A spreadsheet is enough while your scope repeats and the file is small enough to read on one screen. After that it starts costing you money without telling you. Dartmouth researchers audited 50 spreadsheets in real business use, across 270,722 formulas, and found at least one error that produced a wrong number in 43 of them — 86%. An estimate has one bottom line, so a single bad cell anywhere in the file reaches the customer. Five percent light on a $14,000 bathroom is $700 you will never see again. Keep the template while the jobs repeat. Replace it when scope is decided at the address rather than in the file.
When is an estimate template actually enough?
When you are pricing the same work over and over and the file is doing arithmetic you could check by hand.
A template holds up when all four of these are true:
- Your line items barely change between jobs.
- The whole estimate fits on one screen, roughly 40 to 60 rows.
- You are the only person who ever opens the file.
- You can say out loud what month you last updated the material prices.
That is a real share of contractors, and anyone telling you those people need software is selling something. A siding crew running the same four assemblies does not need an estimating platform. It needs a clean sheet and a price list that is current.
The fourth condition is the one that fails first. Not the formulas — the prices. Most estimate spreadsheets I have seen carry a material column somebody typed in during a slow week, and nobody remembers which slow week.
What does an estimate template actually cost?
Not zero, and the licence is the least of it.
| What you are running | Price |
|---|---|
| Google Sheets, personal account | $0 |
| Microsoft 365 Personal (Microsoft list, Sept 2026) | $99.99/year |
| Office Home 2024, one-time licence (Microsoft list, Sept 2026) | $179.99 |
| Building the price list the first time | most of a Saturday |
| Writing each estimate | 30-60 minutes |
Put the licence aside. Twenty estimates a year at 45 minutes is 15 hours, and that is the number worth arguing about, because it is the same 15 hours whether the tool cost you nothing or a few hundred a year. The comparison that matters is not free versus paid. It is what each one gets wrong.
Worth knowing what your competition is looking for, too: over the twelve months to September 2026, US search interest in "estimating software" ran several times higher than "contractor estimate template" or "construction estimate template", which barely lift off the axis on the same chart. Within the software term, "job estimating software" was up 110%. Trends is relative interest and not volume, so treat it as direction. The direction is away from the template.
Where do estimate spreadsheets actually go wrong?
Not where you would guess. Excel's arithmetic is perfect; the file is wrong because of what a human put in it. The Dartmouth audit sorted every fault it found, and three types accounted for almost everything.
| Error type | Share of instances | What it looks like on an estimate |
|---|---|---|
| Hard-coding a number over a formula | 37.7% | The subtotal looked high, you typed $4,200 over it, and that cell has been dead ever since |
| Reference error — formula points at the wrong cell | 32.9% | You added "dumpster and haul-off" at the bottom and the SUM above it still stops at the row it stopped at last year |
| Logic error — the formula does the wrong thing | 21.9% | Markup applied to labor but not materials, or applied twice on the tax row |
Copy/paste, omission and bad data input together made up under 5%.
Look at the second row, because that is the one that costs real money on a job site. You add a line at the bottom of the sheet — the extra electrical the homeowner asked for while you were standing in the kitchen — and the total does not move. The number on the page is still right for the estimate you wrote yesterday. It is wrong for the job you are about to sign, and there is nothing on screen that looks unusual. You find out in week three when the dumpster invoice shows up against a job that never carried a dumpster.
The first row is worse over time. Overwriting a formula is what everyone does under pressure, and then that file gets saved as the template for the next house. Every job after it inherits a cell that stopped calculating in March.
That is the honest difference between a template and software. The spreadsheet is not less accurate than an app at arithmetic. It just has no way of telling you that a line you added is not in the total, and no memory of what the number is supposed to be built from.
How do you make a spreadsheet estimate safe enough to use?
Five habits, and they cost you nothing. This is most of what a paid tool gives you on a small job anyway.
- Keep prices on their own tab. One column of material rates, one of labor rates. Nothing priced inline in the estimate, ever.
- Lock every calculated cell. Protect the sheet, leave quantity cells open. Now you physically cannot type over a total at 9pm.
- Make the total range go past the last row. Sum ten rows further down than you need and keep the spare rows blank. New lines land inside the range instead of underneath it.
- Date the price tab in a cell you can see. "Prices as of 03/2026" at the top. If that date is older than six months when you open it, you are quoting from last year.
- Send a PDF, never the workbook. An .xlsx is editable, shows your markup column to anyone who scrolls right, and there is no version of that conversation you win.
Point three is the one nobody does and it prevents the most expensive failure on the list.
Habits will not save the file forever, though. What none of this fixes is a price list built on an hourly rate that was wrong to begin with — if you have not worked out what your overhead costs per hour, a locked, dated, perfectly summed spreadsheet will produce clean numbers that are too low on every job.
When should you stop using a template?
When any one of these is true. Not all of them. Any one.
Scope is decided at the address, not in the file. Old houses, remodels, anything behind a wall. A template prices what you planned for. What you are pricing is what you found.
Somebody else has to open the file. The moment a second person edits it, you have version drift — two people, two copies, one of them stale, and a real estimate went out of the wrong one. The file called "Estimate_Jenkins_FINAL_v2 (1)" is the one that got emailed. It always is.
You are writing estimates in the truck. Spreadsheets are a desk tool. Anything you postpone to 9pm gets priced from memory, and memory keeps the tile and forgets the soft subfloor. That is the mechanism behind most underestimating and it has nothing to do with knowing your trade.
You cannot say what date the prices were set. Covered above, and it is the most common of the four.
You are losing bids on speed. If the homeowner had two numbers before yours arrived, the quality of your spreadsheet was never the issue. The cost arithmetic for that decision is worked through in whether estimating software is worth it for a one-person shop.
What this misses
Those audits are not construction. Powell and his colleagues audited business spreadsheets — budgets, forecasts, operational models — mostly far larger than a 50-row estimate. Size drives error rates, so your file is safer than theirs. Take 86% as a warning about where spreadsheets go, not a prediction about yours.
The cell error rates are also lower than the scary version you have read elsewhere. Powell's team put them at 0.87% of formula cells under a strict definition and 1.79% under a loose one, against the 5% figure Panko's earlier field audits implied. In their sample, 46% of spreadsheets had error rates under 2%. Worth saying plainly, because an argument that needs the number inflated is not an argument.
And research this specific has an age problem. Panko's field audits ran from 1987 to 2000; the Dartmouth study published in 2009. Nobody has re-run either one on modern files. The failure modes have not changed, but do not treat those percentages as current measurements.
Last thing, on the licence. Software does not remove any of these errors. It removes the ones caused by the file having no idea what it is for — a total that does not include a line, a price nobody dated. It cannot tell you that you never wrote down the soft subfloor. That is what BidWalk is for: getting the line items down while you are still in the room and the homeowner is still talking, so the file you are checking has the whole job in it.
