Lesson 7 of 7
Estimates, cycles and suspension rules
9 min
Listen to this lesson
A strategy that names the levels but not their size is half a plan. In Explore, program leadership asks two questions: how many people, and for how long? ISO/IEC/IEEE 29119-3 gives the test plan sections for testing activities and estimates, staffing and schedule, and the ISTQB Advanced Test Management syllabus treats estimation as a core test management task. This lesson covers the estimate, the cycle calendar of a system conversion, and the rules for stopping and restarting a cycle.
Estimating test effort
Estimation techniques fall into two families. Metric-based techniques use data: ratios (for example, test effort as a share of build effort on a comparable program) and extrapolation (projecting from the first days of a cycle). Expert-based techniques use judgment: Wideband Delphi, where experts estimate separately and converge over rounds, and planning poker in agile teams. Three-point estimation combines an optimistic (a), most likely (m) and pessimistic (b) figure: the expected effort is (a + 4m + b) / 6 and the standard deviation is (b − a) / 6. Whatever the technique, write the assumptions into the estimate. When they change, so does the number.
| Input | Value | Where it comes from |
|---|---|---|
| Test cases in scope | 420 integration and end-to-end test cases | The SIT 2 test plans in Cloud ALM |
| Optimistic (a) | 900 hours | Few defects; data and roles ready on day one |
| Most likely (m) | 1,200 hours | SIT 1 actuals: about 2.9 hours per test case, retests included |
| Pessimistic (b) | 1,800 hours | SIT 1's worst week repeated: environment down, heavy rework |
| Expected, (a + 4m + b) / 6 | 1,250 hours | (900 + 4,800 + 1,800) / 6 |
| Standard deviation, (b − a) / 6 | 150 hours | Reported as 1,250 ± 150 hours |
Priya's three-point estimate for SIT cycle 2 execution, with its assumptions. Numbers are illustrative.
Effort is not duration. 1,250 hours across eight testers giving 25 productive hours a week is a little over six weeks of work, not a four-week cycle. In SAP programs the factors that move estimates most are the quality of the test basis (are process designs and requirements signed off?), the defect and retest rate of the previous cycle, key users who keep their day jobs, finance testers who disappear during month-end close, environment downtime, and how much test data has to be built by hand.
Cycles around the conversion runs
In a system conversion the test calendar is pinned to the conversion runs. Contoso converts a fresh copy of EP1 in CMQ for Mock 1, Mock 2 and the dress rehearsal, about three weeks before go-live. Each run removes the test documents in CMQ client 100, so each cycle must end before the next run, and each run is followed by its validation tests (15 hours after Mock 1, 13 after Mock 2), a smoke test and the reload of test-specific records. Realize runs in two-week sprints, so Priya plans SIT cycles in whole sprints.
Suspension and resumption criteria
Exit criteria say when a cycle is finished. Suspension criteria say when to stop it early, and resumption criteria say what must be true to restart. ISO/IEC/IEEE 29119-3 lists both in the test strategy part of the test plan. Without them, testers keep executing against a broken baseline, and the defect log fills with dozens of failures that share one root cause.
| Suspend when | Resume when |
|---|---|
| CMQ client 100 is unavailable for more than half a working day | Sam confirms the system is back and the smoke test passes |
| One S1 defect blocks more than 20% of the planned test cases | The fix is imported in an announced window and the blocked cases' first steps pass |
| An unplanned import breaks the transport baseline | The baseline is restored or re-confirmed against the entry criteria |
| Converted data fails reconciliation after a conversion run | Dan signs the corrected reconciliation |
Example suspension and resumption criteria for Contoso's SIT. Thresholds are program decisions.
What would you do?
Grace asks Priya to cut SIT 2 from four weeks to two so UAT can start earlier. Priya's estimate is 1,250 ± 150 hours with eight testers. What should she answer?
- Estimate with a named technique: ratios or extrapolation from data, Wideband Delphi or planning poker from experts, three-point for a range.
- Three-point: expected (a + 4m + b) / 6, standard deviation (b − a) / 6; write the assumptions down.
- In a conversion, plan test cycles around the mocks and the dress rehearsal; each run wipes the test client.
- Set suspension and resumption criteria so a broken baseline stops testing instead of flooding the defect log.
Check your understanding · not graded
Lesson check
1. Optimistic 900, most likely 1,200 and pessimistic 1,800 hours. What is the three-point expected effort?
2. Which pair are expert-based estimation techniques?
3. What do resumption criteria define?
4. The cutover team moves Mock 2 forward by a week. How should the strategy treat this?