2026 T2 Construction Manager Analysis
Jobs, inspection timeliness, QC, return fixes and misc work — per CM, for every job whose major tasks finished in the period.
Built September 9, 2026 from exports dated September 9, 2026 · Owner: Ed Marshall <ed@cascade.build>
2026 T2 · May 1 – August 31, 2026 · 944 jobs
How these numbers are built
- The cohort. Every job whose Major Tasks Completion Date falls between May 1 and August 31, 2026 — 944 jobs. That is what the job export was filtered to, and it is the whole population here: every inspection, QC record, return fix and misc task on those jobs counts, whatever its own date.
- Job credit. A job counts once for its Construction Manager of Record. Where the CM changed after the 4-Way the job is split 0.5 to each, revenue too.
- Business days everywhere. Every gap is counted Mon–Fri, so a Saturday finish never counts against anyone. Public holidays are not modelled.
- Inspection on time = the first entry in the CM Inspection Log was recorded within 2 business days of the task’s completion. Only the earliest log entry counts.
- Office-entered inspections are excluded. 143 inspections on cohort jobs were keyed in by the Cascade IT account rather than a CM; they are dropped from every inspection figure. Return fixes and misc tasks are unaffected.
- QC comes from two systems. Cascade moved to the new QC Checkoff system in late July 2026. A job counts as QC’d on the earliest of: its first completed old-system QC Walk task (taken from the complete export of those tasks, not from the job record), or its first QC Checkoff completion. Both checkoff statuses count as walked — Punchlist Items In Progress means the walk happened and defects are being fixed. Where a job has both, the old walk is used. Blank-job-name checkoff rows (test entries) are dropped.
- QC’d % is the share of a CM’s jobs with any QC record. QC on time is measured only over jobs that were QC’d: within 1 business day of major-tasks completion. Read the first before the second.
- Return fixes are raised when a crew got something wrong during the normal task and the inspection did not catch it. They read as inspection quality. 14 are still open and are dated today.
- The task type behind a return fix is inferred, not recorded. A return-fix record names only the job and the crew; the same job plus the same crew identifies the task. 181 of 268 resolve to a single type outright; the rest use the most recent task that crew finished on the job before the fix. Treat the hover breakdown as a strong indication of where rework clusters.
- Misc tasks are “Misc. Work” tasks — extra work outside standard scope. They come from the inspections export, so they are also counted in Task inspections.
- Reorders are purchase orders raised to order material again on a cohort job — 2,740 of them, for reasons of Short (stolen, damaged or not enough ordered) or Plan error (not ordered correctly on the original order). Counted per job and normalised per credited job; $ is the unvoided PO total with tax. Every PO on a cohort job counts, whatever its date.
- Cycle time is business days from Major Tasks Start Date to Major Tasks Completion Date, shown as a median per CM. 26 jobs are left out because both dates are stamped the same day — a rollup artifact in how the start date is derived, not a one-day build. A shared job counts toward both CMs’ medians.
- Jobs with no QC record. If the job is archived, the missing walk counts neither for nor against the CM — it leaves the QC denominator. If the job is still open, the walk is owed and the job is counted as waiting. Archived means the job’s Status field reading Archive or Warranty.
- Profitability comes from TrackVia’s own Job Revenue & Profitability report (view 1942, last night’s sync), one row per job: Labor is fully burdened crew labor (task pay × 1.2), Material is every un-voided purchase order with tax, and Gross profit is revenue minus both — it reconciles on every one of the 932 cohort jobs that have a row. 12 jobs have blank revenue and therefore no P&L; they still count everywhere else. 89 jobs closed at a loss. All three are split 50/50 on shared jobs like revenue. Material here is all POs; the non-standard slice (reorders, will calls, consumables) is a separate breakdown.
- Volume floor. A CM is scored with at least 10 credited jobs and 100 inspections; everyone below is listed under the table with raw counts.
Where the QC figures come from.
717 of the 944 cohort jobs have a completed old-system QC walk, and a further 100 have a QC Checkoff (the new system, live from late July) with no old walk. 127 jobs (13%) have no QC record in either system — 51 of them are archived and count neither for nor against anyone, 76 are still open and are counted as waiting. Of the 503 May/June jobs, 30 have none. “Archived” here means the job’s Status field reading Archive or Warranty. The old-system date is read from the complete export of QC Walk tasks — 2,482 tasks, 2,415 of them completed — rather than from the job record. 19 cohort jobs have an old walk task that was created but never completed; 208 have no old walk task at all. An earlier version of this report (9 Sep, morning) was built from a job export that unintentionally excluded every job with a completed old-system walk, and is superseded by this one.
Two more things to know.
1. The cohort is heavily back-loaded: 229 of 944 jobs (24%) finished major tasks in August, ramping week by week through the month. 2026 T2 is effectively August plus a tail.
2. Inspections on cohort jobs run from December 2025 to September 2026, because work on a job starts well before its major tasks finish. All of them count — the cohort is defined by the job, not by the inspection date.