The admin is a wage you're already paying.
Put your numbers in. Every constant on this page is cited, labelled, and deliberately conservative — the workings sit on the table, not behind a lead form. This is a model, not a measurement of olab: none of these constants was timed on a site running olab, because no site has run olab yet.
Put your numbers in.
What comes out is a model built from the published sources named below — not a result anyone has observed on a site using olab. Treat it as the size of the prize a pilot would have to go and measure.
Sets the denominator for the sanity check in the results — headcount never multiplies the sum.
BLS puts the median US construction manager at $55.28/hr in wages (May 2025) and benefits at ~30% of compensation (ECEC, Mar 2026); UK site engineers run ~£50k+ (Hays, 2024); a 1.25–1.4× loaded multiplier is the standard reference (Hadzima, MIT).
The $/£ switch changes the symbol only — no constant converts. Set the rate to your own currency's number.
That's 0.6 h per person per week — the industry-reported band is 2–6 h/person/week (Procore × Dodge Construction Network, 2026 Specialty Contractor ROI Report).
The model claims less than the surveys say is lost. That is deliberate.
5 reports/wk × 35 min × 52 + 40 snags/mo × 8 min × 12 + 1 update/wk × 38 min × 52 = 14,916 min ÷ 60 = 248.6 h/yr × €55 = €13,673/yr (÷12 → €1,139/mo · ÷52 → 4.8 h/wk)
Recovered capacity valued at your loaded rate — minutes come back as attention on site; only you can turn them into payroll.
Three record types, nothing else.
Each row's basis is stated beside it, at its honest confidence. Two of the three "with olab" figures are our own assumptions and are chipped as such; none of the three has been timed on a live job.
| Item | Manual | With olab | Saved | Basis |
|---|---|---|---|---|
| Daily report | 40 min | 5 min review | 35 min / report | An independent timed test completed a photo-documented daily report in 4 m 38 s vs 40+ min on paper (Contractor ToolStack, 2026). Raken's published case studies report superintendents recovering an hour or more per day (vendor-published) — corroboration only, never summed. |
| Snag round-trip admin | 12 min | 4 min | 8 min / snag | Conservative assumption, stated as ours — no independent benchmark exists; vendor-class claims run higher. The manual round trip: photograph → message → retype → chase. |
| Weekly client update | 45 min | 7 min | 38 min / week | Counted only while the toggle is on. Buildertrend reports client updates falling from 30–60 min to ~6.5 min with AI drafting (vendor claim, June 2025) — we use the midpoint conservatively. |
| Hunting for project information | 2–6 h / person / wk | — | NOT COUNTED | The industry-reported band (Procore × Dodge Construction Network, 2026 Specialty Contractor ROI Report). We use it only as the sanity check above — stacking a generic search-time row into the sum is how savings calculators inflate. Ours doesn't. |
Every constant, sourced
Ranked by how much you should trust it: INDEPENDENT TEST timed by someone with nothing to sell · SURVEYED industry survey data · VENDOR CLAIM a vendor's own published figure · OUR ASSUMPTION ours, sized low — trust it least.
| Constant | Value | Source · year | Status |
|---|---|---|---|
| Daily report, manual | 40 min | Timed paper baseline, "40+ minutes" — Contractor ToolStack, 2026 | INDEPENDENT TEST |
| Daily report, in-app | 5 min | Timed at 4 m 38 s, photo-documented, in the same independent comparison; we round up — Contractor ToolStack, 2026 | INDEPENDENT TEST |
| Superintendent time recovered | ≥ 1 h / day | Raken published case studies (vendor-published) — corroboration only, never enters the sum | VENDOR CLAIM |
| Snag round-trip, manual | 12 min | Ours: photograph → message → retype → chase. No independent benchmark exists | OUR ASSUMPTION |
| Snag round-trip, in olab | 4 min | Ours, sized low — vendor-class claims run higher | OUR ASSUMPTION |
| Client update, manual | 45 min | Midpoint of the 30–60 min Buildertrend reports — June 2025 | VENDOR CLAIM |
| Client update, AI-drafted | 7 min | Buildertrend reports ~6.5 min with AI drafting; we round up — June 2025 | VENDOR CLAIM |
| US wage baseline | $55.28 / hr | BLS, median US construction manager wage — May 2025 | SURVEYED |
| Benefits share | ~30% of compensation | BLS Employer Costs for Employee Compensation (ECEC) — Mar 2026 | SURVEYED |
| UK cross-check | ~£50k+ / yr | Site engineers — Hays salary guide, 2024 | SURVEYED |
| Loaded-cost multiplier | 1.25–1.4× | The standard reference — Hadzima, MIT | SURVEYED |
| Default rate on this page | €55 / hr | Our default, drawn from the four rows above — the slider is yours | OUR ASSUMPTION |
| Sanity band | 2–6 h / person / wk | Procore × Dodge Construction Network — 2026 Specialty Contractor ROI Report | SURVEYED |
There's time to recover because the old way leaks it.
Each figure carries its survey, its year and its sample size — and we never stack figures that share a source family and call it corroboration. One representative number per source, nothing compounded.
Where the model says the time goes.
No stopwatch in this section — it's qualitative on purpose. The model's minutes come from four places where retyping simply stops existing:
Capture is beside the camera
The photo is taken in the app, against the task it belongs to — task, checks, location and person already known. No filing, no renaming, no evening upload session.
The field loop →The log assembles from records
The daily log drafts itself from what was verified today; the person who owns it reviews and signs, instead of typing the day from memory at nine o'clock.
The field loop →Registers print themselves
Permits, ITRs, forms and submittals are filed once, on site — and leave as the register you'd hand an auditor, PDF and CSV, without a retype in between.
The paper layer →The client update is a film
The Friday brief renders from verified records only — the PM's share of it shrinks to review and approve, and the client stops phoning for progress.
The client lens →Judge it on a pilot,
not on this page.
This model is deliberately conservative: it counts only three record types, and ignores rework avoided, dispute defence, and month-end reconciliation entirely — the places a verified record earns its keep. Judge it on a pilot instead.