Research framework · Results forthcoming
A research framework for construction cost control, BOQ reconciliation and AI quantity takeoff. Built around traceable project records—not unexplained headline numbers.
No project dataset or index findings have been published. Definitions below are proposed methodology, version 0.1, subject to validation before data collection and release.
The planned study examines how approved quantities, procurement decisions, material movements and project outcomes differ across eligible capital projects. It will distinguish physical material loss, pricing differences, scope changes and schedule effects instead of calling every cost variance “leakage.”
AI adoption and AI takeoff accuracy will be measured separately. Adoption does not establish causation: any comparison must account for project type, stage, size, region, procurement model and record quality.
16 fields · Proposed definitions · No project rows
| Field / unit | Measurement definition | Validation and exclusions |
|---|---|---|
| Project_IDPseudonymous identifier | Stable research identifier assigned after permission and anonymization; never a client or ERP project code. | Unique, non-identifying identifier; keep the lookup outside public data. |
| Project_TypeCategory | Approved project classification, such as residential, commercial, transport or industrial; retain a controlled vocabulary. | Record classification rationale; do not aggregate unlike project scopes. |
| RegionBroad geographical region | Region at an approved disclosure level; generalize locations when they could identify a project. | Disclosure review and small-cohort suppression before release. |
| Project_ValueINR crore | Approved total project budget at the stated baseline version. Convert foreign currency using a disclosed rate and date. | Positive value; record tax basis, scope, currency and baseline date. |
| BOQ_ValueINR crore | Approved priced BOQ for the measured scope, including documented approved changes as of the observation date. | Same scope, tax basis and observation period as matched actual cost. |
| Actual_CostINR crore | Verified incurred cost for the matched measured scope and period; distinguish incurred, paid and committed amounts. | Reconcile to ledger, receipt and billing records; avoid double counting. |
| BOQ_Variance_%Percent, signed | 100 × (matched actual cost − approved matched BOQ value) / approved matched BOQ value. Cost-weighted variance, not a physical quantity measure. | BOQ denominator must be positive; missing or unmatched scope is unavailable, never zero. |
| Procurement_Deviation_%Percent, signed | 100 × (actual procurement cost − approved procurement baseline) / approved procurement baseline for matched purchase quantities and scope. | Retain baseline, amendments and matched quantities in audit support; distinguish price from volume effects. |
| Vendor_Price_Variance_%Percent, signed | 100 × sum(quantity × (actual unit price − benchmark unit price)) / sum(quantity × benchmark unit price) for comparable approved specifications. | Document benchmark date, freight, taxes and specification; positive weighted baseline required. |
| Material_Leakage_%Percent, nonnegative | 100 × verified unexplained material loss valued at the disclosed basis / material value issued for the measured scope and period. | Reconcile opening stock, receipts, issues, transfers, returns and closing stock; exclude approved waste, scope changes and timing differences. |
| Labour_Productivity_IndexRatio; baseline = 1 | (verified physical output / actual labour hours) / (baseline physical output / baseline labour hours), for a matched task and work package. | Positive hours and baseline productivity; separate unlike tasks; disclose output units. |
| DPR_Delay_DaysElapsed days | Mean elapsed time between reporting-period close and first approved daily progress report submission for eligible reports. | Keep timestamps and time zone; missing reports are reported separately, not counted as zero delay. |
| Schedule_Variance_%Percent, signed | 100 × (actual or forecast duration − approved baseline duration) / approved baseline duration for a matched milestone. Positive means delay. | State actual versus forecast; disclose baseline revision, milestone and as-of date; duration must be positive. |
| Cost_Overrun_%Percent, signed | 100 × (verified final cost or estimate at completion − approved total budget) / approved total budget. Not synonymous with material leakage. | State final versus forecast; same whole-project scope, tax basis and approved change policy. |
| AI_Takeoff_Accuracy_%Percent, 0–100 | Proposed quantity agreement: 100 × max(0, 1 − sum(weight × abs(AI quantity − independently checked quantity)) / sum(weight × checked quantity)). | Match units and measurement rules; predeclare weights, independent ground truth, holdout drawings, sample size and error distribution. |
| AI_Adoption_LevelLow / Medium / High | Proposed rubric: Low = isolated pilots; Medium = recurring production decision support; High = governed cross-workflow execution with monitoring and human approvals. | Evidence-backed rubric assessment; not proof of causation or effectiveness. |
The CSV contains these definitions only. It is not a project dataset and must not be used as evidence of measured performance.
Permissioned deployments are not a random sample of the construction industry. Selection bias, missing records, forecast uncertainty and changing project scope can affect comparisons. Existing Deepsa AI case studies are company-reported engagements, not verified observations from this forthcoming index.
Verified anonymized records, publication permission, a reviewed methodology and an approved license are required before release. Dataset metadata and a data download will accompany genuine published records—not this framework.