Setting up a calibration system from scratch
No register, no records, no procedure, and probably an audit on the horizon. This is the plan the quality-forum veterans converge on, step by step, with their own words as evidence.
You have been asked to set up calibration for a company that has never had a formal system. There are instruments on the floor, some with stickers from years ago, some with nothing, and it is now your job to turn that into a program an auditor will accept. Nobody handed you a procedure, and the internet gives you forty opinions per question.
The good news is that the shape of a working calibration program is not controversial. People have been asking this exact question on the quality forums since 2002, and the veterans give the same answer every time. This guide is that answer, laid out in the order that works, with nothing that requires a budget until the last step.
“After trawling through endless information on the internet on various sites my head is starting to spin... Basically I need to set up the calibration of these gauges from scratch”
The full scope, from someone who has built many
Before the steps, it helps to see the whole machine at once. A calibration program is not a list of due dates: it is a register of what you own, a rule for when each item is calibrated, a record of every calibration that happened, somewhere certificates live, a way of being told before things fall due, and a plan for the day something comes back out of tolerance.
A calibration veteran on Elsmar Cove gave a newcomer the honest version of that scope:
“you will also need to create an entire calibration program. This will include setting up a calibration recall program, establishing a calibration history database, maintaining the calibration - and traceability - of your lab standards, producing and storing calibration records, training programs, a recall system for out of tolerance standards, etc. ... And you need to do this well enough to... be 'audit proof'.”
Start with the register, not the software
Every failed setup starts the same way: someone evaluates tools before they know what they own. The register comes first because every other decision hangs off it. How many instruments, of which types, in which places, used for what: those numbers decide your intervals, your budget, your lab lead times, and whether you need software at all.
The situation you are avoiding is the one this quality manager at a small AS9100 shop walked into in June 2026:
“I recently inherited a situation involving hundreds of pin gages ... There has never been a formal identification, tracking, calibration, or verification system for them.”
The whole requirement, in one buyer's sentence
When people who run calibration in small companies describe what they actually need, the list is short. Not modules, not workflows: a reliable warning before things fall due, and one place where the history and the certificates live. Everything else in this field is elaboration on those two jobs.
One Excel-outgrower on Elsmar Cove compressed the entire specification into a sentence:
“We are using excel but we need a program that would provide notification on when calibration is due and an area to scan certificate / history of gauges.”
Simple systems work, until a knowable point
You do not need software on day one, and anyone who tells you otherwise is selling something. A spreadsheet or even a card file does the job for a small fleet, provided it has a named owner and a written checking cadence. The honest question is where that stops, and the field has had a consistent answer for twenty years.
A metrologist and auditor drew the line back in 2004, and the number has held up:
“A recall system using Excel or cards in a box can work well.....for small amounts of equipment. Larger amounts of equipment, say 300 items or more, is best managed using software designed for that.”
Whatever you do, do not write your own program
Somewhere around step seven, if you can code a little, you will be tempted to build the system yourself: a macro that emails, an Access database, a small app. The forums are a graveyard of those projects, and the failure mode is always the same. The build works, the builder leaves, and the company inherits software with exactly one former maintainer.
A veteran moderator on Elsmar Cove put the case against it bluntly, all the way back in 2004:
“I would NEVER waste the time and energy to write my own program ... Guys who only know Excel are like little kids with hammers - everything starts to look like a nail. ... Even the most expensive commercial software product is less expensive than the time it takes to define and structure a program then write it and maintain it and train others how to use it ...”
The plan, in the order that works
Steps 1 through 9 need no software and no budget, only time and a walk around the floor. Step 10 is a decision, not a purchase. Do them in order: every step depends on the ones before it.
Inventory everything that measures
Walk the whole floor, including drawers, toolboxes and the goods-in bench. List every device used to accept product, reject product or control a process: make, model, serial number, location, and who uses it. Do not filter yet; a complete ugly list beats a tidy incomplete one.
Sort the list: calibrate, verify, or reference only
Not everything needs the full treatment. Instruments whose readings decide product quality get calibrated. Some tools only need periodic verification against a known reference. Devices used for rough indication can be marked as reference only and excluded, with a durable tag saying so. This single sorting step is what keeps the program affordable.
Give every instrument a permanent identifier
Assign a unique ID to each item that stays in the program, and fix it to the instrument: engraved, etched or on a durable label. The ID is what connects the physical item to its row in the register, its records and its certificates. Serial numbers alone fail here because duplicates and unreadable plates are common.
Set a starting interval per type, and write down why
With no history, use published starting recommendations per instrument type and record the source next to the interval. Treat every interval as a risk decision you will revisit: guidance documents such as ILAC-G24 and OIML D 10 describe adjusting intervals from usage and calibration history, not fixing them forever.
Decide internal versus external per type
Some checks you can do in-house against a reference you own, if you write down the method. Anything beyond that goes to an accredited laboratory. Call two or three labs early and ask about lead times and pricing per type; lead time is the number that decides how early your recall list must warn you.
Define the record you will keep for every calibration
One record per calibration event: instrument ID, date, who performed it, the procedure or reference used, the as-found and as-left results, pass or fail, and the next due date. Keep the certificate with the record. If a form does not capture as-found and as-left, fix the form now; that pair is what makes history usable later.
Build the recall list and give it an owner
Put every due date in one place, sorted by date, and write one short instruction: who checks the list, how often, and what they do when something comes due. This is the component that fails silently in homegrown systems, so name a deputy as well. A list nobody is assigned to check is a list that will be discovered by an auditor.
Write the procedure, one page if you can
Scope, identifiers, how intervals are set, what records are kept and where, how recall works, and what happens when an instrument is overdue or fails calibration. Short enough that people read it, specific enough that a stranger could run the program from it. Your future successor is the reader you are writing for.
Plan the out-of-tolerance response before you need it
Decide now what happens when a calibration comes back with a failed as-found: who assesses which product was measured with that instrument since its last good calibration, who decides on containment, and where that decision is recorded. Improvising this after the fact, with product already shipped, is the worst place to design a process.
Decide what the system lives in, using your real numbers
Now you know the instrument count, the types and the lab cadence. Under roughly 150 instruments, the spreadsheet or card file you built in the earlier steps is honestly enough if it has a named owner. Around 300 and beyond, the people who run these programs stop recommending manual systems. Choose deliberately, and revisit when the count grows.
Where Gaugelog fits, honestly
Gaugelog is the register, records and recall list from this plan as one piece of software. Instrument types come with sourced starting intervals built in, so step 4 starts from documented recommendations instead of a blank cell. The recall list runs itself: escalating email reminders at 30 days, 14 days, 7 days and on the due day, then weekly while anything is overdue, whether or not anyone opens the register. If you started in a spreadsheet, you upload it as it is and your history imports with it.
- If your inventory lands under roughly 150 instruments, a well-kept spreadsheet with a named owner is honestly fine. The free calibration log template below is that spreadsheet, structured to match this plan.
- Gaugelog does not calculate measurement uncertainty and is not built for ISO 17025 laboratory scopes.
- Nothing in the 10 steps requires software of any kind, this product included.
Questions, straight answers
Do I need software to start a calibration program?
No. A program is a register, intervals with a written basis, records with certificates, and a recall list someone owns. A spreadsheet holds all of that for a small fleet. Software earns its keep when the fleet grows past what one person reliably checks by hand, which the field has long placed around a few hundred instruments.
How do I choose calibration intervals when I have no history?
Start from published recommendations for each instrument type and write the source next to the interval. Then let your own as-found results move it: instruments that keep passing can earn longer intervals, instruments that drift get shorter ones. Guidance documents such as ILAC-G24 and OIML D 10 describe exactly this adjustment loop.
What records does an auditor expect for each calibration?
For each event: which instrument, when, who performed it, against what reference or procedure, the as-found and as-left results, the outcome, and the next due date, plus the certificate when a lab did the work. Auditors follow the chain from register to record to certificate, so the identifier has to match across all three.
Can we do calibrations ourselves instead of using a lab?
Often, yes. In-house calibration or verification against a reference you own is legitimate if the method is written down and the reference itself is calibrated externally on its own schedule. Keep the reference's certificates with your records; that link is the first thing a careful auditor pulls on.
What should this cost to start?
The plan itself costs time, not money: the inventory, the sorting, the procedure and the recall list are all free. Real costs are external lab calibrations, which scale with how well you did the sorting step. Software, when the count justifies it, starts free: Gaugelog's Free plan covers 30 gauges with no credit card and no per-seat pricing.
Doing this job with an audit on the calendar?
Gaugelog is in development and launches in 2026. Until then you can generate a clean calibration certificate PDF with our free tool, no account needed, and compare the plans on the pricing page.