Philippines staffing guide · 9 min read · Published November 9, 2026

Set an Error Severity Scale for Offshore Quality Reviews

Use consequence-based severity levels so reviewers respond consistently without turning every defect into the same emergency.

On this page

The short answer

An error severity scale helps reviewers agree on how quickly a defect must be contained and who must be told. Define three or four levels by consequence, give each level a response window and an owner, and calibrate the levels with real examples. Keep the scale separate from the correction note, which explains the specific fix. Use offshore staffing services and the research library to align severity with the risk of the role. Review disputed calls monthly so the scale stays useful rather than ceremonial.

Define severity by the consequence of the error, not by the effort to fix it.
Give every level a response window, a containment step, and a named owner.
Calibrate with the same examples across reviewers before go-live.
Review disputed calls and revise the examples, not the whole scale.

Start with consequence, not effort

Teams often label an error severe because it is difficult to fix or because a senior person noticed it. That makes the scale a measure of internal discomfort rather than customer or business impact. Start instead with what the error could cause: a wrong payment, a missed legal deadline, a privacy exposure, a delayed customer reply, or a cosmetic inconsistency. Two errors can take the same repair time yet deserve different severity because one reaches a customer and the other does not. Write the consequence first, then decide how fast the response must be.

Define four severity levels

Four levels are usually enough: critical, high, moderate, and low. Critical means an immediate stop, containment, and manager notification because money, safety, legal exposure, or a customer relationship is at risk. High means the item must be corrected before it leaves the team and the client may need a notice. Moderate means a normal correction within the review cycle. Low means a quality note that does not block delivery. Define each level with one sentence and one concrete example so a reviewer can choose without a long debate.

Tie response time to severity

A severity label without a response time is only a mood. Give each level a containment window, a correction window, and a person who must be informed. Critical might require a stop within fifteen minutes and a manager notice in the same hour. High might require containment before the end of the shift. Moderate and low can follow the normal review cadence. State the business hours and the holiday rule, and name a backup for the notification path. If the team cannot meet the window, the scale should say what to do instead of quietly missing it.

Calibrate examples across reviewers

Before using the scale, give every reviewer the same ten fictional items and ask them to assign a level. Compare the answers and discuss the disagreements. A split on a payment error usually means the consequence statement is vague, while a split on a formatting issue may mean the low level is doing too much work. Revise the examples until most reviewers agree on the critical and high cases. Perfect agreement is not the goal, because judgment matters. The goal is a shared boundary for the decisions that trigger urgency and notification.

Record severity at the point of review

Ask the reviewer to select a severity when the defect is logged, not after the correction is finished. The first assessment reflects the information available when the problem was found, which is the right basis for a response decision. Add a field for the final severity if new facts change the picture. Record who changed it and why. This preserves the difference between a bad initial call and a reasonable call made with incomplete information. It also prevents hindsight from making every missed defect look obvious and every reviewer look careless.

Review disputed severity calls

Set a short monthly review of calls that were disputed, upgraded, or downgraded. Read the item, the original level, the reason for the change, and the response time. Look for patterns by task type, shift, or reviewer. If one task repeatedly produces disagreement, the acceptance standard may not distinguish a blocking defect from a cosmetic one. If one reviewer consistently downgrades, check whether the examples are clear or whether the reviewer is protecting the queue. Keep the conversation about the evidence and the rule, not about who was right.

Report severity mix, not only totals

A total error count hides whether the team is producing small formatting notes or missing payment checks. Report the count by severity and by task type, with the number of critical and high items stated separately. Show the time to containment for each high-consequence item. Compare the mix over several weeks rather than one. A rise in low notes may mean the review became more thorough, while a rise in high items is a different signal. Report the definitions beside the numbers so a reader outside the team can interpret them. State the reporting period, the data source, and the number of items reviewed. A reader should be able to reconcile the totals without asking for a separate extract.

Keep the scale small and current

Review the scale when a new client, system, or regulation changes what an error can cause. Add an example rather than a new level unless the response path genuinely differs. Too many levels create false precision and slow decisions. Keep the document to one page with the levels, examples, windows, and notification owners. Put a version date on it and name the person who can approve a change. Train new reviewers on the current version, and remove old copies from shared drives so nobody applies a retired threshold.

Sources and further reading

NIST Cybersecurity Framework 2.0Framework used to frame consequence-based response and escalation.
CISA Cyber EssentialsPublic guidance used for containment and notification discipline.
ISO 9001 quality management principlesQuality principles used to frame defect classification.
SBA guide to hiring and managing employeesPublic guidance on consistent performance feedback.

Questions managers ask

Can a low-severity error become high later?

Yes. Record the change when new facts show a larger consequence. Keep the original level and the reason for the change so the response can be reviewed without hindsight bias.

Should severity be set by the worker or the reviewer?

The reviewer sets the level because the reviewer is deciding the response. The worker can flag a concern, and the reviewer should record that input before finalizing the call.

A practical decision framework

Define the task, evidence, authority boundary, reviewer, and stop condition before assigning live work. Test a normal case and an exception, then expand access or volume only after the evidence is consistently reviewable.

Compare offshore staffing services, review the research library, or request a role plan grounded in your systems and approval boundaries.