Small businesses, side projects and household budgets share an irritating feature: the same kind of problem keeps resurfacing, and each time it feels brand new. An invoice goes out with a wrong detail. A delivery arrives a day late. A booking clashes with another. Each incident is handled, apologised for and forgotten. This step-by-step method shows you how to stop treating each as a one-off and how to find the recurring cause using nothing but a short log and a little discipline.
The approach is evidence-led. You will not guess at causes; you will collect a handful of observations, compare them, and let the comparison point you toward the fix. The goal is to repair the underlying cause once rather than patch the symptom every time it returns.
Step 1: Decide what counts as the same kind of problem
Before you record anything, write a one-line definition of the problem you want to watch. “Customer complaints” is too broad. “Orders that ship with a wrong address” is specific enough that two people would agree whether an event belongs on the list.
A good definition has three properties: it describes an outcome, not a cause; it can be judged yes or no; and it is narrow enough that you could realistically see it three times within a couple of months. If you cannot meet the last condition, pick something that happens more often. Patterns need repetition to become visible.
Step 2: Create a log with only four columns
Open a sheet or a notebook and give it four columns: date, what happened, which step in the process it occurred at, and what else was going on that day. That is all. Every extra column lowers the chance that you will keep filling it in.
The third column matters most. Problems that look different on the surface often cluster at a single step, such as data entry, handover to a courier or final approval. You will not see that clustering unless you record where each event happened.
Step 3: Log every event for a fixed period
Choose a window, say six weeks, and record every qualifying event during that time, including the small ones you would normally ignore. Resist the urge to analyse as you go. Early analysis tends to lock you into an explanation after two entries, and two entries are not evidence.
Stay even-handed in what you write down. Should you only log the dramatic cases, the sample will be skewed toward whatever was most memorable instead of whatever was most common.
Step 4: Wait for the third occurrence
This is the rule that keeps the whole method honest. One occurrence is noise. Two might be coincidence. Three of the same kind is the first point at which it is reasonable to say “this is a pattern worth investigating”.
Waiting is uncomfortable, especially if the problem is annoying. But acting after the first event usually fixes the wrong thing, and acting after the second often does the same. If you feel pressure to act sooner, make a temporary workaround and keep logging while it is in place.
Step 5: Sort the entries and count
When the window closes, group entries by the step column. Count how many fall under each step. Then group them again by the “what else was going on” column and count again. You are looking for lopsidedness: one step or one circumstance that accounts for far more entries than the rest.
Keep this analysis modest. With a few dozen entries you cannot calculate anything fancy, and you do not need to. If seven of ten events sit under one step, that is a clear enough signal to act on. If they spread evenly across every step, the cause may be something broader, such as workload or training, and the log has still told you something useful.
Step 6: Write the pattern as a testable sentence
Put your finding into a single sentence that events might later prove mistaken. For example: “Most wrong-address orders were entered during the Friday afternoon batch.” A testable sentence names a place, a time or a condition, and it predicts what should happen if you change it.
Avoid vague summaries such as “we need better communication”. They sound reasonable and cannot be checked. A sentence you can disprove is a sentence that can guide you.
Step 7: Change one thing and keep logging
Pick a single change aimed at the pattern. If the problem clusters at Friday batches, add a second look before sending, or move the batch earlier. Then continue the log, same columns and same definition, for another stretch of time.
Changing only one thing is what makes the result interpretable. If you alter five things at once and the problem disappears, you will never know which one mattered, and you cannot carry the lesson to the next situation.
Step 8: Judge the result honestly
Compare the new window with the old one. Did the target step still account for most of the entries? Did the total count fall? Did a different step suddenly grow, suggesting the problem moved rather than vanished?
If the change worked, keep it and write down why. If it did not, you have ruled out one explanation, which is progress, and you can return to the sorted entries and test the second most likely cause.
Where this method fits in ordinary business life
The same eight steps apply well beyond one workplace. Money is a natural place to use them. Recurring overdrafts, late fees and forgotten subscriptions are classic repeating events, and the guidance in our piece on modern cash flow and the leaks that quietly eat into reserves pairs well with a log like this one, since the log tells you where the leak is and the guide helps you plug it.
Travel planning is another. If you coordinate trips for a team or a family, the same sort of delay probably shows up repeatedly. Anyone organising long drives will find parallels in the long guide to hiring a vehicle and driving across Canada, where collection, licence checks and underestimated distances are the spots that keep tripping people up.
A note on bias
Even a well-kept log can mislead, so keep a few cautions in mind:
- Memory bias. Dramatic events feel more frequent than they are. The log corrects for this only if you record the boring ones too.
- Observer bias. Once you start watching for a problem you will notice it more. Compare with a baseline period if you can.
- Confirmation bias. It is tempting to sort entries in a way that supports your first theory. Do the grouping before you have a theory.
- Small samples. Ten entries can point in a direction, but they cannot prove anything. Treat the result as a lead, not a verdict.
Borrowing techniques from software teams
Engineers who run online services have turned this exact habit into routine practice. They bundle similar failures, examine what the group shares and fix that shared cause. Curious how that thinking gets taught? Engineers can wander through DevOpsSchool.org, a community-run site that welcomes anyone by its own account and keeps study routes, role maps, practice labs, a glossary and a tool listing under one roof. Non-technical readers will not need it, though the underlying idea of grouping repeated failures translates cleanly.
A small-shop illustration
Picture a shop owner notices that customers occasionally receive the wrong item. In the first week there is one case. In the second week, another. By the fifth week there are four. The log shows that three of the four happened on days when a temporary helper packed orders, and that all four involved products stored on the same shelf.
Looking only at the first case would have suggested a careless mistake. Looking across all four reveals a shelf with look-alike products and a handover that depends on memory. The fix is a label change and a short packing checklist for helpers, a single adjustment that removes the whole class of error.
Notice what was not needed here: no new software, no long meeting, no blame. Just a small record and the patience to read it as a set.
Common mistakes to avoid
- Logging too many things at once. Pick one problem at a time.
- Stopping the log early. If you quit as soon as things improve, you will not know whether the improvement lasted.
- Treating the fix as final. Keep a light version of the log going, because patterns can return.
- Skipping the “what else” column. Context often explains the cluster.
Summary
Define the problem, log it in four columns, wait for the third repeat, sort and count, write a testable sentence, change one thing, and check the result. Those steps convert a stream of irritating surprises into a short list of causes you can fix. It is a modest routine, but it swaps repeated firefighting for one proper repair, which pays back far more than reacting each time.

