Multi-Part RFQ Quoting for CNC Shops, Done Faster

The RFQ that ruins your week doesn’t look dangerous. It’s one email with one attachment — a zip, maybe, or a single PDF with a table. Inside: twelve parts. A couple of simple turned components, a few milled brackets, one genuinely nasty 5-axis housing, quantities ranging from 5 to 500, and a tolerance table that’s different on every line.

Your customer thinks they sent you “a quote.” They sent you twelve. And that gap — between what an RFQ looks like and what it actually costs to price — is why multi-part RFQs are where manual quoting quietly falls apart.

An RFQ is not one quote

Here’s the thing that doesn’t show up in the inbox: there is no economy of scale in quoting by hand. Pricing twelve parts is not meaningfully faster per part than pricing one. Each part still needs the full treatment:

• Open the model, read the geometry, work out the stock.
• Read the drawing for threads, tolerances, finish and notes.
• Decide the process, setups and fixturing.
• Derive cycle time operation by operation.
• Roll up material, setup, tooling, finishing, overhead and margin.

If a single non-trivial part is one to three hours by hand, a twelve-part RFQ is — honestly — a multi-day job. The work doesn’t compress just because the parts arrived together. It stacks.

That’s the core problem in one sentence: a multi-part RFQ forces you to re-derive every part from scratch, and the cost is linear in the part count.

The four things that make it worse

The per-part re-derivation is the baseline pain. Mixed RFQs pile on four specific failure modes.

Mixed quantities break setup amortisation

One line wants 5 pieces; another wants 500. Setup is a fixed cost spread across the batch, so it lands almost entirely on each of the 5 and is rounding error across the 500. Get the amortisation wrong on even one line — quote the 5-off as if setup were spread thin, or the 500 as if each carried full setup — and that line is mispriced. Across a dozen lines under time pressure, at least one usually is.

Mixed tolerances reset your assumptions every part

A part-to-part jump from “general tolerances, bead-blast finish” to “±0.01 mm on three features, inspection report required” completely changes the machining strategy and the cost. You can’t carry assumptions from the last part forward. Every line resets the clock on careful reading — and careful reading is exactly what suffers when you’re on part nine of twelve at the end of the day.

Shared admin gets repeated instead of reused

Customer details, your branding, payment terms, lead time, the cover note — none of that changes across the RFQ, but in a manual spreadsheet-and-document process it tends to get rebuilt or copy-pasted per part anyway. It’s not the expensive part, but it’s pure waste, repeated a dozen times.

One hard part holds up eleven easy ones

There’s always one. The 5-axis housing with the deep pocket and the awkward tool access stalls the whole RFQ, because by hand you’re working serially. The eleven straightforward parts that you could price in minutes sit waiting behind the one that needs real thought. So the customer waits on all twelve.

What it costs you to be slow

Add it up and the multi-part RFQ is the quote that slips. It’s the one that moves from “I’ll have it to you this afternoon” to “early next week,” because there’s no afternoon big enough for twelve parts on top of everything else.

And late is expensive in a way that’s easy to underestimate:

• The first credible quote often wins. On a competitive RFQ, the shop that replies same-day is shaping the decision while you’re still on part four.
• Big RFQs are big jobs. The quotes most worth winning are exactly the ones that take longest to produce by hand — so your slowest quoting is concentrated on your highest-value work.
• Rushed multi-part quotes leak the most. When you finally push twelve parts out the door at 7pm, that’s where the missed callouts and the wrong setup amortisation hide. You can lose the job and lose money on the version you win.

Quoting the whole RFQ in one pass

The fix isn’t quoting faster by hand. It’s removing the per-part re-derivation entirely — doing the reading and the arithmetic for all twelve parts in a single pass, and keeping the judgement where it belongs.

Read every part at once. Drop in all the parts — STEP files and drawings together. Best-in-class AI models recognise the machinable features in each part’s CAD — the holes, pockets, faces, threads and harder 5-axis geometry — directly from the STEP, while the 2D drawing is read in parallel for the threads, tolerances, finish symbols and notes that move the price. Twelve parts get read together instead of one-at-a-time-after-coffee. The hard 5-axis housing no longer blocks the eleven simple parts, because they’re all in the same pass.

Price every part against your shop. Then a deterministic engine — fixed, transparent formulas, not a model’s guess — prices each part against your configuration: your machines and rates, your material costs, your setup times, your overhead and margin. Crucially, it applies the right batch economics per line automatically: the 5-off carries its setup, the 500-off spreads it, and you don’t have to remember which is which. Mixed quantities and mixed tolerances stop being a source of error and become just data the engine handles.

Assemble one quote, not twelve documents. Out comes a single branded quote: a per-part breakdown so the customer can see each line’s material, cycle time, setup, tooling and finishing, plus a roll-up total. The shared admin — customer, branding, terms, lead time — is filled once and applied across the whole RFQ. The thing that used to be twelve copy-paste jobs is one document.

Catch the problem parts up front. Because every part has actually been read, manufacturability issues are flagged per part before anything goes out — a tolerance that’s tight for the feature, a thin wall, an ambiguous note. And when something is genuinely unclear, the system asks you a question rather than guessing. You resolve the one problem line without freezing pricing on the other eleven.

The honest framing

This is deliberate, and it’s worth being precise about it: leading AI models do the reading, a deterministic engine does the pricing. The reading — twelve parts’ worth of geometry and drawings — is where speed and pattern-recognition win decisively. The pricing is where you want repeatability and an audit trail, identical on the first part and the twelfth. You get both, and you get a per-part number you can defend line by line.

What it doesn’t do is take the decision away from you. The whole RFQ comes back as a transparent breakdown, and you own it. Adjust a rate on one line, flex the margin on the parts you actually want, add a note on the awkward housing, drop a part you don’t want to bid. Every number recalculates instantly, and the whole RFQ goes out under your branding.

Next-week to same-day

That’s the real shift on multi-part RFQs. The toil — twelve parts of reading, twelve cycle-time derivations, twelve setup amortisations, the repeated admin — collapses into roughly a minute of compute plus your review. The judgement — which parts you want, at what margin, with what caveats — stays yours.

A twelve-part RFQ stops being the quote that ruins your week. It becomes the quote you turn around the same day the customer sent it — while they’re still deciding who to give the job to. On the RFQs most worth winning, that’s the difference that wins them.