Construction & Trades

The job is on site. The margin is in the workbook

McKinsey found that large construction projects typically run around 20% over schedule and up to 80% over budget, and that sector productivity has barely moved in decades. The cost information that would have changed those outcomes usually exists. It just arrives after the decision. SheetAI does the reconciliation while it can still matter.

Try it on your sheet See real workflows

Audit-ready trailNo data trainingReversible per action

$177bn

US labour lost to bad project data

Searching for project data, resolving conflicts and fixing mistakes take 35% of construction professionals' hours, an estimated $177.5 billion of US labour in 2018. From FMI and PlanGrid's Construction Disconnected survey of 600 professionals.

Source: FMI and PlanGrid (Autodesk): Construction Disconnected (2018)

30%

Say most of their project data is bad

30% of construction professionals said more than half of their project data is bad, and only 12% always use project data in decisions. From an Autodesk and FMI survey of over 3,900 professionals worldwide.

Source: Construction Equipment, reporting Autodesk and FMI: Harnessing the Data Advantage in Construction (2021)

£25bn

UK yearly cost of avoidable error

Avoidable error costs the UK construction sector up to £25 billion a year, more than public spending on maintaining roads and railways. From the Get It Right Initiative's submission to the Competition and Markets Authority.

Source: Get It Right Initiative: submission to the CMA civil engineering market study

Construction & Trades: illustration for SheetAI's construction & trades use case

Large construction projects run up to 80% over budget and around 20% late.

McKinsey's construction productivity research found the sector's output per worker has grown at roughly a quarter of the rate of the total economy, and that megaprojects overwhelmingly overrun. Rework and poor data are repeatedly identified as leading causes. A US industry study put the cost of rework and miscommunication at around $177 billion a year in labour alone.

Source: McKinsey Global Institute: Reinventing Construction

This page is about the commercial and cost work that determines whether a job makes money: estimating and take-off checking, tender comparison across subcontractors, cost-value reconciliation, variation and instruction tracking, valuation and payment applications, retention schedules, and the monthly cost report that arrives after the money has been committed. SheetAI works in the workbook the commercial team already uses, joining the site data to the ledger, reconciling committed against forecast, and flagging where a variation has been worked but never priced. It does not price risk for you, it will not warrant a quantity, and it is not a substitute for a quantity surveyor's judgement.

The state of construction cost control

Four numbers that explain why the commercial team is always reporting on decisions already made.

Large projects exceeding budget

Up to 80%

McKinsey's megaproject analysis; overruns are the normal outcome rather than the exception, and the pattern holds across geographies and project types.

McKinsey Global Institute

Annual US labour cost of rework and poor data

$177bn

FMI / Autodesk research attributing rework to miscommunication and information that arrives too late or in the wrong form to be acted on.

FMI & Autodesk, Construction Disconnected

Construction productivity growth versus the wider economy

~1% vs 2.8%

Two decades of near-flat output per worker while manufacturing and the broader economy compounded. Information handling is repeatedly identified as a leading structural cause.

McKinsey Global Institute

Commercial reporting produced after the cost is committed

Usually

The monthly CVR reports on a period that has closed. By the time the overspend appears, the labour has been worked and the materials have been ordered.

Practitioner experience

The pattern: the problem is not that the sector cannot count. It is that the counting happens a month behind the committing, in a workbook rebuilt each period from exports that do not agree. Closing that gap is a data-joining exercise, which is exactly the thing to do in the sheet rather than after it.

Anatomy of a monthly cost report

Where the commercial week goes on a mid-size contract, before any of it is automated.

Phase 1: Gather the period's data

›Labour allocations and plant returns from site
›Subcontractor applications and material invoices
›Ledger extract from accounts

⚠ Four sources, four cut-off dates, and none of them use the same cost code convention. Reconciling them is two days before any analysis starts.

Phase 2: Reconcile committed against actual

›Match invoices to orders and to the ledger
›Identify goods received not invoiced
›Chase the costs booked to the wrong code

⚠ Miscoded costs distort every package's position, and finding them means going line by line through a ledger extract.

Phase 3: Value the work done

›Measure progress against the programme
›Value variations and instructions
›Assess the work in progress not yet applied for

⚠ Valuing variations is where the money is, and it is the part done last, at speed, on Thursday evening.

Phase 4: Forecast to completion

›Project remaining cost per package
›Assess risk and contingency drawdown
›Compare against the tender allowance

⚠ The forecast is last month's forecast adjusted, which means it inherits every optimistic assumption in it and corrects slowly.

Phase 5: Report and defend

›Produce the CVR for the commercial review
›Explain the movement since last period
›Answer the question about package 14

⚠ Movement analysis is done from memory because nobody recorded why the previous figure was what it was.

Cautionary tale

Late information is the same as no information

The Scottish Parliament at Holyrood was initially estimated at between £10m and £40m and completed in 2004 at around £414m. The Fraser Inquiry that followed found the primary causes were not construction failures but compounding scope change, unclear decision-making, and cost information that consistently reached decision-makers after the decisions had been taken. The building was not badly built. The commercial information cycle simply ran slower than the commitment cycle, so every decision was made against a picture that was already out of date.

The lesson: This is the ordinary condition of construction cost control, not an exceptional failure. The commercial position is reported monthly and committed daily. Anything that shortens that gap (reconciling the ledger to the cost codes weekly rather than monthly, flagging worked-but-unpriced variations as they appear, forecasting from live committed cost rather than last period's report) changes outcomes in a way that better reporting formats never will. SheetAI performs those reconciliations on demand against whatever data you have today.

Source: The Holyrood Inquiry (Fraser Report): Scottish Parliament

Who it's for

If the commercial position of your job lives in a workbook, this is the page.

Quantity Surveyors

Rebuild the cost-value reconciliation monthly from exports that use different cost codes.

Estimators

Compare tender returns across subcontractors who each priced a different scope breakdown.

Project & Commercial Managers

Report a position that is a month behind the commitments driving it.

Small Contractors & Trades

Run job costing in a spreadsheet with no commercial team and no system.

Developers & Clients

Track drawdown, variations, and forecast final account across multiple contractors.

Real construction workflows

The prompt, the reconciliation it performs, and the position it leaves you able to defend.

construction_workbook.xlsx - SheetAI

You ask

“I've attached six subcontractor quotations for the same package. Extract each into a comparison sheet against our bill of quantities in BOQ!A:E, mapping their line descriptions onto our items. Show rate and total per item per bidder, flag items any bidder has excluded or qualified, and highlight rates more than 25% from the median.”

SheetAI does

✓Extracts each quotation's priced items with their descriptions.
✓Maps differing line descriptions onto your bill items.
✓Flags exclusions and qualifications as separate columns rather than burying them.
✓Identifies rate outliers against the median across bidders.

Formula written

=IF(ABS(C2-MEDIAN($C2:$H2))/MEDIAN($C2:$H2)>0.25,"OUTLIER","")

Result

Six tenders on one comparable grid. The apparent low bidder had excluded two items worth £41k and qualified a third, which moved them to third place once the scope was made like-for-like.

Everything the commercial team needs, in the workbook they already run

Tender comparison on a like-for-like basis, ledger reconciliation against cost codes, variation ageing that catches unpriced work, and forecasts built from live committed cost.

✓Tender normalisation and outlier detection

✓Ledger to cost-code reconciliation with exceptions

✓Variation register with stage-level ageing

✓Cost-to-completion forecasting by package

✓Retention schedules with release dating

✓Self-checking payment applications

Plays well with your stack

Accounting and ledger exports (Sage, Xero, COINS)
Estimating exports and bills of quantities
Subcontractor quotations as PDF or Excel
Programme exports (Asta, P6, MS Project)
Site labour and plant return sheets
Any CSV of costs, orders, or valuations

What a monthly CVR looks like

A representative mid-size contract with 18 packages, before and after the reconciliation moves into the sheet.

Before SheetAI

~41 hours

Mon
Gather four data sources with four different cut-offs
~7h
Tue
Reconcile ledger to cost codes; chase the miscodes
~10h
Wed
Value work done; price the variations at speed
~9h
Thu
Adjust last month's forecast; build the CVR
~9h
Fri
Explain the movement from memory at commercial review
~6h

✓

With SheetAI

~11 hours

Mon
Sources joined; cut-off differences flagged automatically
~2h
Tue
Ledger reconciled; exceptions listed and valued
~2h
Wed
Variation register aged; unpriced instructions surfaced
~2h
Thu
Forecast rebuilt from live committed cost, not last month's
~3h
Fri
Movement analysis generated with reasons attached
~2h

~73% reduction on a representative 18-package monthly cost-value reconciliation.

What SheetAI will not do

Construction commercial work carries contractual consequences, so the limits are worth being exact about.

Warrant a quantity or a take-off

It can check a take-off's arithmetic, compare quantities across revisions, and flag inconsistencies between a bill and a drawing schedule you supply. It cannot measure from drawings and it does not warrant quantities. Measurement is a quantity surveyor's professional responsibility and the liability attached to it does not transfer.

Price risk or contingency for you

It can show contingency drawdown against the programme and model scenarios you define. Deciding what risk a job carries and what allowance it needs is commercial judgement informed by things not in the spreadsheet: the client, the ground, the subcontractor, the weather, and what you know about all four.

Interpret the contract

It tracks instructions, dates, and values against the register you keep. Whether a variation is properly instructed, whether a notice was validly served within the contractual window, and what the payment provisions require are contractual questions. It flags the dates; a QS or a solicitor determines what they mean.

Replace site data

It reconciles the records you have. If labour allocations are estimated on Friday afternoon or progress percentages are optimistic, the reconciliation is precise about unreliable inputs. It will flag the patterns that suggest this: round-number allocations, progress that never falls, packages with no returns, but the fix is on site.

“

The variation register was the one that paid for itself. Nine instructions worked and never priced, sitting there for months. Everyone knew it happened and nobody had a way to see it while there was still time to do anything. Now it flags at twenty-eight days and someone deals with it.

Commercial Manager, regional contractor (£40m turnover)

Frequently asked

What commercial teams ask before they let this near a live job.

Can it do a take-off from drawings?

›

No. It does not measure from drawings and it does not warrant quantities. That is a quantity surveyor's professional responsibility with liability attached. What it does well is check the arithmetic of a take-off you have done, compare quantities between revisions to show what changed, and flag inconsistencies between a bill and a drawing schedule you provide. That is the checking layer, not the measuring layer.

Will it work with our cost code structure?

›

Yes, it reads the structure you use rather than imposing one. Every contractor's cost coding is different and most have accumulated inconsistencies over years. It will also flag costs booked to codes absent from your cost plan, which on most jobs surfaces something interesting within the first reconciliation.

How does tender comparison handle different scope breakdowns?

›

That is the main problem it solves. Six subcontractors price six different breakdowns of the same package, and comparing headline totals is meaningless. It maps their line descriptions onto your bill items, then makes exclusions and qualifications explicit columns rather than footnotes, which regularly reorders the ranking once the comparison is genuinely like-for-like.

Can it catch unpriced variations?

›

Yes, and it is usually the highest-value thing on the list. Ageing instructions by stage (issued but unpriced, submitted but unagreed, agreed but uncertified) surfaces the leakage that everybody knows happens and nobody can see in time. Flagging at 28 days from instruction turns money that historically was written off into money that gets submitted.

Does it handle retention and payment applications?

›

Yes: retention at your contractual rate with sectional release dating, the application built from your valuation data, and the arithmetic self-checked so the ties are verified before it goes out. The release calendar routinely surfaces retention that has fallen due and was never claimed.

Our site data is not great. Will this still help?

›

Directionally yes, and it will tell you where the data is weak. Round-number labour allocations, progress percentages that never go down, packages with no returns in a period. These patterns get flagged, which is useful evidence in itself. Precision requires better site records; the reconciliation exposes exactly which records are the problem.

Is this a replacement for our commercial system?

›

No. If you run a full commercial platform, keep it. This works alongside it on the analysis that ends up in Excel anyway, which for most contractors is the CVR, the tender comparison, the variation register, and the forecast. For smaller contractors running jobs entirely from workbooks, it is a considerable upgrade to what already exists.

Can it forecast the final account?

›

It projects cost to completion from committed cost, measured progress, and the remaining programme, and compares that against tender and current budget. Treat it as an early-warning indicator rather than a settlement position: the final account depends on entitlement, negotiation, and contractual interpretation that no run-rate captures.

Sources & further reading

Every benchmark and statistic on this page is drawn from publicly available research. We cite our sources because we read theirs.

Get started

Stop reporting last month.
Start seeing this week.

Drop your ledger, your orders, and your cost plan into one workbook and ask for the reconciliation. The exceptions list is usually the most interesting thing you will read this period.

All use cases

Pay as you go, no subscription required