The cost of undocumented process exceptions in offshore teams

Research findings

Every operating process has exceptions, and most teams handle them well enough in the moment. The cost appears later, when the person who handled the case is unavailable and the record shows only that the item was completed. An undocumented exception transfers knowledge from the company to an individual, and that transfer has a price even when it is invisible in the monthly accounts. That transfer is the reason an exception log belongs beside the queue rather than in a manager's memory.

This review defines an undocumented exception as a completed item whose path differs from the written procedure in a way that no record explains. The population is items completed in a declared window, and the unit is one item plus its recorded path. The question is descriptive: how often did the written procedure fail to describe what actually happened? It is not a claim that exceptions are wrong.

Exceptions are not automatically defects. A good exception may be the fastest correct response to a source that arrived in the wrong format. The problem is the missing explanation, because the next reviewer cannot tell a thoughtful deviation from a shortcut. The cost is paid in rework, delay, owner interrupts, and weaker evidence when a customer or auditor asks.

Make the exception visible without asking people to confess. Add a small field to the queue that says the written procedure did not fit, with a short reason and the name of the person who decided. Keep the field separate from the quality score so a coordinator is not punished for reporting reality. A hidden exception is more expensive than a recorded one.

Estimate cost in components rather than a single number. The first component is the time spent re-explaining the case to the next reviewer. The second is the delay while the owner is located. The third is the correction if the shortcut was wrong. The fourth is the review effort required to prove the item was handled properly. Each component can be observed, even if roughly.

Discussion

A defensible estimate needs a denominator and a window. Count the eligible items, the items with a recorded deviation, and the items with no explanation for a deviation the reviewer observed. Report the three counts separately. A single percentage that blends recorded and unrecorded exceptions conceals the very gap the review is trying to expose.

Owner interrupts are often the largest hidden cost. Each question that reaches the owner costs a context switch, and the cost rises when questions arrive in different channels. Track how many exceptions required a decision from a named owner and how long the item waited. That wait is a concrete measure that does not require a made-up hourly rate.

Implications and limits

Do not price the finding with an invented wage. If a monetary estimate is needed, use the company's own payroll and contractor records for the roles involved and state the assumptions. Labor cost surveys describe broad averages rather than your team, and they should be labeled as context. A transparent range with stated assumptions is more honest than a precise number with no basis.

The practical output is a short exception log and one process change. The log should show the item, the written step that did not fit, the action taken, the decision owner, and the destination record. The change should remove the most common reason for deviation or make the exception path explicit. Retest after the change with the same definition.

The limits matter. A small sample cannot support a claim about an entire provider, industry, or country. The review observes one queue in one window with the records that were available. Some exceptions are appropriate and should stay. The value is in making the pattern visible so the company can decide which deviations to formalize and which to stop.

Making exceptions visible

Start with a definition the coordinator can apply during the work. A short phrase such as the written step did not match the source is easier to use than a long taxonomy. Give two examples and one non-example, then review the first week of entries with the team. Adjust the wording until people use it without asking.

Keep the log close to the queue and far from the performance review. The purpose is process learning, not individual scoring. If the log feeds a penalty, entries will disappear and the cost will return to the owner's calendar. Publish the aggregate pattern and keep individual entries restricted.

What the cost estimate can support

A cost estimate built from observed time, wait, and correction counts can support a decision about where to invest in process documentation. It cannot prove that outsourcing is cheaper or more expensive in general. State the boundary in the memo so the number is not reused for a purpose it cannot support. A reader should be able to trace each input back to a recorded event.

Use ranges rather than point estimates when the underlying time data is rough. Show the assumptions beside the range and identify the one assumption that would change the conclusion. A decision that flips under a small change in assumptions needs a smaller pilot before a larger commitment. State that limit in the memo so the number is not reused for a different decision.

Turning the log into a control

Choose the most frequent exception reason and rewrite that part of the procedure. Add the source example, the decision rule, and the destination check. If the exception needs an owner, name the owner and the response time. Then remove the reason from the log only when the new procedure actually covers it.

Review the log monthly for new patterns and for entries that have stopped. A log with no entries after a process change may mean the change worked, or it may mean the team stopped recording. Check a small sample of completed items to tell the difference. Keep the revised procedure short enough that the coordinator can apply it during the work.

Reviewing the log with the provider

Review the log with the provider on a fixed cadence rather than only after a failure. Share the aggregate reasons and ask which ones the written procedure can absorb. The provider may see a pattern the client misses, especially when the exception involves tooling. Record the proposed change and the owner.

Do not use the meeting to review individual performance. The log is a process record, and turning it into a scorecard will dry up the entries. If a specific case needs review, use the normal quality channel and keep the person informed. Separate the two conversations in the calendar as well as in the notes.

Agree on one change per cycle and write it into the procedure. Give the change a review date and a way to tell whether the exception rate moved. If the rate does not move, check whether the team is still recording before concluding the change failed. A silent log is ambiguous evidence.

Questions