The metre you priced and the metre you dug

You priced the trench at $18.50 a metre. The crew is digging at something else. Nobody in the business can tell you what, and the number that would settle it will arrive with the final account — which is to say, after the last decision it could have changed.

That is the whole problem, and it is smaller than it sounds. It is not a forecasting problem, or a discipline problem, or a problem with your estimator. It is one division sum that nobody performs while the work is live.

This page is about that sum: what goes into a unit rate, how to calculate the one you bid and the one you are achieving, and why the second one is the only number on a live site that predicts anything. There is a worked example throughout, from the pilot scenario in our own specification, and every figure in it is checkable on the back of an envelope. We think you should check them.

1. What a unit rate actually is

A unit rate is a price for one unit of output — one metre of trench, one chamber, one home connected. It is the atom of a construction bid, and it has a specific shape.

The components of a unit rate, for the trenching line on a fibre build:

Two things follow from that list, and they are the whole reason unit-rate analysis exists.

First: almost every component is a function of the ground. Change the surface from soil to asphalt and plant goes up, disposal goes up, reinstatement goes up, traffic management appears where it did not exist, and labour goes up because the metres come slower. One variable moves seven lines. Our own data model treats surface type as the dominant driver of a trenching unit rate, and names getting it wrong at bid time as the leading cause of variance — a design judgement in a specification, stated as one, not a published statistic.

Second: the rate you bid is an average, and the site is a specific. You priced $18.50 because that is roughly what a metre has cost you across the ground your crews have historically worked. Bahnhof Street has not been opened yet. The bid is not wrong because someone was careless; it is wrong because an average met a particular. Why that error leans the same direction for seventy years is a separate piece. This one is about the arithmetic that catches it while you can still act.

2. How to calculate a unit rate — both of them

There are two unit rates on any live site, and confusing them is where the money goes.

The bid rate is the one in your tender: line amount ÷ planned quantity.

The achieved rate is the one the ground is charging you: cost attributable to the activity ÷ quantity actually delivered.

Both are divisions. The difficulty is never the arithmetic — it is that the numerator and denominator of the second one live in different places, and on most sites neither is written down until the job is closed.

The worked example

Our pilot scenario prices SITE-03, a 92-home distribution area with 1,150 m of trench, bid as soil, at $35,000. Here is the full build-up:

Check the top line: 1,150 × $18.50 = $21,275, and $21,275 ÷ $35,000 = 60.79%. One multiplication and one division. That is the most consequential pair of sums on the site, and we have not met a PM who had done them.

Three units, one rate

Because half this industry prices in metres and half in feet or miles, and because the numbers look wildly different in each:

The all-in column is the bid divided by trench length — every line, not just the trench. Both columns are legitimate and they answer different questions, which is exactly why fibre construction cost per mile is not a number you can compare between two contractors without asking what is inside it. A $48,980-per-mile figure and a $29,773-per-mile figure describe the same site.

3. The sum that makes progress honest

Now the part that changes how you read every progress report you have ever been sent.

Trenching is 60.8% of SITE-03’s money. So the afternoon the crew passes 575 m — exactly half the trench, a genuinely good day, and the crew is right to be pleased — the value delivered is 575 × $18.50 = $10,637.50, against a $35,000 bid.

That is 30.4% of the job. Not 50%.

Nothing has gone wrong. Nobody has lied. The crew dug half the trench and reported half the trench. But if your progress number counts tasks or metres, it now says something close to 50% while the money says 30.4%, and you are about to make decisions on a figure that is nineteen points optimistic.

Run it the other way and it is worse. Suppose everything except the trench is finished: duct, chambers, blowing, splicing, reinstatement, mobilisation — six of seven lines closed out. By task count you are 85.7% done. By money you are 39.2% done. A forty-six point gap, and the number that flatters you is the easy one to produce.

This is why our specification derives progress as

progress_pct = Σ(quantity_done × line unit_price) ÷ bid_total

— value-weighted, not task-counted. Nine of ten tasks complete means nothing when the tenth is most of the money. It is the single cheapest correction available to a PM, it requires no new data collection whatsoever, and you can apply it to your current job this afternoon in a spreadsheet.

4. Unit rate analysis: the comparison, not the number

A unit rate on its own tells you very little. The ratio between the two is the instrument: achieved rate ÷ bid rate.

Above 1.0 you are losing money per metre. Below 1.0 you are making it. And unlike a variance in dollars, the ratio extrapolates — it is the only thing on a live site that lets you say something defensible about metres nobody has dug yet:

forecast = cost to date + (value of remaining work at bid × rate ratio)

That is the simplest forecast we could defend, and the simplicity is deliberate: a forecast a PM cannot explain to their own boss in one sentence gets ignored.

What that looks like on the rock

Day 3 on SITE-03, the crew hits solid rock at the 205 m mark. A breaker attachment goes on hire at $1,450 a day, four days booked — $5,800, committed the moment it is booked rather than when it is invoiced, because a hire you have committed to is already spent. The rock stretch yields 145 m in four days, about 36 m a day against roughly 105 in soil — a third of the rate.

Now do the component sum. $5,800 of breaker hire across 145 m of rock is $40.00 a metre of plant, on a line bid at $18.50 all-in. Plant hire alone, on that stretch, is 2.16× the entire bid rate — before a single hour of labour, a load of spoil, or a square metre of reinstatement.

Our specification puts the breaker hire on that stretch at $40–45 a metre — 2.16× to 2.43× the bid rate for the plant line alone, which is exactly the $40.00 the invoice arithmetic above lands on. That is the one rock figure we will put a number to, and it is deliberately the component rather than the total. It is a planning band in a spec, not a measurement from a job.

We are not going to give you an all-in rock rate, and the reason is the subject of this page. An all-in figure needs a labour number for those four days, and the honest answer is that we have two — what the crew’s day actually cost, and what a day of digging is worth at the bid rate. Adding invoiced plant hire to a bid valuation produces a number that looks like an achieved rate and is not one. That is the exact confusion §2 is about, and we are not going to commit it in our own copy to reach a rounder figure. The component is checkable against a hire invoice. The total would only be checkable against an assumption.

The point is not the exact forecast — it is that a ratio above 1.0, applied to the metres still unbuilt, turns one bad stretch into a number about the end of the job. In our scenario that warning fires on day seven of a twenty-one-day site, off two inputs: dollars spent on trenching, and metres of trench.

Where the waste hides

Two more entries from the same site, and they are the ones that never appear on an invoice as themselves. Half a day of standby when the excavator’s hydraulics fail: $900. Forty metres of duct laid short on depth and pulled back out: $2,100 of rework.

$3,000 — 8.6% of the entire bid, and 11.4% of the final overrun, from two events that produced no output at all. Spread across the site’s 1,150 m that is $2.61 a metre, or 14.1% on top of your bid rate, purely for work that had to be undone or was not happening. Most cost coding has nowhere to put that, so it gets absorbed into labour and disappears. Ours gives rework and standby_downtime their own categories for exactly this reason: a number you cannot see is a number you cannot bid against next time.

5. Cost per home passed, and what it is really made of

Cost per home passed is the number your client’s board asks about, and it is a ratio of two things you already have:

cost per home passed = cost ÷ homes passed

Across the scenario’s six sites that runs from $380 to $802 a home on a deployment averaging $504.85. A 2.1× spread on one build, priced by one estimator, in one town.

The instinct is to blame route density: some areas are just more spread out, more metres of trench per home. Test it. Trench length per home across those six sites runs 11.25 m to 12.71 m — a spread of only 1.13×. Put the most expensive site’s route density on the cheapest site’s rate per metre and you get $386.78 a home against $380.43. Density explains 1.7% of a 111% gap.

Which means cost per home passed is almost entirely a rate story wearing a density story’s clothes, and it decomposes exactly:

cost per home = (trench metres per home) × (cost per trench metre)

That identity is worth writing on something. The left side is what you get asked about in the monthly review. The right side has one term you basically cannot change after design freeze, and one term that moves every single day the ground surprises you — and it is the second term that your reporting does not measure.

6. Why nobody calculates the achieved rate

Everything above is arithmetic a competent PM could do in Excel. So why does it not get done?

Not because the sums are hard. Because both halves of the division arrive as messages, and neither becomes a number without somebody retyping it.

The metres come in as “did 110 m today on site 3” at 6pm, sometimes as a voice note. The hire comes in as “breaker hire booked 4 days 1450 a day.” The asphalt comes in as “the stretch past the crossing is all blacktop not dirt, thats not what the plan says” — a design error, a repricing event, and a rate observation, reported within minutes of being found, in a complete sentence.

None of it is missing. It simply never becomes data. Between the voice note and the division sits a person opening a thread, working out which site and which shift, and typing. In a quiet week they get most of it. In a week with a rock in it they get the dramatic parts and lose the boring ones — and the boring ones are the denominator. A plain day of 110 m in soil is the most useful row you will collect all month and the one most likely to go untyped, because nothing happened.

So the achieved rate survives as a story. “Rock’s always bad on the north side.” Probably true. Not a rate. Not something you can multiply by the 400 m you have left.

The failure is mechanical, not cultural. The structuring step is manual, so it is optional, so it is skipped exactly in the weeks the data would have been worth most.

7. What this does not do

These are our own numbers, and there is no customer in them. Harmis is pre-pilot: nothing described here is shipped software, and every product statement on this page is design intent from our specification. Every figure on this page comes from the pilot scenario in our own MVP specification — seed data we build and test against — or is arithmetic performed on it. It is illustrative, not a case study and not a benchmark, and we publish no industry averages because we have none worth publishing.

The achieved rate is an attribution, not a measurement. Dividing trenching-phase cost by metres trenched puts some cost on the trench that a strict allocation would put elsewhere. Cleaner would be per-activity coding of every cost — which requires a crew lead to attribute every receipt to a task, and that will not happen reliably, ever. Our specification names this as an open question because it is one. A rate labelled as an attribution beats one pretending to be a measurement.

A ratio from a thin window lies in both directions. Measure the achieved rate over two days and one bad afternoon forecasts a catastrophe; pick the wrong two days and it forecasts a triumph. What makes it usable is the window — ours is fourteen calendar days, and when the window holds no reported metres the ratio defaults to 1.0 and forecasts the rest at bid rate rather than inventing a trend. We clamp the ratio as a week-one backstop, and would rather say plainly that a clamp does not make one bad day safe than imply it does.

And it does not fix the bid you have already submitted. What a metre-level rate history does is make your next bid in that ground an observation rather than an opinion, and it needs one clean deployment of data before it is worth anything. Anyone selling you historical bid intelligence today, without your data in it, is selling you somebody else’s averages.

8. The one sum to take away

Find the biggest line in your current bid. Divide it by the bid total.

If it is 60% of the money, then half of it done is 30% of the job, and any progress figure that says otherwise is costing you every week between now and the final account.

Then do the other division — that line’s cost to date, over the quantity actually delivered — and compare it to what you priced. That ratio is the only number on your site that knows something about the metres you have not dug yet.

Both sums take a minute. The reason they do not get done is not the arithmetic. It is that the metres and the money are sitting in a WhatsApp thread, and somebody has to type them in.

Frequently asked

What are the components of a unit rate in construction? Plant, labour, materials, disposal, reinstatement, traffic management, preliminaries and overhead, and risk and margin. On a trenching line, almost all of them move together when the ground changes — which is why surface type is the dominant driver of a trenching rate and why a rate bid on soil does not survive asphalt.

How do I calculate the actual unit rate mid-project? Cost attributable to the activity, divided by the quantity actually delivered, measured over a window long enough not to be dominated by one day. The arithmetic is trivial. Getting a trustworthy numerator and denominator on the same day is the entire difficulty, and it is a data-capture problem rather than a calculation problem.

Is this earned value management? It borrows EVM’s most useful habit — weighting progress by value rather than task count — and drops the static baseline. A rate you are achieving this fortnight is a live input; a baseline is a memory. You do not need EVM software to weight your progress by value. You need your bid lines and one division.

Isn’t cost per metre too crude to forecast with? It is crude, and it is the number that moves. A crude measure that tracks the dominant variable beats a sophisticated one that needs data nobody will enter. If your reporting cannot produce cost per metre, it certainly cannot produce anything finer.

Why is your cost per home passed so different between sites? In the scenario it is almost entirely rate rather than route density — density accounts for 1.7% of a 111% spread. That is worth testing on your own build, because if the same holds, per-home reporting is telling you about the ground and you have been reading it as a design story.

Do you have results from customers? No. Harmis is pre-pilot, with no customers and nothing from the field to show you. Every number on this page is from our own documented pilot scenario, labelled as such, or is arithmetic on it. When we have a pilot’s numbers they will be that pilot’s numbers, and they will say whose they are.

Check the arithmetic

Every figure on this page is derived from the pilot scenario in our MVP specification by multiplication and division, and we would rather you verified it than took it. The four that carry the argument: