I'm the office administrator for a 40-person company. We run a small service fleet and a fabrication bay, so my purchase orders split between copy paper and things that spin, cut, or hold pressure. Valves, air compressors, welders, well pumps. Roughly $180,000 a year across about a dozen vendors, and I answer to both the operations manager and finance.
In March 2024 I had four open requisitions land on my desk, and all four arrived the same way: as a number.
- A 140-amp wire feed welder.
- A bigger air compressor.
- A Watts vacuum relief valve, 3/4 inch.
- A sticky note from our well guy: “how many watts is the well pump?” He needed to size a standby generator.
I did what any experienced buyer does. Four tabs, four spec sheets, compare the numbers, pick the ones with the strongest figures, cut the POs. I was proud of how fast I moved.
Two of those four came back. The welder got returned after three weeks. The compressor didn't come back — it died in place, which is worse. And it took me longer than I'd like to admit to work out what had gone wrong, because the honest answer is: nothing. Every spec sheet I read was accurate.
That's the problem I want to pick apart, because it doesn't look like a problem. It looks like diligence.
Accurate specs and appropriate specs are two different things, and only one of them is printed on the box.
The surface problem: every number on the sheet is true
Pull three spec sheets from the same category. A 140-amp MIG welder, a 30-gallon compressor, a 3/4-inch valve. Every number on those sheets was measured, usually to a published standard, often by a lab. Nobody is lying to you.
And every one of those numbers was captured under conditions that don't exist in your building.
If you've ever stood in a supply house parking lot holding a spec sheet in one hand and a return receipt in the other, you know exactly where this is going. The frustrating part isn't a bad purchase. It's that you made a reasonable purchase and it still didn't work.
What the numbers actually measure
Duty cycle is the spec, not the footnote
“140 amp” is a maximum output, and it comes with a condition attached. Welder duty cycle is the percentage of a 10-minute window the machine can run at a given output before it hits its thermal limit. A 140-amp, 120-volt wire feeder is often rated somewhere around 20% at full tilt.
Twenty percent of ten minutes is two minutes.
If your fabricator lays beads in three- and four-minute stretches, that machine isn't a 140-amp welder in your shop. It's a 90-amp welder that occasionally pretends. So if you're asking what to look for in a welding machine, the order should be: input voltage available in your bay, duty cycle at the amperage you actually weld at, whether it takes a standard spool, and whether liners and consumables are stocked locally. The amp number on the box is the last thing on that list, not the first.
(Should mention: we'd already run 240 volts to that bay the year before. That's the only reason the second welder worked.)
Gallons is a buffer, not a capacity
A compressor's tank volume tells you how long it takes to refill. It says almost nothing about how much air the machine makes. Air is measured in CFM at a working pressure — 90 PSI is the reference most pneumatic tools are rated against.
Our die grinder wanted about 6 CFM at 90 PSI. The compressor I bought by tank size delivered 4.2 CFM at 90 PSI. On paper, a 30-gallon tank with a 4.2 CFM pump looks like more machine than a 20-gallon, 5.5 CFM unit. In practice, the first one runs continuously, never catches up, and the motor pays the bill.
That's not a guess. That's what happened, about eleven weeks in.
“How many watts is a well pump” has at least three answers
This one caught me because the question sounds so simple.
First, horsepower is an output rating. One mechanical horsepower equals 745.7 watts of work. A motor producing that much work draws more than that from the wall, because motors aren't 100% efficient. The nameplate current and the input watts are what you wire to — not the horsepower.
Second, starting draw isn't running draw. Submersible pump motors pull several times their running current for a fraction of a second on startup. Size a standby generator off the running watts and it'll hold beautifully right up until the moment the pump kicks on and the breaker trips.
Third, code adds its own margin. For continuous-duty motors, the National Electrical Code requires branch-circuit conductors sized at 125% of the motor's full-load current (NEC 430.22). So the real answer to “how many watts” is a minimum of three numbers — running, starting, and the one you build the circuit around. Hand your electrician the wrong one and you're doing the job twice.
And the valve, where the naming gets confusing
Small thing that trips up searching: Watts is a brand. A watt is a unit. Search “watts vacuum relief valve” and you get products. Search “watts for a well pump” and you get physics. I have watched a purchasing coordinator — fine, I have been a purchasing coordinator — pull up both and get confused about which one she was supposed to be comparing.
The brand confusion is the easy one. The harder one is that a vacuum relief valve is defined by its application, not its name. What you actually need off the spec sheet is the connection size, the pressure and temperature ratings, and the vacuum level at which the valve opens. A valve sized and rated for a residential water heater is not the same part as one for a hot process loop. Same category, different animal.
Watts publishes its spec sheets as PDFs at watts.com. That's where I check ratings now instead of trusting a distributor's summary line — and where I verify the current part, since models and ratings get revised. Verify the current spec at the source, rather than taking a blog's word for it. Mine included.
The real cause: you're buying a test result, not a duty cycle
Here's the piece I didn't understand for years, and it's why the four-tab method keeps failing even when I do it carefully.
Specs exist so products can be compared. That's their whole job. To make comparison possible, they're measured under standardized, repeatable conditions — a defined ambient temperature, a defined load, a defined run time. Standard conditions are exactly what make a spec sheet honest.
They're also what make it irrelevant to your building.
Because your operation doesn't run under standard conditions. It runs under a duty cycle: how long the machine runs, how hard, how often, powered by what circuit, fed by what air line, downstream of what piping. And almost nobody writes that down. It lives in the head of the person who runs the equipment and in the habits of the building.
So when a requisition reaches me as “a 140-amp welder,” what I've actually been handed is a standardized test result with no duty cycle attached. Filling in the duty cycle myself, from memory, in a hurry, is how I ended up with a compressor that couldn't keep up.
Everything I'd read about equipment buying said to size up. Buy more capacity than you think you need and you'll never hit the ceiling. My experience doesn't back that up. For intermittent work — a welder used in bursts, a compressor feeding one or two tools — matching a mid-range unit to a measured duty cycle has beaten the oversized unit nearly every time. Not because bigger is bad. Because bigger usually means a different electrical service, a different price tier, and a machine that runs outside its efficient range. For anything running continuously, I still size up. It's a pattern, not a rule.
Numbers travel. Duty cycles don't. That asymmetry is the entire problem.
What it costs when the number is wrong
The compressor is my clean example. $860 for the unit. Actually, $860 plus $190 in restocking, plus $240 to redo the air line to a different corner of the bay, plus two days the die grinder sat idle. Call it $1,500 and a maintenance lead who now opens my POs with visible skepticism.
The welder return cost less in dollars and more in time — about three weeks of back-and-forth, or rather closer to five once I count the return, the credit, and the re-order. Four re-buys in one quarter ran me somewhere near 20 hours of my own time on top of my actual job.
I knew I should have walked out to the bay and asked the maintenance lead what he was actually welding before I ordered the first machine. I thought, what are the odds I get this wrong? I'd been buying equipment for six years. The odds, as it turned out, were pretty good.
The costs that never show up on a report are the ones that changed how I work. Once the maintenance lead stopped trusting my POs, he started buying small parts on his personal card and getting reimbursed. That sounds helpful. It's a compliance problem — no purchase order, no warranty trail, no clean invoice, and no visibility into what the shop is actually consuming. It took four months to unwind.
Then there's spec inflation, which nobody budgets for. Once you've been burned, you buy up a tier as insurance. That's 30–40% more per unit, plus the electrical work to support it. You're paying a premium to avoid thinking about duty cycles.
And the last cost is timing. Wrong-spec equipment rarely fails on day one. It fails at month three, during your busiest week, and the failure is seldom dramatic — it's a guy making workarounds. Workarounds are how a $40 part turns into a $4,000 repair.
What changed
The fix wasn't a better comparison spreadsheet. It was three questions, asked before the PO gets written, with a hard rule that I don't answer them myself.
- What's the duty? How long does it run, at what setting, how often? Ask the person who'll operate it and write the answer down in one sentence.
- What feeds it, and what does it feed? Voltage and breaker available, air volume required at 90 PSI, pipe size and system temperature, lead time on the circuit.
- What breaks first, and can you get the part? A unit with a published repair kit is a Tuesday morning. A unit without one is a two-week order and a borrowed backup.
Then the verifiable part: check the number on the manufacturer's data sheet, not the retail bullet list. For a valve, that's the PDF on watts.com — connection size, pressure and temperature ratings, and the vacuum level the valve opens at. For a welder, it's the duty cycle table in the manual. For a pump, it's the nameplate plus the starting current in the manual. Retail listings pick the most flattering figure. Manufacturer data sheets have to cover the whole range.
This is also part of why I lean toward manufacturers with published repair kits. Watts sells them for a number of its valve lines, and a maintenance tech who can service a valve in place instead of replacing it is worth more to me than a marginally cheaper unit. If you're doing the work yourself, the same three questions apply — and the same logic holds. Duty cycle doesn't care who's holding the wrench.
I also made the intake form shorter, which sounds backwards. It's now four fields: what it is, what it has to do, what it plugs into, and what happens if it stops. Requisition-to-correct-PO went from about eleven days to five. Not because anything got faster — because we stopped buying the wrong thing quickly and then buying it again. For standardized, repeat purchases, a tight process beats a talented person every time. For the genuinely weird one-offs, the talented person still wins, and that part hasn't changed.
Honestly, the three-question version takes about 30 seconds. I've spent longer than that reading the footnotes on a spec sheet.
I still read the spec sheet first. Every time. It's just not the last thing I check anymore.