Thursday night, three certificates and a bank statement
The job is a side extension with a knocked-through kitchen and a re-planned bathroom above it, taken on a fixed price over about four months. Three interim certificates have been issued. The valuation for the fourth is on Friday morning, and on Thursday evening the contractor is at his own kitchen table with the file open in front of him, adding up the three certificates and comparing them with the bank.
The certificates come to a little over two-fifths of the contract sum. The account says something close to half of it has gone out. He looks at that for a while and then does the only thing available: walks the job in his head — steels in, roof on, first fix done upstairs, plaster next week — and decides it feels about right. It might be. It might also be a job that is quietly losing money at a rate that will be undeniable in six weeks and unrecoverable in ten.
The trouble is not that he lacks discipline. It is that the two numbers on his table cannot be subtracted from one another. One is what somebody else agreed to pay for, net of retention and settled at four o'clock on a Tuesday. The other is what has cleared a bank account, which is neither all of what has been spent nor only what has been spent on this month's work. Nothing useful comes out of comparing them, and a fortnight of instinct gets built on the answer.
Spent and complete are two different measurements
The check everyone does is percent spent against a feeling of percent complete. The first half of that is arithmetic and the second half is a judgement made by the person with the most reason to be optimistic, standing in a building he has been inside every working day for eleven weeks. Familiarity is not the same as measurement, and the specific direction familiarity errs in is towards done.
What makes the comparison legal is a third quantity: the work physically complete, valued at the rates it was priced at. Not what it cost to build — that is the other number, and mixing them is how the whole exercise becomes circular. Not what the client has certified either. Value the finished work at your own budget rates and you have a figure denominated the same way the budget is, which means it can sit beside the budget and be subtracted from it. That is earned value, and everything else on this page is either how to get it honestly or what to do with it once you have.
It needs no new paperwork on a fixed price, because the document it is built on already exists. Any job with monthly valuations has a schedule of values — the contract sum broken into lines with a figure against each, whether that came out of the tender build-up, a JCT payment schedule or an AIA G703 continuation sheet, whose columns are exactly this exercise: the scheduled value of each line, what was complete at the last application, what has been added this period, what is stored on site, and the balance to finish. Earned value is that sheet filled in against the building rather than against the negotiation.
So keep two ledgers and never let them touch. The certificate ledger tells you what money is coming and when; it is a settlement, and settlements contain rounding, goodwill, materials on site and retention. The earned value ledger tells you whether the job is working. They will not agree, they are not supposed to agree, and a small contractor who tracks only the first one discovers his loss in month five, from an accountant, in a year-end meeting.
Decide how a line earns before you go out and measure it
The measurement is worth nothing if the rule for it is invented at the moment of measuring, because at that moment the answer you want is already known. So the earning rule goes against every line of the schedule of values once, at the start, and it does not change afterwards. The methods are not a matter of taste; ANSI/EIA-748 and the PMI standard on earned value both set out the same small family of them — fixed formula, weighted milestones, units complete, percent complete, apportioned effort and level of effort — and the whole skill on a small job is picking the right one per line and then not arguing with it.
Percent complete by judgement is the one to be miserly with, because it produces the ninety per cent that lasts three weeks. A line reported at ninety and then at ninety-two and then at ninety-five, while the same two men go back to the same room every Thursday afternoon, has stopped being a measurement and become a mood. Units complete cures it outright: two hundred and forty square metres boarded out of three hundred and ten is a number a stranger with a tape could confirm, and it does not care how anyone feels about the job.
Preliminaries need their own rule and it is the one most often got wrong. Scaffold on hire, welfare, skips, the van, supervision and the working foreman's time are time-related; they earn at level of effort, which means they earn with elapsed weeks and never with the progress of the building. Let prelims earn alongside the blockwork and you have built an instrument that hides exactly what it was installed to detect, because falling behind will quietly reduce the cost of falling behind. Prelims are the alarm. Do not wire them into the thing they are supposed to be watching.
The last rule is about scale. Nobody running a four-month job is going to maintain sixty control accounts, and the ones who try stop in month two. Five to eight lines carry most of the money on a domestic extension; measure those properly and put everything else on nought-or-a-hundred, which earns nothing until the line is finished. NEC4's Option A takes that to its logical end for a whole contract — under a priced contract with an activity schedule, an activity that is not complete is worth nothing at all, so there is no percentage to argue about — and while that is harsher than most small works need, it is the right instinct for the tail of the schedule.
| Line | How it earns | The measurement that settles it |
|---|---|---|
| Groundworks, drainage, muck away | Units complete against the take-off | Linear metres of trench dug and laid, loads removed against the ticket file |
| Substructure and superstructure masonry | Units complete against the take-off | Square metres built to a stated course, measured, not estimated from the scaffold |
| Steel, lintels, padstones | Weighted milestones | Delivered, set, bearings loaded, props struck — four dated events, no percentages |
| Roof covering and flashings | Units complete for the field, milestones for the details | Area covered; then abutments, verges and the chimney tray as separate events |
| First fix mechanical and electrical | Nought or a hundred, per room or per circuit | The line earns on inspection, not on completion — the two are weeks apart in consequence |
| Plaster, second fix, decoration | Nought or a hundred, per room | A room is finished or it is not; there is no partial plastering worth tracking |
| Preliminaries and time-related plant | Level of effort | Elapsed weeks against programmed weeks, whatever the building has done |
| Making good, snagging, handover | Held at nil until the list exists | Nothing earns here before the snagging list is issued and dated |
The cost side breaks first
Every warning written about earned value is about the earned value input, and on a large programme that is where the corruption is. On a four-month job for a domestic client the cost side goes wrong first, and it goes wrong for a dull reason: a bank statement is a cash measure and cost is an accrual measure, and on a small job the gap between them is enormous relative to the numbers involved.
Count what is missing on a normal Thursday. Last month's blocks and lintels are on a thirty-day account and the invoice arrives at the end of this one. The groundworker has not applied for month two because he is behind on his own paperwork. The scaffolder invoices when he strikes, which will be in August. There is retention held on two subcontract accounts, which is money incurred and not paid. Each is cost belonging to work already earned and each is invisible in the account. Leave them out and this month's cost performance is beautiful; next month three land together and the job appears to collapse overnight, which it has not. The bookkeeping has simply caught up.
The rule that fixes it is a sentence long and it is the only rule here that is not negotiable. Actual cost must cover the same scope and the same cut-off date as earned value. Not the same approximate period. The same date, applied to both sides, chosen before the exercise starts rather than after somebody has seen how the first answer looks.
Materials delivered and not installed are the special case, and on a small job they are large enough to swing a month on their own. A pallet of windows stacked in the garage is cost incurred and nothing earned, because nothing about the building has changed. A month containing a glazing delivery will show a cost performance that looks like a disaster and is not; the month after will show one that looks like a recovery and is equally unreal. The accounting standards land in the same place for the same reason — both IFRS 15 and Topic 606 of the FASB codification require an input method that measures progress by cost incurred to be adjusted where materials are delivered but not yet installed, on the ground that the delivery does not depict performance. Worth knowing, because your accountant is already making that adjustment somewhere and you can borrow the figure.
Then there is your own gang, which is the line most consistently under-costed and the line you have the most power over. Costing your own men at what you charge for them turns a labour overrun into an invisible margin transfer; costing them at the bare hourly wage does the same thing more slowly. The figure that belongs in actual cost is the hourly cost of employment — wage, employer's national insurance, holiday accrual, pension, sick pay, tools, the van, and the non-productive hours between jobs — applied to hours taken off a timesheet rather than remembered on a Thursday. Under-costing your own labour conceals the one problem you could still do something about on Friday.
- Fix the cut-off date first, in writing, and apply it to both the measurement and the cost. A valuation that measures to Friday and costs to last Tuesday's bank balance has measured nothing.
- Accrue delivered but uninvoiced materials from the delivery notes, not from the invoices you happen to have.
- Accrue subcontract work done but not applied for or not yet certified, and add back retention held on those accounts — cost is incurred at the work, not at the payment.
- Book scaffold, plant and welfare to the end of the hire period that crosses the cut-off, whether or not it has been invoiced.
- Cost your own labour at the hourly cost of employment, from timesheets, including the hours nobody was on the tools.
- Take back out anything delivered and not yet installed, and record it on its own line so that next month is not flattered by a delivery it did not earn.
Four numbers, two indices, and two honest warnings
With the budget at completion, the planned value to the cut-off, the earned value and the actual cost all measured on the same basis, the arithmetic is four subtractions and divisions and takes about a minute. Cost variance is earned less actual. Schedule variance is earned less planned. The cost performance index is earned over actual, and the schedule performance index is earned over planned. Say the index out loud in trade terms and it stops being jargon: a cost performance index of 0.91 means you are getting ninety-one pence of budgeted work for every pound that leaves the account.
Take the working set for this extension — planned 44 per cent of the contract sum, earned 41, spent 45, which is arithmetic for illustration and not a benchmark for anything. That gives 0.91 on cost and 0.93 on schedule, which sounds mild. Project it and it stops sounding mild: dividing the whole budget by 0.91 forecasts an outturn about a tenth above the contract sum, and on a fixed price that tenth comes out of one place only. Whether that is survivable depends entirely on the margin in the price, which is a number only the contractor knows and no site on the internet can tell him.
The number that ends the argument is the to-complete performance index: the efficiency the remaining work would need to still land on the original budget. Here it is about 1.07 against a demonstrated 0.91, which is asking for an eighteen per cent step change in productivity, on the same site, with the same gang. That gap is the most useful thing this arithmetic produces, because it converts an optimistic intention into a quantity somebody has to justify. If nobody can name what changes on Monday to deliver it, it is not going to be delivered.
Two warnings belong with the numbers rather than after them. The schedule index is measured in money and not in time, and it converges on 1.00 as a job finishes regardless of how late it is — a job handed over three months late still ends at exactly 1.00, because by then everything planned has been earned. It is a rough mid-job indicator and worthless at the end; time performance comes from the programme, which is the argument the CIOB's guidance on the management of time in major projects makes at length. The second warning is about the rule of thumb that a cost index settles by twenty per cent complete and never recovers. That came out of studies of very large, multi-year defence acquisition programmes and it is quoted far outside them. A four-month job has four data points. Four points are not a trend, and treating the third one as destiny is as wrong as ignoring it.
Put the four figures in as at your cut-off date: the contract sum plus approved variations as the budget, the baseline value of what should have been done by now, the measured value of what has been, and the accrued cost of doing it. The indices and the forecast fall out together, and the to-complete index in the breakdown beneath them is the one to take to Friday.
Total authorised budget for the work.
Budgeted cost of work scheduled to date.
Budgeted cost of work actually completed.
What has actually been spent on that work.
Estimate at completion
$1,166,667
Over budget and behind schedule on the figures given. TCPI shows the efficiency the remaining work would need to still land on budget.
- Cost performance index (CPI)
- 0.86
- Schedule performance index (SPI)
- 0.9
- Cost variance (CV)
- $-60,000
- Schedule variance (SV)
- $-40,000
- Variance at completion (VAC)
- $-166,667
- Estimate to complete (ETC)
- $746,667
- To-complete performance index (TCPI)
- 1.1
- Percent complete
- 36 %
- Percent spent
- 42 %
They open the calculator with your figures already in it
Earned Value Management (EVM) Calculator: 1,166,667 currency — shown in imperial, US market. The link sets both, so the result they see is the one on your screen.
What this calculation does not cover
- Every figure is only as good as the earned value input. Optimistic percent-complete reporting produces a healthy CPI on a failing project, and it is the standard failure mode of EVM in practice.
- SPI is measured in currency, not time, and it converges to 1.0 as a project finishes regardless of how late it is — a project delivered a year late still ends with SPI = 1. Use the schedule network for time performance, not this index.
- EAC here assumes current cost performance continues. Other sanctioned formulas assume the remainder runs to plan, or weight cost and schedule together; they give materially different answers on a troubled project.
Where the variance lives decides which forecast to believe
There is more than one sanctioned way to forecast the outturn and they disagree, sometimes violently. Dividing the budget by the cost index assumes what has happened keeps happening. Adding the remaining budget to the cost already incurred assumes the rest of the job runs exactly to plan. Weighting the remainder by cost and schedule performance together assumes lateness costs money, which on a job carrying time-related prelims it certainly does. All three are legitimate. On a job with three valuations behind it, choosing between them is a judgement about cause, not a choice of formula.
So open the variance by line before choosing. If the whole of it sits in groundworks — an obstruction, a soft spot, a day lost to a drain nobody had on a drawing — and the groundworks are finished and backfilled, then projecting that inefficiency across the plastering forecasts the recurrence of an event that physically cannot recur. If instead the variance is spread thinly across five lines, that is not bad luck; it is a rate problem. The labour constants used to price the job are wrong, they are wrong by roughly the same proportion everywhere, and they will do precisely the same thing to every line that has not started yet.
Which is why, on a job this size, the forecast worth trusting is usually built by hand and takes twenty minutes: finished lines at their actual outturn, lines in progress at whatever measurement you actually believe, and unstarted lines at budget adjusted for anything the first three months taught you about your own rates. The indices do not replace that. They tell you which month is the month to spend the twenty minutes, and stop you spending it in a month when nothing was wrong.
The baseline has to be the job you are actually building
The commonest way this arithmetic lies is not a bad measurement, it is a baseline that has quietly stopped describing the job. Three variations have been signed since March. If the budget at completion is still the March contract sum, then the cost of building the amended job is being compared with the budget for the original one, and the difference shows up as an overrun that nobody caused. The budget must carry every approved variation, the planned value must carry them at the point in the programme they were actually instructed for, and the date the baseline moved has to be written down — the GAO's cost estimating guidance treats the integrity of that baseline as something actively maintained rather than assumed, which on a small job means naming who owns it.
The failure in the other direction is worse because it feels like housekeeping. Moving the baseline to match what has been spent produces a cost index of exactly 1.00 and a page of numbers containing no information at all. The distinction that keeps it honest is simple to state and requires discipline to apply: a variation is a change of scope, priced and agreed, and it moves the baseline; an overspend is the same scope costing more, and it does not. The instruction and pricing of the change itself is a separate subject with its own page on this site. What belongs here is only the reconciliation — on the day of the valuation, the net of the variation register and the budget at completion are the same figure, and if they are not, every index underneath them is decoration.
The contingency question is not how much is left
Which brings the evening to its second half. The pot was sized before a spade went in the ground, against unknowns spread across the whole job. AACE International's Recommended Practice 40R-08 is unambiguous about what it is for: uncertainty within the scope already defined, priced as an amount that is expected to be spent, and not a discount anybody gets to keep for being careful. Read that way, the remaining balance answers nothing on its own: it is a number without a denominator.
The pair of numbers that does mean something is how much of the pot has gone against how much of the job is done. Forty-one per cent earned with seventy per cent of the contingency spent is a job in trouble whatever the balance looks like, because the burn rate says the remaining fifty-nine per cent of the work will need more than the pot can carry. But forty-one per cent earned with a barely touched pot is not automatically the comfortable case either. On a refurbishment where the first floor has not yet been opened up, that pot has not been saved; it has simply not been tested.
Track it in three columns, not one, because the middle column is the one that catches people. Drawn is money already spent against a named risk that materialised. Committed is a risk that has materialised, been priced and not yet paid — an order placed, a subcontractor instructed, a specialist booked. Open is what is genuinely still standing against risks that have not yet resolved. Committed contingency is entirely invisible in a bank statement and is already gone, and a contractor who reads only the balance will spend it twice.
What keeps the three columns meaningful is charging correctly in the first place, and it is worth being pedantic about because the error flatters two numbers at once. A change the client instructed is scope; it is priced, agreed and paid for separately, and it must not touch the contingency. A latent condition inside the fixed price — the drain under the foundation, the rot behind the render, the wall that turned out to be single skin — is a drawdown and that is what the pot exists for. Charge a client variation to contingency and the pot looks healthy and the margin looks intact simultaneously, which is a comfortable way to arrive at a loss. Then, and only then, re-size: list the risks still open against the work that has not been done, price them, and compare that total with the open column rather than with a percentage of a contract sum agreed in March.
| What the ledger shows | What it actually means | The check it should trigger |
|---|---|---|
| Drawn well ahead of percent complete | The early risk was under-priced, and the same estimator priced the rest | Re-price the open risks on the remaining work before assuming the rate improves |
| Drawn roughly level with percent complete | The pot is behaving as sized — which is the intended outcome, not a surplus | Confirm the risks retired so far were the expensive ones, not the cheap ones |
| Barely drawn, with the opening-up still to come | Untested rather than saved; the risk profile is still in front of you | Do not release it into the margin, and do not spend it on specification |
| A large committed column and a small drawn one | Money already gone that no bank statement is showing yet | Add committed to drawn before any conversation about what is left |
| Drawdowns with no named risk against them | Overspend being reclassified after the fact, which erases the signal | Re-code them to the line they belong on and rerun the cost index |
Set the base to the work not yet earned — budget at completion less earned value, which is the budget still standing against work that has not happened — and set the percentage from the risks on your list that are still open. The headline figure adds the two together; the buffer itself is the contingency amount on the breakdown line beneath it, and that is what the remainder of this job would need. Put that beside the open column of the drawdown; the gap between the two is the item for Friday.
Your planned budget before adding a buffer for the unexpected.
The extra buffer to add for unexpected issues.
Total budget with contingency
$23,000
Contingency is a planning buffer, not a guarantee — projects that uncover major surprises (structural damage, code-required upgrades) can still exceed even a generous contingency.
- Contingency amount
- $3,000
They open the calculator with your figures already in it
Project Contingency Calculator: 23,000 $ (total recommended budget) — shown in imperial, US market. The link sets both, so the result they see is the one on your screen.
What this calculation does not cover
- The percentage is applied to the base budget as one flat multiplier, so every dollar of the job is treated as carrying identical risk. A $20,000 kitchen made up of $14,000 of fixed-price cabinetry already on order and $6,000 of demolition into an unknown wall gets the same $3,000 buffer at 15% as one that is speculative end to end. Where the risk sits in a single part of the scope, size a buffer against that part and add it to the rest rather than smearing one rate across the total.
- Nothing in the arithmetic is a fixed amount: the buffer is purely proportional, so it shrinks with the budget while many of the surprises it is meant to absorb do not. A failed inspection, half a day of extra excavation or an emergency call-out costs roughly the same on a $3,000 job as on a $300,000 one, yet 15% sets aside $450 on the first and $45,000 on the second. Small jobs are the ones a percentage rule quietly under-buffers.
- Whatever is missing from the base figure stays missing from the answer. The base budget is read as a single opaque number, so if permits, disposal, delivery charges or temporary accommodation were never counted in it, a 15% buffer on that total does not fund them — it scales an incomplete estimate rather than completing it.
- The output is a lump sum with no timing in it. No term asks when the money is drawn or how long ago the base was priced, so a buffer taken on a year-old estimate is a percentage of a stale number. Re-running the figure part-way through a job would need the remaining scope and the buffer already consumed, and neither is tracked here.
- The percent field accepts whole numbers from 5 to 50 and the base accepts $100 to $2,000,000; those are input bounds, not guidance about where your job belongs. The commonly cited 10-20% range is a general renovation figure, and nothing in the calculation weighs building age, how much structural work is involved, or how firm your quotes are to place you within it.
What the third valuation is actually for
None of this fixes anything by itself, and it is worth saying, because a page of indices can feel like an action. What the arithmetic buys is time. At the third of five valuations there is still a buying decision to change, a sequence to re-plan, a gang size to adjust, and a claim to evidence while the cause of it is still standing on site and can be photographed. At the fifth, all of those have expired and the only remaining option is to absorb the number.
On a fixed price the levers are almost entirely internal, which is unwelcome but clarifying. The client is not going to fund your labour constants being wrong. What can still move is what you are buying at and from whom, the hours per unit your own gang is achieving, the time-related prelims that run whether anyone is productive or not, and the sequencing that has two trades waiting on each other. The one genuinely external lever is a properly evidenced change, and it needs the records made this month — dated photographs, the site diary, the delivery tickets, the instruction that was or was not given — rather than reconstructed from a phone in month five.
There is also a date in the diary worth knowing about if the client is not a homeowner. On a commercial job the payment notice and pay less notice machinery under Part II of the Housing Grants, Construction and Regeneration Act 1996 attaches hard deadlines to the valuation, and the number in an application that goes unanswered can become the sum payable by default. On a contract with a residential occupier none of that applies and the position rests entirely on what was written down, which is a different guide on this site. Either way, the measurement has to be finished before the meeting, not during it.
On the table before Friday morning
Six things that decide whether a monthly valuation tells you something or merely records a settlement, assembled on the Thursday evening while there is still a month left to act on the answer.
- One cut-off date, applied to both sides — Every measurement and every pound of cost as at the same day, chosen before the first figure is written rather than after somebody has seen how the answer looks.
- The schedule of values with an earning rule against every line — Units complete, weighted milestones, nought-or-a-hundred or level of effort, decided at the start and not revisited; preliminaries earn with elapsed weeks whatever the building has done.
- The accrual sweep: delivered and uninvoiced, applied and uncertified, retention held — Cost is incurred at the work and not at the payment, and three missing invoices buy one flattering month followed by one alarming one that describes nothing real.
- Your own labour at the hourly cost of employment, from timesheets — Wage, employer's national insurance, holiday, pension, the van and the non-productive hours — under-costing the line you control most is how a recoverable problem stays hidden.
- A budget at completion equal to the contract sum plus approved variations, dated — Reconciled to the net of the variation register on the day of the valuation; if the two figures differ, every index calculated underneath them is decoration.
- The contingency in three columns — drawn, committed and open — Read against percent complete rather than as a balance, and re-sized against the risks still open on work not yet done rather than as a percentage of a number agreed in March.
Opens the calculators above on one screen with the dimensions from this article already filled in. Quantities only — this site publishes no price list, because local prices vary too much to publish honestly.