Compare SLA attainment when ticket priority mix changes

The overall SLA percentage can improve because the month contains more easier tickets, even when performance within each group is unchanged or worse. Before celebrating the headline, compare the same kinds of obligations under the same rules.

The SLA glossary owns the definition. This page addresses a comparison problem: different ticket mixes across periods or teams. It does not set service-credit rules or redefine customer obligations.

Establish comparable groups

Select strata before looking for a favorable result. Priority is a useful starting point only when its meaning is stable. A P1 incident under a one-hour response target is not interchangeable with a P1 incident under a four-hour resolution target. Consider account target, obligation type and calendar as well as severity.

Use completed cycles or another explicitly agreed population consistently. The denominator guide explains why open and unknown outcomes need separate treatment. Show the count in each stratum and retain excluded cases. If the priority assignment process changed, document that before attributing a movement to service delivery.

Atlassian's documentation allows different SLA goals to apply to different groups of work. That is a reason to inspect configured targets, not evidence that all organizations use the same priority scale. SLA goals documentation.

An example in which mix cushions deterioration

These synthetic completed-cycle counts use the same targets and eligibility rules in both periods.

The raw total falls six points, although each group falls ten points. October contains far fewer urgent cases, partly cushioning the decline. Fixed September weights of 50% urgent and 50% routine give October a standardized rate of 0.5 x 70% + 0.5 x 80% = 75%. The standardized change is minus ten points.

Now consider another October mix with the same group rates but 10 urgent and 190 routine cycles. Raw attainment becomes (7 + 152)/200 = 79.5%. Performance inside each group is unchanged from the first October mix; only the composition moved. This makes the reason for retaining both rates visible.

Calculate the two views

The raw rate is total passes divided by total eligible observations under the selected population. It uses the period's actual counts. The standardized rate is the sum of each group's rate multiplied by a reference weight held fixed across all compared periods.

Choose reference weights from an agreed baseline, pooled comparison population or operational planning mix. Label that choice and freeze it for the comparison. Choosing a different reference for each period removes the purpose of standardization. Weights must sum to 100%, and every referenced group must have a usable rate.

For the first example, September's standardized rate is 85% and October's is 75%. October's actual rate is 79%. The four-point difference between its actual and standardized results reflects the weighting scheme; it is not an additional four points of customer breaches.

Handle empty and thin groups honestly

A group with no eligible observations has no observed rate. Do not enter 100% or carry forward last month's result without labeling an imputation. Either restrict the comparison to a disclosed common set, extend the observation window or report that the standardized result cannot be calculated as specified.

Small groups are volatile. One additional pass in a ten-cycle group moves the rate ten points. Display counts and avoid ranking teams by tiny differences. Repeated cycles from one account may also be dependent; the counts alone do not establish statistical certainty.

Use the timestamp audit to check whether apparent group differences arise from different logging quality. A team missing first-response events should not be declared slower solely because its evidence is weaker.

Keep contractual and management conclusions separate

An account's contractual result is evaluated against its own applicable terms. A standardized management comparison answers what performance would look like under a shared reference mix. It does not change the account's achieved result, create a credit or remove a breach.

A useful report therefore shows raw totals, within-group rates, fixed weights, standardized results and unknown coverage together. The per-account guide remains the starting point for customer reporting, while contract-version reporting addresses changed terms.

For a scoped review, supply two comparable authorized extracts and the priority/target definitions. Discuss the comparison before assuming that a particular adjustment or recurring report is available automatically.