technology to mechanize old ways of doing business.” His answer for large corporations was radical redesign. In a small company the same lesson is smaller and cheaper: decide who owns the step and what the rule is, and only then let a tool carry it.

Write the new version as a short procedure — the trigger, the owner, the steps in order, the exception path — and keep it where the work happens. A procedure stored three folders deep does not get opened in a busy moment; the guide on how to write an SOP that gets followed covers the format.

Illustrative example, with hypothetical round numbers: a service company maps its job-to-invoice step and finds that finished jobs wait an average of nine days to be billed. The map shows why. Completion notes arrive by text to whoever is around, nobody owns turning them into an invoice, and the owner reviews every invoice before it goes out. The fix is three decisions, not software: the office manager owns invoicing, completion notes go into one form the same day, and invoices under a set amount go out without the owner’s review. If the average wait four weeks later is two days, the improvement is real. If it has not moved, the map was wrong about where the wait lives, and it gets redrawn.

Steps 5 and 6: measure before and after, then decide what to automate

A change without a baseline can only be liked or disliked. Before changing anything, measure the process for two to four weeks from the records. After the change, measure the same things for the same length of time.

• Volume — how many times the process ran in the period.
• Wait time — the gap between the event that starts a step and the moment someone acts on it.
• Misses — instances never handled at all: the quote never followed up, the job never invoiced.
• Rework — instances that had to be corrected, re-sent or apologized for.
• Money waiting — the dollar value sitting at the slow step on a given day.

Only now does automation become a sensible question, and it is a narrower one than it was at the start: which part of the improved, owned, written-down process is frequent, rule-based and repetitive enough to hand to software? A reminder when a quote has had no reply, an invoice drafted the moment a job is marked complete, a payment reminder on the due date — these make good candidates because the rule already exists and someone owns the exceptions. The guide to what to automate first in a small business runs each candidate through a five-question filter, and the business process automation page describes how we build once the numbers justify it. AI can be one of those tools, but it is never the starting point.

Lean, Six Sigma or a process improvement consultant — and when to hire no one

Lean and Six Sigma are serious methods with long track records in manufacturing and large service operations, and a small company can borrow from both without adopting either. Lean focuses on flow and on removing waste — waiting, rework, handling that adds nothing the customer pays for. The EPA describes Six Sigma as “a set of statistical methods for systemically analyzing processes to reduce process variation,” run through a sequence called DMAIC — Define, Measure, Analyze, Improve, Control — with improvement efforts led by specialists known as black belts. It also notes that a Six Sigma quality level equates to about 3.4 defects per million opportunities.

That last figure shows where Six Sigma fits: high-volume, repeatable work where defects are counted in the thousands and small variation is expensive. A company that sends a few dozen invoices a month does not have the volume for statistical process control to reveal anything a column of timestamps cannot. Belt training, statistical software and a multi-month program are overkill at that size. What carries over is the habit: define the problem, measure before changing anything, find the cause, change one thing, and hold the gain.

A process improvement consultant is worth hiring when the map cannot be drawn from inside — the owner and the team disagree about where work waits, the records are spread across several tools, or nobody has the hours to do the reading. Whoever is hired should work from the company’s own records, put a number on each problem, and hand over a plan the company owns.

Hire no one when one process is plainly broken and the fix fits in a sentence; fix it by hand this month. Hire no one when there are no records to read; start capturing the timestamps first. And hire no one before putting a rough number on the problem; if the number is small, keep the money.

If the number is not small and the map will not draw itself, the Revenue Review does this reading across the whole chain in two weeks, with read-only access to the company’s own records. It ends with a ranked list of what to fix by hand, what to automate and what to leave alone; any build is scoped and priced separately. The 20-minute call is free, and part of its job is to say when the Review is not worth it.