The $28 Router Bit That Cost Us $2,900

In March 2024, one of our foremen sent me a photo of eight maple stair treads. About $2,100 in material, and a 3/4" roundover profile that looked less like a machined edge and more like something had chewed on it. His message was one line: "Bad batch of maple?"

It wasn't the maple.

We ran the same boards on the same shaper, same feed rate, same RPM, with a reference bit we keep for exactly this kind of argument. Clean edge. No chatter, no burnishing, no fuzz. So I pulled the bit he'd actually been running — a house-brand 3/4" roundover he'd picked up to save about $28 — and put it on the indicator.

Total indicated runout: 0.006". Our incoming spec is 0.002" or better. And that was cold, before the bit had done any work.

Six thousandths of an inch doesn't sound like much. It's roughly two sheets of copy paper. You could almost convince yourself it's noise in the measurement.

It's not noise. At 22,000 RPM, a bit with 0.006" of runout has one cutter doing most of the work and the other doing almost none. That loaded cutter runs hot, dulls early, and starts tearing wood fibers instead of slicing them. Which is exactly what we were looking at on those treads.

Here's the part that gets me: nobody on that crew made a careless decision. They made a reasonable one. The bit was $28 cheaper, it looked identical in the package, and the supply house had it on the shelf. That's the trap.

What's Actually Different Inside the Bit

Strip the packaging off a $22 roundover and a $95 one and they look the same. Same profile, same color, same 1/2" shank. The difference lives in three places you can't see from six inches away.

1. The carbide

Cutting tools use micrograin carbide, but "micrograin" covers a range. Grain size affects how sharp an edge you can grind and how long that edge survives in abrasive material — and hardwood with silica in it is abrasive. Cheaper bits tend to run coarser grain and a softer binder. They take a fine edge out of the box and lose it faster once things heat up.

You can't measure that with a caliper. You find out at linear foot 600.

2. The braze

The carbide tip is brazed to a steel body. If the braze isn't clean or the tip isn't seated properly, you get voids, and voids mean the tip shifts slightly under load. Not enough to throw the bit. Enough to change your cut. Bits that run hot tend to fail at the braze first.

3. Geometry — where spiral bits earn their price

A straight-flute bit hits the wood at ninety degrees. Every cut is an impact. Freud spiral router bits present the edge at a shear angle, so the cut is a slicing action rather than a chipping one. That shear angle is why a spiral leaves a cleaner edge in figured maple, and why it clears chips out of the cut instead of packing them in.

It's also why they cost what they cost. Grinding a helical flute on a solid carbide blank is slower, more tooling-intensive, and generates more scrap per unit than grinding a straight flute. There's no trick to it. It just costs more to make.

Same story on the saw side. A Freud blade with proper tooth geometry ground into premium carbide isn't priced that way because of the logo on the plate.

The Spec Sheet Is Not Marketing

I have a slightly odd background for this. Before millwork I spent a few years in facilities and shop maintenance, and it ruined me for "just eyeball it."

Take a question I still get from homeowners and, honestly, from a couple of our own guys: how many turns on garage door torsion spring? The answer is a number, not a feel. The usual rule of thumb is one full turn per foot of door height, so a standard 7-foot door lands around seven turns — 28 quarter-turns — with the exact figure depending on the drum and the spring rate. You look it up on the chart for your specific drum. You don't guess, because guessing wrong gets you either a door that won't balance or a spring that lets go, and one of those can put somebody in a hospital.

Cutting tools are less dramatic. The principle is the same: there's a specification, there's a tolerance, and your opinion about it doesn't change the physics.

Runout has a spec. Balance has a spec. Chipload is arithmetic, not preference:

Chipload = Feed rate ÷ (RPM × number of flutes)

At 18,000 RPM with a 2-flute bit and a 200 IPM feed, chipload works out to about 0.0056" per tooth. Drop the feed to 100 IPM and you're at 0.0028" — thin enough that the cutter starts rubbing instead of slicing. Rubbing makes heat, heat kills the edge, a dull edge burns the wood. That's the entire failure chain, and it starts before a single board gets ruined.

Which is why I stopped accepting "we ran it the same way we always do" as a quality answer.

What That Bit Actually Cost

Let me do the arithmetic the way it showed up on the job.

Saved: $28.

Spent, on that same job:

• $1,400 in material value. Eight treads couldn't be sold as stain grade. We downgraded them to paint grade and ate the difference.
• About six hours of shop time re-cutting replacements and re-running the profile on the salvageable pieces. Roughly $540 loaded, and it pushed two other jobs back a day.
• Around $380 in finishing rework. Tearout that doesn't show until stain goes on is the most expensive kind, because by then you've already paid to sand and finish it once.
• A $620 rush charge on replacement maple, because we were already inside the install window.

Call it $2,900, against a $28 saving. The budget bit was roughly a hundred times more expensive than the one I would have bought on purpose.

And that's the clean version of the story. The messy version is the one where the defect ships, the customer finds it three months later on a stair they walk on every day, and now you're talking about a callback, a reputation, and a general contractor who quietly stops calling.

The $62,000 Tool Box in the Room

Here's the thing about tooling budgets: people spend money where they can see it.

I sat in a shop owners' meeting once where a $62,000 tool box got approved in about four minutes. Stainless top, full-extension slides, the works. It's a genuinely nice piece of equipment and I'm not knocking it. It's also furniture. It doesn't touch the work.

The same group then spent forty minutes debating whether an extra long ratcheting wrench set was worth the upcharge for the extra reach on one specific hydraulic fitting. That's at least a functional argument, so fine.

And at the end of that meeting, a router bit order got approved on price with no discussion at all. A single 3/4" roundover was cutting every stair tread in the building that month. It was the cheapest line item in the room and the one item that could take the whole month down.

That's the pattern. The box is visible. The bit isn't. So the box gets scrutiny and the bit gets a purchase order.

I'm not saying buy the most expensive thing on the shelf. I'm saying the number on the invoice is the smallest number in the decision, and most shops never bother to look at the rest of them.

The Math We Run Now

We don't do anything clever. We stopped pretending the purchase price was the cost.

Cost per linear foot is the number we track: bit cost divided by the linear feet it cuts before it's pulled, plus the scrap it generated, plus the labor to fix what it ruined. The house-brand roundover at $22 might give us 700 to 800 feet before it starts burning. The Freud version gets us past 4,000 feet on the same material with close to nothing in the offcut pile. On those two numbers alone the expensive bit is cheaper. The scrap is what widens the gap.

Three things changed on the floor:

1. Incoming inspection. Any bit headed for a production machine gets checked on a dial indicator. Over 0.002" TIR, it goes back. Not a judgment call anymore.
2. Standard profiles are standardized. Our three highest-volume profiles run a Freud router bit set, full stop. No mid-run substitutions, no "this one's basically the same."
3. A simple log. Linear feet per bit, per profile. It's a spreadsheet, not a system. But after two quarters we had numbers instead of opinions, and most of the arguments stopped on their own.

Tooling is maybe 2% of our consumables budget. It's the 2% that decides whether the other 98% was worth anything.

The cheapest part of a cut is the bit. That's exactly why it's the one people cut corners on.