How UK Contractors Standardise Estimates with NRM Codes

Most margin loss comes from scope and risk gaps, not wrong rates. A 2023 RICS cost prediction study put pricing errors at about 30% of final account variance. The other 70% came from unallowed prelims, logistics, temporary works, compliance, and late design change.

Early estimates only work if you treat them like what they are, a range. RICS guidance puts Stage 1 feasibility at about plus or minus 20%. Stage 3 is about plus or minus 10%. Stage 5 tightens to about plus or minus 2% if scope is stable and the supply chain is locked.

Prelims are where “right rates” still turn into a bad job. Mid-market UK work often lands in the 8% to 15% band for prelims. The spread is driven by time and access, not contract value, so a simple percentage uplift is a gamble.

Labour burden is the other silent killer on specialist packages. Travel, overtime, supervision ratios, and non-productive time can shift the real rate by 10% to 25% if you only price hours at base rate. That gap usually shows up late, once the programme and resourcing get messy.

Run one discipline this month: compare estimate to final account by NRM work section, not just the total. Then lock a single structure from estimate into budget and cost-to-complete, with agreed prelim rules and labour burdens. Archdesk helps by keeping that structure consistent, so your next tender gets the benefit of your last final account.

NRM only pays you back when it becomes your cost code system, not your BoQ format. The quickest way to lose margin is breaking the link between estimate, buyout, and cost report. Keep one coded structure that follows the job from first pass to final account.

NRM 1 and NRM 2 do different jobs, so treat them as a bridge, not a choice. NRM 1 is for early-stage elemental cost planning. NRM 2 is for measured work sections that you buy, value, and change on site. Tie them together once, then stop rebuilding your estimate structure every time design moves.

Inconsistent coding quietly kills your benchmarks. Free text like “conc footings” versus “C35 founds” creates false comparisons, so your team argues about buckets instead of fixing the number. Enforced codes turn personal memory into corporate memory, even when your best estimator leaves.

The win is speed and control. If “RC frame” is always coded the same way, you can spot whether the gap is access, sequence, spec, or productivity. You can also align buyout packages and valuations to the estimate without manual rework.

Run one rule for eight weeks: no estimate line, PO, or valuation goes in without an NRM code. Archdesk enforces that structure, so “11.1 In-situ concrete” means the same thing in the tender, the order, and the final account. That is how your cost history stops being anecdotes and starts becoming a price book you can trust.

Most tender waste comes from using the wrong method at the wrong time. A 2024 CIOB survey found 43% of mid-market contractors used the same approach regardless of project stage, and those firms ran three times more re-estimate cycles. Match the method to what is actually defined, or you measure too early and guess, or stay high-level too long and get hit at buyout.

Elemental estimating (NRM1) works at RIBA Stage 2 to 3, when drawings are outline only and you need to stress-test the big cost drivers before committing resource to measurement. Take a £4m school extension: you break the budget across NRM1 elements, substructure, superstructure, finishes, services, and externals, using cost-per-m² benchmarks from comparable completed schemes. The process takes 18 to 22 hours on a £5m job. The failure point isn't the method itself. It's that abnormals, abnormal foundation depth, restricted access, or phasing around a live school, sit outside the standard element allowances and get missed unless someone has explicitly listed what's different about this job versus the benchmark.

Comparative estimating is the fastest tool available, at 6 to 10 hours for repeat scope. Price a standard 8m-eaves portal-frame warehouse shell by pulling the cost-per-m² from two or three completed warehouse schemes, adjusting for location factor, current steel price, and any site-specific constraints. The process is: identify the closest comparator, strip out any distortions from the original job (unrecorded disruption, abnormal prelims, a one-off negotiated subcontract rate), then apply the adjusted rate to the new scope. The failure point is stale or dirty data. If the "similar job" ran 14 weeks late and the delay costs were absorbed into the preliminaries line rather than called out separately, the benchmark is wrong and you won't know it until buyout.

Unit-rate estimating (NRM2) earns its place when scope is measurable, procurement will follow the same structure, and valuations will be argued line by line. A 20-unit housing development is the right context. The process runs through take-off, producing measured quantities for every trade package against NRM2 work sections, then rate build-up, combining materials, labour, plant, and waste factors for each item, then preliminary pricing, risk, and QA. That takes 65 to 80 hours on a £5m tender, with measurement and rate build-up accounting for over two-thirds of that time. The failure point is time pressure. Teams run out of measurement time and start filling gaps with guessed rates in the final 48 hours. Those gaps are exactly where margin leaks at buyout.

Assembly and package estimating suits MEP, roofing, façade, and fit-out packages, where labour productivity, access, testing, and commissioning belong together and can't be sensibly separated into individual unit rates. Price an HVAC package by building it as a system: equipment supply, pipework or ductwork runs, insulation, controls, commissioning, and builder's work in connection, all priced as a single assembly with defined scope boundaries. The process takes 12 to 18 hours if the library is current. The failure point is library drift. BSRIA tracked a 6% HVAC equipment cost shift in a single quarter in Q1 2025. If the assembly rates haven't been reviewed since last year, that margin is already gone before the tender goes out.

The decision on which method to use comes down to four things: RIBA stage, data availability, how well scope is defined, and procurement route. On a design-and-build (D&B) tender at RIBA Stage 2 with outline drawings only, elemental or comparative is the right call. On a traditional contract at RIBA Stage 4 with a full specification and construction issue drawings, unit-rate or assembly is the right call. Most tenders need a blend: elemental for the early go/no-go, then unit-rate or assembly on the packages that drive the most cost or risk. The blend only works if the rates, productivity assumptions, and scope boundaries are standardised and traceable across both methods.

Revision control decides whether your NRM2 takeoff stands up at buyout and final account. A 2023 RICS disputes analysis found 41% of quantity disagreements traced back to measurement from superseded drawings, or missing revision notes in the original takeoff. The maths was fine. The takeoff just drifted away from the design team’s latest issue.

Assumptions buried in a spreadsheet cell are where claims die. Most manual takeoffs mix the raw dimension, the waste allowance, and the rate into one formula. You can’t then show what changed when a supplier quote lands, or when the CA asks what your rate included. Split them: net quantity, explicit waste factor, and a rate build-up that shows materials, labour, and plant as separate inputs.

Run one rule on your next tender: no line goes in without a drawing reference and revision tag, plus a named rate source. Archdesk supports this by linking each quantity to the drawing issue used, and carrying the rate build-up through to package orders and valuations. That gives your QS a clean audit trail when the design moves, and when the final account turns into a line-by-line argument.

Prelims priced as a flat percentage break the moment the programme slips. A 2024 CIOB benchmarking study found firms using percentage prelims overspent their allowance by 19% on projects longer than 20 weeks. Firms that built prelims from a resourced programme overspent by 4%.

Build prelims in three buckets: fixed, time-driven, and value-driven. Hoarding install, permits, and mobilisation are fixed. Supervision, welfare, security, and temporary services are time-driven, so price them as “£ per week” against the programme weeks. Insurances tend to be value-driven, so price them separately so they don’t hide time risk.

Run one discipline on every tender: write down your weekly burn, then stress-test it against a four-week slip. Archdesk helps by linking prelim cost lines to the programme, so a change in duration shows up as cost before you submit, not at month four in the cost report.

Flat contingency fails because you can’t show what it was meant to cover. A 2023 CIOB risk management survey found firms using a single lump-sum contingency recovered only 38% of risk-related costs through the variation process. Firms that allocated contingency against named risk headings recovered 71%.

Building Safety Act work must be priced as scope, not “general risk”. Dutyholders on higher-risk buildings (HRBs) need a golden thread of building information, plus gateway submissions and evidence at handover. Those tasks have known cost headings, so treat them as a ring-fenced allowance and keep contingency for uncertainty like ground risk, design development, and interfaces.

Risk-allocated contingency only works if each line has a trigger and a release point. Tie “design development” to a design freeze, “ground risk” to the SI report sign-off, and “interface risk” to a resolved sequence between trades. That turns contingency from a sinking fund into a managed budget you can close out and give back when the risk doesn’t land.

Archdesk makes this easy to run because compliance lines, risk allowances, and work packages sit in the same coded budget from estimate to cost report. Set one tender rule and stick to it, no allowance goes in without an owner, a trigger, and a cost code the site team can book to.

Most tender “savings” come from under-priced labour burden, not smarter buying. A base rate can look competitive, then fail on site once you add employer costs, travel, supervision, and non-productive time. Treat every labour rate in your estimate as a build-up you can audit, not a single number you hope will hold.

CIS errors hit twice, once on compliance and again on cash flow. HMRC requires you to verify every subcontractor before first payment under the Construction Industry Scheme (CIS), then apply the correct deduction rate, 0%, 20%, or 30% if they are not verified. If you only model what the subcontractor invoices, your payment forecast is wrong from the first valuation.

Reverse charge VAT breaks tender comparisons and short-term working capital. Reverse charge VAT applies on most B2B construction services inside CIS, unless the client is an “end user”. Strip VAT out of every subcontract quote before you compare, then model the cash correctly. If reverse charge applies, VAT doesn’t sit in your bank as a buffer, so peak funding often lands earlier than your old forecasts suggest.

Put the rules inside your estimate, not in someone’s spreadsheet. Archdesk keeps labour burden, CIS status, and VAT treatment tied to the same cost lines you buy, value, and report later. Set one rule for the team: no labour line goes in without its burden method and tax handling, so your tender and your cash plan match the job you are about to build.

Re-keying is the hidden cost of Excel, not Excel itself. A 2022 RICS digital practice study found mid-market estimating teams spend about 35% of tender time re-entering data that already exists elsewhere. That time doesn't improve the price. It just increases the chances your buyout, budget, and cost report start from different numbers — and once those three diverge, you spend the rest of the job chasing the gap.

Start by governing your rates, not by rebuilding your templates. Pick your top cost drivers by spend and lock them into one Price Book in Archdesk, with a named owner per trade or package. Every rate needs a source, a date, and a short "included" note. That stops two common margin leaks: pricing the same item twice, and missing the small scope that always sits around the edge of a package. A concrete frame rate, for example, should state whether it includes rebar fixing labour or treats it as a separate line. If it doesn't say, every estimator will assume something different.

Keep the chain unbroken from estimate to final account. Archdesk works best when the tender structure becomes the job budget and stays intact through procurement, valuations, and cost reporting. That is where you protect margin. Invoice coding, application checks, and variation recovery all depend on the same headings the job was priced on. If your buyout uses different cost codes to your estimate, you can't tell whether a package overspent or was just coded wrong. That ambiguity costs you money at every interim valuation and again at final account.

Run the change in parallel, then switch submissions last. Price your next live tender in Excel as normal, but build the same job in Archdesk from the governed Price Book. Compare totals at work-package level and fix the library, not the tender. After two to three live tenders, convert the next win into a live budget straight from the estimate. That is the moment re-keying stops. It's also the moment your QS stops spending the first three weeks of a project reconciling the tender sum to a budget that was rebuilt from scratch after award.