Features
Forensic-quality analysis for construction and infrastructure programmes — from a single health check to tracking one project across every monthly update. Primavera P6, MS Project, and Asta Powerproject, in minutes.
Schedule Health Analysis
Comprehensive quality assessment of any construction or infrastructure schedule, delivered in minutes. Built on the DCMA 14-Point assessment — the defence-grade standard now used across capital projects, energy, and rail.
- Schedule Quality Grade — a single A–F letter rolled up from the DCMA-14 result
- DCMA 14-Point compliance score with pass/fail per check
- Per-check offender detail — every failing item lists the activities behind it (tunable thresholds on a tracked project)
- Logic density and open-end detection
- Total float distribution and critical-path integrity checks
- Constraint usage audit (hard, soft, mandatory) and calendar validation
- Calendar reconciliation — restates float across mixed 5-day / 6-7-day work patterns so you chase the right path first
- Progress forecast with an earned-value completion projection
- Remaining-plan feasibility signal — the production rate the plan requires vs the rate demonstrated so far
- Excel appendix with full activity detail
Schedule Comparison
Upload two versions of the same schedule — a baseline and an update — and get a forensic report identifying exactly what changed and why the completion date moved.
- Three-number delay decomposition — activity-level vs scope-impact vs topology, not one smeared 'net delay' figure
- Delay-attribution reconciliation — separates genuine acceleration from duration booked down on paper
- Critical delay path tracing and activity-level categorisation
- Concurrent delay-period detection and absorption analysis
- Net delay assessment after accounting for concurrency
- Responsibility tagging → supported-EOT range (tracked projects): you classify events Employer / Contractor / Neutral; the engine aggregates the position
- Activity matching that survives renumbering (ID, then name + WBS, with confidence flags)
- SCL Protocol Appendix B records-adequacy assessment
- Delay waterfall and timeline visualisations, plus a plain-language narrative
- Excel appendix with full activity detail
Series Analysis
Upload three or more chronological updates and see how delay has developed over the life of a project. Built for monthly programme submissions on construction and infrastructure jobs.
- 3 or more schedule files in chronological order
- Period-by-period delay tracking and cumulative waterfall
- Schedule-delay-over-time chart — observed end-date movement against engine-attributed critical delay on one axis
- Completion-date trend, and where the delay first appeared
- Forecast track record — backtests the project's own past completion forecasts (realized accuracy once complete; convergence while in flight — never conflated)
- Remaining-plan compression trend — a schedule holding its end date on paper while feasibility drains
- Baseline-evolution ledger — re-baselined programmes measured against the plan in force for each era, re-anchored at every re-baseline
- Cross-update activity matching that survives renumbering
- Excel appendix with cross-period filtering
Update-integrity checks
A monthly update can be made to look healthier than the works actually are — by a settings change, a shaved duration, a forced date, or a rewritten actual. This is the category no other schedule tool we've found puts front and centre. ScheduleLens runs a suite of checks aimed at exactly it — every flag states the mechanical evidence it rests on and leaves judgement to you. It never claims intent, fraud, or responsibility.
-
Scheduling-mode changes
Flags a switch between Retained Logic and Progress Override across updates — a settings change that can move the critical path and float without touching a single activity — and quantifies the divergence between the two modes on the same data.
-
Concealed slip
Quantifies slip hidden by durations shaved between submissions, by restoring each activity to the prior update's observed values. Reported as a defensible lower bound on the comparison and series reports and tracked on the dashboard.
-
Rewritten actuals
Flags recorded actual start / finish dates that changed after they were first reported — a recorded actual is a historical fact and should be frozen — plus percent-complete values that contradict the recorded remaining duration.
-
Critical-path churn
Measures how much critical-path membership turns over between updates, and flags the sharp tell: an activity that leaves the critical path in the same update its forecast finish slips later. Membership is from our own recomputed CPM, not the file's stored flags.
-
Expected Finish forcing
Detects per-activity Expected Finish dates that override logic-driven scheduling, and highlights recorded progress that contradicts the forced date.
-
Constraint and float tells
Surfaces As-Late-As-Possible constraints, hard-constraint usage on the critical path, and float-consumption patterns across updates — including periods where float churn re-times work rather than eroding the buffer.
Multi-format Support
We parse the formats your team already uses — no conversion steps needed.
.xer
Oracle P6 XER
Native Primavera P6 export format, fully parsed including calendars and relationships.
.xml
Primavera P6 XML
P6 XML exchange format with complete project data fidelity.
.xml
MS Project XML
Microsoft Project XML format for teams using MS Project Professional.
.mpp
MS Project (binary)
Native MS Project binary format — MPP14, used by Project 2010, 2013, 2016, 2019, and Microsoft 365.
.pp
Asta Powerproject
Native Asta Powerproject format common in UK construction. Some older format versions may not be supported — you'll get a clear error if so.
The engine computes. The AI writes.
The schedule analysis itself is not performed by AI. Every number in a ScheduleLens report — DCMA checks, critical-path trace, float, delay attribution, BEI, CPLI — is computed by a deterministic Python engine that produces the same result for the same file, every time.
The AI's job is the writing: it turns the engine's already-computed findings into an executive summary and finding-level narratives. It is grounded in the standards — each finding carries a citation, the model is given verbatim excerpts of the cited clauses (DCMA 14-Point, AACE RP 29R-03, the SCL Protocol), and it can quote only those excerpts. If the AI is ever unavailable, the report still ships with template prose; the numbers never change.
Your schedule data is never used for model training.
Track one programme over time
Upload each monthly update of the same project and watch schedule quality, forecast finish, and the integrity signals move between submissions. Built for planners, PMs, and consultancies running multi-month construction and infrastructure programmes — one project and its own versions over time, not a portfolio of distinct projects.
- Forecast finish date — and how it's moving update to update
- Schedule Quality Grade (A–F) and DCMA score, trended with deltas
- BEI, float consumed, and critical-path duration across snapshots
- Integrity signals trended: concealed slip, critical-path churn, rewritten-actuals count, scheduling-mode changes
- Remaining-plan compression ratio — feasibility drift over time
- Forecast track record — how past forecasts held up
- One-click open of any historical analysis, and a per-upload delta digest email
View sample Project dashboard
Four snapshots tracking a delay narrative
Re-baselining, handled honestly
Real programmes get re-baselined — an EOT award, a contract amendment, an accepted revised programme. Measure every update against one original baseline and a snapshot from after the re-plan reads as "ahead of a plan that didn't exist yet". ScheduleLens models the baseline as an ordered set of eras instead.
-
Each era measured against its own plan
Every snapshot is measured against the baseline in force for its era — not one baseline smeared across the whole history. The vs-baseline trend re-anchors at each re-baseline boundary, shown with a visible marker.
-
Baseline-evolution ledger
The series report shows slip accrued within each era and what each re-baseline reset at its boundary — 'slipped X vs original → re-plan reset Y → slipped Z vs amendment' — with a per-era headline instead of one net figure across plans that changed.
-
"Was this re-baseline agreed?"
Record each era's instruction reference and an agreed / disputed / unknown status. The baseline-integrity callout then resolves the question: an agreed re-plan reads as the plan of record; a disputed one keeps the warning. Evidence, never intent.
-
Still one programme, not a portfolio
This is multiple baselines of the same project over time — surfaced on the dashboard timeline, the series report, the Excel export, and the schedule agent. It is not portfolio analytics across distinct projects.
Ask the schedule, not a spreadsheet
Two grounded, read-only assistants answer questions over the numbers the engine has already computed — one over a single report, one across every snapshot of a tracked project. They answer only from the analysis; they cannot invent a figure. Inference runs on a zero-retention provider, and your schedule is never used to train a model.
Grounded Q&A
"Which activities drove the two-month slip?" "How did BEI trend across the last four updates?" "Was the November re-baseline agreed?" Answers cite the computed result they came from.
Bounded what-if
"If activity X takes 10 more working days…" recomputed through our own calendar-aware CPM and reported as movement relative to the same engine's base run — a bounded scenario on the network, never an absolute date or a forecast.
Privacy by design
Schedule data is sensitive. We treat it that way at every stage of the pipeline.
Deleted after processing
Uploaded schedule files are removed immediately after the analysis completes. A failsafe sweep deletes any orphans within 24 hours.
Encrypted at rest
Reports are stored with AES-256 server-side encryption, scoped strictly to your account, and deletable by you at any time.
Never used for training
Your schedule data is never used to train AI models, aggregated across users, sold, or shared with anyone other than the inference provider that generates report narrative.
Higher-isolation options — self-hosted inference and full on-premise deployment — are on the V2 roadmap for government, dispute, and defence projects.
Read the full privacy policy →