Quality sampling: how many outsourced items to check

Key takeaways

A quality sampling plan says how many finished items you check, how you choose them and what happens when errors rise. For outsourced repetitive work, start high, checking every item for the first week, then step down as accuracy holds. Tie the rate to risk: a payment needs more checks than a record update.

• Week 1: check every item.
• Weeks 2 to 4: check about 1 in 5.
• Steady state: about 1 in 20, plus all high-risk items.
• Step the rate back up if errors rise.

Why sample instead of checking everything?

Checking every item forever means paying twice for the same work. Sampling checks enough items to know the error rate with reasonable confidence, and saves full checks for high-risk outputs. The point is to catch patterns early, not to find every slip, which is the job of controls built into the method.

• How UK SMEs can use outsourcing to handle repetitive tasks
• Quality assurance for outsourced teams

What sampling rates should you use?

These are GBL working rules of thumb rather than a statistical standard. Start at 100% in the first week or the parallel run. Move to about one in five for the next three weeks, then about one in twenty once accuracy holds at your target. Always check items above a value or risk threshold in full.

• Go-live week: 100% of items.
• Weeks 2 to 4: about 20%.
• Month 2 onwards: about 5%, chosen at random.
• High-risk items, such as payments over a limit: 100%.

• Parallel running: test an outsourced task safely
• The four eyes principle for small businesses

What does a sample look like in numbers?

In an illustrative example, a team processes 800 records a month. At 5% you check 40, roughly 2 a day. If 1 of the 40 is wrong, the estimated error rate is 2.5%. If the target is 2%, step the rate back up to 20% until the cause is fixed. These are example figures.

• Right first time rate: the quality number that cuts cost
• Outsourcing efficiency KPIs: prove the work got faster

When is sampling not enough?

Sampling is not enough when one error is costly or irreversible, such as a supplier payment to the wrong account or a regulatory filing. Use a full second check or a built-in control on those items. Sampling is also weak when volume is very low, because a few items say little.

• Mistake-proofing admin processes: design errors out
• Company secretarial admin: confirmation statements on time

What does this look like in practice?

A pattern we see: sampling stays at 100% for months because nobody agreed when to step down. Writing the step-down rule into the scope from day one saves those hours.

• See client work
• BPO Solutions for UK businesses

Sampling plan checklist

Agree these points before go-live.

• Accuracy target per task.
• Sampling rate for each phase.
• Random selection method.
• Full checks for high-risk items.
• The step-up rule when errors rise.
• A weekly log of errors by type.

• Quality drift: why outsourced routine work slips over time
• Scope of work for an outsourced task: what to write down

Next step

Tell us the task, its monthly volume and what an error costs. We will suggest a sampling plan and an accuracy target to match.