A custom automation request often begins with a proposed machine: “We need a robot to load this fixture,” or “Can you add vision to this station?” Those ideas can be useful, but they leave the hardest question unanswered: what must happen when production is not ideal? Parts arrive skewed. A tray has an empty pocket. A test result is inconclusive. The downstream station stops. A well-defined cell has a response to each condition; an attractive demonstration may not.
For manufacturers and integrators, a strong request for quotation (RFQ) should describe the task boundary before the equipment architecture. That boundary includes normal input, acceptable output, variation, exceptions, and recovery. It helps bidders solve the same problem and gives the buyer a way to test the delivered system.
First, locate the constraint
Map the current movement of material and information, including waiting, rework, and operator decisions. NIST’s Manufacturing Extension Partnership (MEP) describes value stream mapping as a way to see those flows and identify waste and opportunities for improvement. Its automation guidance likewise starts with an assessment of operations, priority opportunities, a business case, and rigorous measurement of results. NIST: Value Stream Mapping; NIST: Robotics and Manufacturing Automation.
A simple, hypothetical example shows why this matters. A manual fixture-loading task takes 30 seconds. An automated concept reduces it to 12 seconds, but an upstream machine releases one part every 50 seconds and sets the line pace. Assuming no other constraints, the line still cannot produce more than about 72 parts per hour (3,600 ÷ 50). The project might still improve ergonomics, consistency, or operator availability. It should not be sold internally as an 18-second reduction in line takt. Measure the benefit at the line or labor-allocation level relevant to the business case, not only at the new cell.
Describe the input envelope, not the “typical” part
Give bidders a part-family matrix. List each SKU and its relevant dimensions, mass, tolerances, finish, flexibility, and allowed contact surfaces. Describe how parts arrive: tray, tote, conveyor, hand-loaded fixture, or bulk feed. Record orientation variation, presentation repeatability, and the condition of packaging after transport. If the task uses vision, include examples of color, reflectivity, markings, and contamination that the camera will actually see.
The goal is not to document every theoretical defect. It is to separate variation the cell must accept from conditions it may reject. For a fixture-loading station, a slightly rotated part might be normal input; a cracked part or two parts in one pocket might require rejection. If that distinction is absent from the RFQ, one bidder may include sensing and recovery while another assumes perfect presentation. Their prices will not represent comparable scopes.
Ask the people who run the current station to review this matrix. NIST MEP cautions that occasional process inconsistencies can complicate integration and recommends involving the employees who know the manual process. Their practical knowledge often reveals the exceptions missing from an engineering drawing. NIST MEP: How to Make Your First Robot Integration a Success.
Turn exceptions into decisions
An exception list is most useful when each row has four fields: detection, response, owner, and restart condition. Consider these illustrative entries:
| Condition | Required decision |
|---|---|
| Empty pocket | Skip and record it, or stop and request replenishment? |
| Two parts presented together | Reject to a defined location, or pause for operator removal? |
| Part cannot be located confidently | Retry within an agreed limit, then require review? |
| Downstream fixture unavailable | Hold the part safely, return it, or stop upstream feed? |
| Power interrupted mid-cycle | How is the part’s location established before restart? |
These are design questions, not claims that a particular Lineverra product provides those functions. The RFQ should assign who supplies the reject container, operator interface, upstream stop signal, traceability record, and recovery procedure. An exception without an owner becomes an integration gap.
Write acceptance criteria that include recovery
A target cycle time alone is a weak acceptance test. State the measurement boundary: from part available to accepted output, or from machine start to machine start? Specify the product mix, sample set, and whether changeovers and stops are included. Ask for both a typical cycle time and a distribution measure such as the 95th percentile if long-tail delays matter. If inspection is involved, define how false accepts and false rejects will be evaluated on known-good and known-bad samples.
Then test the exception table. At factory acceptance, run representative parts from the limits of the envelope, not just the easiest SKU. At site acceptance, include real upstream and downstream signals, operator access, changeover, jams, and restart after interruption. Record the time and steps required to recover. NIST notes that measurement capability must be considered before an experiment; the same logic applies here: agree on counters, time stamps, and quality evidence before using them to judge the cell. NIST Engineering Statistics Handbook: Is the Measurement System Capable?.
Safety must be part of the scope from the start. The cell layout should identify normal operator access, clearing points, and service access so the appropriate guarding and hazardous-energy controls can be designed for the actual installation. OSHA’s general machine-guarding and lockout/tagout requirements are relevant to these decisions; the applicable requirements must be assessed for the specific equipment and work. OSHA: 29 CFR 1910.212; OSHA: 29 CFR 1910.147.
A better basis for comparing bids
A concise RFQ can attach five items: the current process map; part-family and presentation matrix; exception list with owners; required interfaces and floor constraints; and a draft factory/site acceptance plan. Ask each bidder to identify assumptions and exclusions against those same documents. The resulting comparison is about how each proposal handles the real task, rather than whose demonstration looks more polished.
Focused automation succeeds when the boundary is narrow enough to engineer and broad enough to include ordinary production variation. Defining that boundary before selecting hardware is often the most valuable work the buyer can do.