Why No Instant Online Pricing

Custom manufacturing sites rarely show prices next to their work because a custom part has no standing price to show. A price for machined, cut, formed, or printed work is produced per part, from the geometry, the material, the tolerances, the quantity, and the process route a supplier would actually run. Two careful suppliers can reach different but honest numbers for the same file. Some online platforms do return automated prices. That model is real and legitimate, but it works by estimating inside a defined envelope, and the envelope is the story. This page explains both models, what each actually does, and why much of the industry still prices work the slower way. It offers no quote, and it promises no timeline for anyone.

Why a custom quote is a decision, not a lookup

A catalog product carries a price because it is identical to the last unit made, on a line that was paid for long ago, in a quantity planned months back. A custom part shares none of that. Every request arrives as a new geometry with its own material, its own tolerances, and its own quantity. Before any number exists, someone or something has to decide how this specific part will be made. That decision is the quote. The cost-driver anatomy behind it, setup, material, cycle time, tolerance and finish, quantity, and finishing, is unpacked in the companion guide on how manufacturing quotes are calculated. This page stays on the question of why a price cannot simply be looked up.

What has to be decided before a number exists

• Whether the geometry is makeable at all in the offered process, including deep pockets, thin walls, and internal corners that a drawing shows but a tool may not reach.
• How the part will be held: the fixturing approach, the number of setups, and whether a vise, a fixture, or a vacuum table changes the route and the risk.
• Whether the tolerance stack as drawn is achievable in that material, and how each critical dimension will be inspected.
• Whether the material is available in the size and quantity needed, because a hundred parts may fit stocked bar while ten thousand need a mill order with its own minimum and timeline.
• Which process route wins for this quantity, machining against forming against additive, and where that crossover sits for this part.
• What the finishing chain adds, deburring, passivation, anodizing, plating, assembly, and how each step touches cost, fit, and delivery.

Skip any one of those decisions and the number that survives is a guess wearing a price tag.

Why the same part prices differently at two shops

An analysis in the manufacturing trade press catalogued why identical drawings draw different numbers from different shops, and the reasons travel well beyond error. Equipment and route differ, so one shop reaches a part in a single five-axis setup that another covers in three operations on a simpler mill. Regional labor and overhead differ. So does current load: a busy shop bids higher than a quiet one for the same work, because capacity is what is actually being sold. Interpretation differs as well. An experienced estimator knows which callouts on the drawing matter and which can be held by default.

Skill plays its part too: a better fixturing idea genuinely lowers one bid. So do material relationships, because long-standing buyers get better stock pricing. Even the envelope matters, since shops are competitive inside the part sizes they are built around. And sometimes a low bid is simply a mistake. Load, metal prices, and interpretation all move week to week, so a price published on a page today would describe a moment, not the work.

What automated pricing tools actually do

The automated model deserves a fair description, because it is a real segment of the market and not a caricature. In the typical flow, a buyer uploads a CAD model and software reads the geometry: bounding box, volume, machined surfaces, holes, wall thickness. Rules or trained models then estimate machine time and material use, and rates are applied. A price comes back, often with automated manufacturability feedback that flags a thin wall or an unreachable corner before any money moves. Some engines compute from rules, cycle time multiplied by shop rates plus material. Others predict from historical orders, pricing a new part against thousands of similar ones already made. Both approaches estimate. Neither one plans the job the way a machinist, a programmer, or an estimator does, and that distinction drives everything that follows.

Where automation is strong

Give these tools the standard middle of the market and they perform a genuinely useful job: common alloys, 3-axis milling and turning, sheet cutting, the common printing processes, parts inside a tested size envelope, quantities in ranges the system has seen. The estimate rides on well-trodden ground, the number is usually a fair budget signal, and the attached analysis often improves the part. For a designer iterating on a concept, a price that appears while the file is still open is worth real money, because the design loop shrinks from days of waiting to the length of a coffee.

Where automation steps aside

Push outside the envelope and the behavior changes in predictable ways. Complex multi-axis geometry, several processes combined in one part, certified material with paperwork attached, assemblies, and unusual finishes tend to come back as a range rather than a number, as a note that an engineer will look at it, or as a polite decline. The word instant also means different things across the market. One tool returns a number in seconds, another after an automated analysis pass, another after a person has studied the file. A fast first number is sometimes revised once that person finds what the geometry did not say. On uncertain work, automated estimates carry padding for the uncertainty, which is rational pricing of risk rather than a flaw. Published research on estimating machining time from drawings reports meaningful error even for careful parametric methods, which sets a realistic ceiling for any tool built on the same idea. None of this makes the model bad. It makes the model a fit for a territory, with edges.

What an engineering-reviewed quote does differently

In the reviewed model, the request for quotation goes to a person who opens the model and the drawing and decides how the part will be made before committing to a number. Will it clamp safely, and where. Will a thin wall relax once the material around it comes off. Does this feature need a second operation or a different toolpath. Which of the stated tolerances actually drive cost, and which are decoration. Will an anodized layer of thickness push a tight fit out of tolerance, and which faces need masking. Will a process that produces one good part hold across five hundred.

The number that eventually arrives is attached to a specific plan for making the part. A good response states its assumptions: which material was presumed, which tolerance class was applied, how many setups were counted. The buyer can then challenge a line instead of a total. The trade is time. Review costs queue time and human attention, and on simple standard work that attention buys very little. On edge cases it is the difference between a price and a plan.

The questions geometry alone cannot answer

• How the part will be held, and what moves when material is removed.
• Which tolerances are functionally critical and which merely look careful.
• What the datum scheme implies about inspection method and cost.
• What a coating or plating does to mating dimensions after the cut.
• Whether a process that works once will repeat at the production quantity.
• What the drawing leaves unsaid, and whether the unsaid parts carry risk.

Every one of those is a judgment about intent, and intent lives in the application, not in the file. That is the honest core of the reviewed model: it buys interpretation.

The trade-off, stated plainly

Neither model is a trick, and neither is universally better. They price different risks, and a buyer who understands which risk is being priced can pick the right tool for the moment.

• Speed. Automated tools return a number with no wait, and the reviewed model spends human time and queue time to produce its plan.
• Standard work. On common parts, common processes, and common materials, both models tend to land in a similar band, which is exactly the territory automation was built to own.
• Unusual requirements. Tight or stacked tolerances, mixed processes, certified materials, assemblies, and odd stock are where automation pads or hands off, and where review earns its keep.
• Transparency of assumptions. An automated number embeds its assumptions in software, mostly out of sight. A reviewed quote can lay them out, which makes the number arguable in a way an opaque one is not.

The practical read is that the models converge on easy work and diverge on hard work, and the buyer’s job is knowing which kind of work is on the screen.

Why many job shops stay with the RFQ model

The deeper reason so much of the industry still prices by review is economic, not nostalgic. A job shop is a make-to-order or engineer-to-order business: nothing exists to hang a price on until a buyer sends a file, there is no catalog, no part number, no price list to publish. Each accepted job is inserted into a live schedule and displaces other work. The product being sold is a slice of capacity, and the price of that slice moves with the load on the floor. Material behaves the same way, as the stocked-versus-mill-order split above shows, so even the material line resists a flat published number.

Shops that carry audit staff, inspection equipment, and certifications spread those fixed costs across the work they quote, which is another input no published price could track. And quoting itself is unpaid labor. Estimators spend hours on requests the shop will never win, a real cost shops absorb because, for their kind of work, a number without a plan behind it is worse than no number. Programs like the NIST Manufacturing Extension Partnership exist largely for this high-variety, made-to-order world of small and mid-sized manufacturers, which is a fair signal of how structural the model is. Nothing about automated pricing breaks these economics. It simply prices a different, more standardized slice of the market, the slice described in the guide to low-volume manufacturing and the process-specific quote pages.

What makes any quote faster, whichever model

The levers that shorten the distance between a request and a dependable number sit mostly on the buyer’s side. They are identical for both models. These are inputs, not offers.

• Send a complete package: a 3D model with units stated, a 2D drawing with critical dimensions, tolerances, and a revision level, the exact material and temper, the finish, and the quantity. The detail behind each input is in the file preparation guide.
• Choose standard materials and stock sizes, because stocked bar and plate are sourced and priced differently from specials, in every model.
• Apply tight tolerances only where function demands them, since a blanket tight class raises the number under either model and invites scrutiny under review.
• Batch the questions: quantity options, finish alternates, and the timing context, sent once, so the request is answered once instead of entering a loop.
• State realistic quantities that match how material is sold and how setup amortizes, which lets the supplier or the tool pick the right process the first time.

Incomplete requests come back as questions rather than numbers. Much of the waiting that gets blamed on suppliers happens in that loop, not in the shop.

When each model is the right tool

When automated pricing fits

Simple geometry, a standard process, a common material, a quantity inside tested ranges, and a moment when speed is worth more than the last percent of accuracy. Early prototypes, brackets, housings under iteration, and budget signals for a design study all fit. There the automated number is the right tool, and waiting for a person would buy nothing the number lacks.

When reviewed quoting earns its keep

Tight or stacked tolerances, mixed processes in one part, certified materials with traceability, assemblies with mating features, production volumes where the process must repeat, and any drawing where what is not written matters as much as what is. There the plan is the product, and the slower number is the dependable one.

What this page does and does not do

This page describes two pricing models used across custom manufacturing. It names no company, assesses no supplier, compares no live prices, and offers nothing: no quote, no upload step, no timeline. For the vocabulary of quoting, see the manufacturing glossary. For how costs compare across processes, see the cost comparison by process page. For the process-level detail, there are the CNC machining quote and sheet metal quote guides.