Publish Time: 2026-11-18 Origin: Site
Interface planning guide · Published by I.C.T · For process engineers, project buyers and plant planners
Most PCBA factory projects fail at exactly the same place, and it is not the equipment. It is the seam between two parties who each believed the other one was holding a decision. The air drop gets installed at the wrong pressure because nobody owned the regulator. The MES interface stalls for three weeks because the customer's IT team and the machine vendor each assumed the other would supply the tag list. The container sits at the port because the delivery term said DAP and the site had no forklift rated for the crate. A PCBA factory project responsibility matrix exists to close those seams before they open, by naming one accountable party for every interface and writing down what evidence will prove the interface works.
The matrix is not a contract and it does not replace one. It is the working document that sits underneath the contract and answers a narrower question, interface by interface: who decides, who supplies, who approves, and who merely needs to know. This guide covers how to build one across six roles — equipment supplier, customer, utilities contractor, MES/IT, logistics and the local regulatory advisor — with worked examples, a template you can copy, and the arithmetic for deciding which interfaces deserve a formal gate rather than a phone call. The production line itself is only one row of the matrix, and it is worth reading the line configurations that sit inside one row of this matrix with that in mind: the machines are the easy part to specify and the hardest part to leave unattended. Every number below is a stated assumption, not a measurement from a named project.
A responsibility matrix answers three questions for every interface on the project: who makes the decision, who supplies what the decision depends on, and who signs that the result is acceptable. Those three questions sound administrative until you watch a project stall on them. In most projects the cost of an unclear interface is not the cost of the work — it is the cost of the days spent deciding who does the work, and those days land on the critical path in a way that no purchase order anticipates. A PCBA factory project responsibility matrix is therefore a schedule document as much as a management one. It is written early, it is written in one place, and it is written in language that a person who was not in the kickoff meeting can act on six months later.
The first failure is the missing owner: a decision that everybody assumes belongs to somebody else. Compressed air quality is the classic case, because it sits between the facilities contractor who installs the pipe and the equipment supplier who specifies the consumption, and neither of them is wrong to think the other one is buying the dryer. The second failure is the doubled owner: two parties who both believe they are accountable, act independently, and produce two incompatible results — two different network segments, two different ground references, two different answers on whether the line needs an uninterruptible supply. The third failure is the disempowered owner: a name in a cell who has the responsibility but not the budget or the authority to close the item, so every decision migrates upward and waits for a weekly meeting.
All three failures produce the same symptom at a different magnitude: latency. Assume, purely as an illustration, that a project has eighteen interfaces, that four of them are unassigned at the start, and that each unassigned interface costs an average of six working days of elapsed time before someone notices and resolves it. If those four resolve in parallel the project loses six days. If they chain — because the network drop blocks the MES test which blocks the acceptance test — the loss is closer to twenty-four days. The matrix does not make the work faster. It decides whether the delays land in parallel or in series, and that is where the calendar is won or lost.
An org chart is organised by who reports to whom, and a factory project is organised by who touches which system. Those two structures cut across each other. The maintenance manager may report to the plant director while owning the compressed air network, the spare parts store and the acceptance of the line's safety interlocks — three interfaces that involve three different external parties and three different evidence standards. A matrix built by copying the org chart inherits its blind spots, because the org chart has no column for "the party outside the company who also has to act".
The practical fix is to build the rows from the interfaces rather than from the people, then hang the names on the rows afterwards. Walking the product flow is the most reliable way to generate rows: bare board in, paste printed, components placed, reflow, inspection, through-hole or coating if the product needs it, test, pack out. Every transition in that flow is an interface, and so is every utility, every data path and every document that has to exist for the next step to start. That flow-first habit is the same one used in choosing the process route the whole project is built around, and it produces a row list that is complete enough to argue about, which is more than most first drafts manage.
It is not a contract. A contract allocates risk and money between legal entities; a PCBA factory project responsibility matrix allocates decisions between working teams. When a matrix cell and a contract clause disagree, the contract wins and the matrix is wrong, which is why the matrix should be reviewed against the commercial documents before it is issued rather than after the first dispute. It is also not a RACI chart in the consulting sense, where responsibility is often shared and accountability is nominal. In an equipment project a shared cell is a defect: if two parties are both marked as owning the air quality decision, the project has no owner for air quality, regardless of what the spreadsheet says.
Finally, the PCBA factory project responsibility matrix is not a one-time deliverable. It is a live register that changes when scope changes, and its value depends on somebody being allowed to edit it without a change-control meeting. A matrix that is frozen at kickoff and never touched again will be accurate about the project as it was imagined and silent about the project as it is being built. The version that survives to handover is the one that has a named editor, a change log, and a rule that no interface is closed without a dated line in that log.
Six roles cover almost every interface on a PCBA project, and the reason to fix the list at six is that a longer list stops being read. The equipment supplier, the customer's own operations and engineering team, the utilities or facilities contractor, the MES and IT function, the logistics provider, and the local regulatory advisor. Each of them controls something the others cannot, and each of them has a characteristic failure mode when left unassigned. The four columns then force the conversation that most projects avoid: for this interface, which of these six makes the call, which one hands over the input, which one signs, and which ones only need to be told. One letter per cell, one vote per row.
The distinction that matters is between what a role controls and what it can only influence. The equipment supplier controls the machine's internal wiring, its own documentation set, its factory acceptance test results and the interfaces it exposes. It does not control the building, the incoming power quality, the network policy or the customs broker, and a supplier who agrees to "take care of" those things is usually agreeing to coordinate them, not to own them. The customer controls access, schedule, budget approval and the acceptance signature. The utilities contractor controls the physical installation of air, power, exhaust and water, but not the design values those installations are sized against — that is the interface where most of the argument lives.
MES and IT control the data model, the network, user accounts and the retention policy, and they are almost always engaged later than they should be. The logistics provider controls movement, customs documentation and delivery sequencing, but not the unloading equipment at the destination, which is why the last hundred meters is a permanent hazard. The local regulatory advisor controls nothing physical at all; what it controls is your ability to predict what an inspector will require, and its value is measured in permits issued on time rather than in drawings produced. A role that controls nothing physical is still essential, because the alternative is discovering the requirement during the inspection.
Role | Controls directly | Cannot control | Typical failure if unassigned |
|---|---|---|---|
Equipment supplier | Machine internals, documentation set, factory test results, exposed interfaces | Building services, incoming power quality, network policy, customs | Utilities assumed to be "included" and never installed |
Customer operations and engineering | Access, schedule, budget approval, final acceptance signature | Vendor delivery dates, contractor labour, permit timing | Acceptance criteria invented at the end of the project |
Utilities / facilities contractor | Physical installation of air, power, exhaust, water, flooring | Design values the installation is sized against | Correct pipe, wrong pressure; correct cable, wrong earth |
MES / IT | Data model, network, accounts, retention, firewall policy | Machine-side protocol support, production priorities | Connectivity tested two weeks before handover |
Logistics provider | Movement, customs paperwork, delivery sequencing | Unloading equipment, site access rules, storage at destination | Container arrives before the forklift exists |
Local regulatory advisor | Predicting inspection requirements, permit applications, local documentation | Anything physical, and the authority's actual timetable | Permit discovered late, commissioning waits on paperwork |
Four columns are enough: owns, supplies, approves, informed. "Owns" means one party makes the call and is accountable for the outcome. "Supplies" means a party must hand over an input — a drawing, a flow rate, a tag list, a certificate — without which the owner cannot decide. "Approves" means a party can say no, and its silence is not consent. "Informed" means the party needs the outcome but has no veto. The discipline is that every row has exactly one mark in the owns column, at least one in supplies when the decision depends on an external input, and a named approver only where a real veto exists.
Two rules keep the grid honest. First, a cell can carry one letter, not two: the moment a cell reads "owns/supplies" the row has stopped being a decision and become a description. Second, no row may be completed by the same party in both owns and approves, because that is not approval, it is a notification. In practice the second rule is what forces the customer to stay engaged: if the customer owns a decision and the supplier approves it, nothing has been checked, and the interface will be signed with whatever the supplier proposed.
A cell that says "contractor owns air" is worth nothing at handover, because it does not say what the contractor is on the hook for. A testable cell says who decides, what value they must deliver, how it will be verified, and by when: the contractor owns the air supply, must deliver 6 bar at the line's inlet manifold with a dew point suitable for the specified instruments, verification is a flow and pressure test at the inlet with the line running its heaviest simultaneous demand, and the test is witnessed by the customer's maintenance lead before the acceptance meeting. That sentence is longer, and it is the difference between an interface that closes and one that reopens in month three.
The test of a well-written cell is whether a person who has never met the project could read it and know when it is done. If the answer depends on a conversation, the cell is under-specified. This is also where the matrix starts to overlap the commercial documents: the acceptance evidence column should match the acceptance criteria in the purchase order, and if it does not, one of the two documents is about to be renegotiated. Checking that alignment before signature is cheap; checking it after the line is installed is not.
Take a project with an SMT block and a through-hole block, delivered as separate shipments, installed in a leased hall with an existing compressed air ring main and no existing MES. That gives roughly a dozen interfaces worth writing down: incoming power, air supply, exhaust, floor loading, network drop, machine-to-machine conveyor handshake, MES data feed, barcode or laser marking, spare parts storage, crate removal, customs clearance, and permit to operate. Marking one owner per row takes about two hours with the right six people in the room, and produces a document that will be argued with — which is the point.
The board hand-off between the two blocks deserves its own row rather than being folded into "line integration", because the SMT block and the through-hole block are usually shipped and commissioned by different crews on different weeks. That seam is the one most often left to the last team on site, and it is worth looking at a DIP line where the SMT-to-through-hole hand-off is the structural seam to see how many decisions it actually contains. The same applies to the commercial side: the matrix and the input package a turnkey proposal gets priced against should describe the same scope, or the quotation will be answering a different question than the one the project is asking.
Interface | Owns | Supplies | Approves | Informed |
|---|---|---|---|---|
Incoming power capacity and earth | Customer | Utilities contractor, equipment supplier | Customer engineering | MES/IT |
Compressed air at line inlet | Utilities contractor | Equipment supplier | Customer maintenance | Supplier |
Reflow and wave exhaust | Utilities contractor | Equipment supplier | Local advisor | Customer |
Floor point loading and access route | Customer | Logistics, equipment supplier | Building owner | Contractor |
Network drop and IP plan | MES/IT | Equipment supplier | Customer IT security | Operations |
Conveyor handshake between blocks | Equipment supplier | Customer | Customer engineering | MES/IT |
MES data feed and tag list | MES/IT | Equipment supplier | Customer operations | Supplier |
Marking and traceability content | Customer | MES/IT | Customer quality | Supplier |
Spare parts storage and access | Customer | Equipment supplier | Customer maintenance | Logistics |
Crate removal and disposal | Logistics | Customer | Customer facilities | Contractor |
Customs clearance and duty | Logistics | Equipment supplier | Customer finance | Local advisor |
Permit to operate | Local advisor | Customer, contractor | Authority | All parties |
A simple health score makes the grid reviewable without reading every cell. Count the rows, count how many have exactly one owner and at least one approver, and divide. In the table above, twelve rows with a single owner each and a named approver on every row scores twelve out of twelve. A realistic first draft usually scores eight or nine, because three or four rows have either no owner or two. Assume, as an illustration, that each defective row costs five working days of elapsed delay and that the project can absorb two such delays in parallel but not four: fixing the score from nine to twelve is the difference between a four-day slip and a twelve-day slip, for two hours of meeting time.
Utilities are where a PCBA factory project responsibility matrix earns its keep, because they are the interfaces that no single party considers its own. The equipment supplier knows what the machine consumes but does not install the supply. The contractor installs the supply but did not choose the values. The customer pays for both and owns neither, in the sense that no individual in the customer's organisation wakes up responsible for the dew point of the air. Add data and site access and the pattern repeats: the interfaces that cross a boundary between organisations are the ones that default to nobody. This section handles them in the order they usually cause trouble, with the values that need to be written into the matrix rather than agreed verbally on site.
Compressed air splits cleanly into three decisions that are often made by three different parties. The first is demand: what flow the line needs, at what pressure, at what dew point and at what oil class. The equipment supplier owns that number because it comes from the machine specification. The second is supply: compressor capacity, receiver volume, dryer type, filtration and the ring main. The contractor owns that, but only if it is given the demand figure and a margin. The third is the last meter — the drop, the regulator, the filter and the quick-release coupling at the line inlet. That is where projects fail, because it is small enough that both parties assume it is included.
The matrix should therefore carry two rows, not one: "air supply to the hall" and "air at the line inlet", with different owners and different acceptance evidence. As an illustration, assume a line whose air-operated devices total 900 litres per minute at a 40 percent duty cycle, which averages to about 360 litres per minute, and add a 30 percent allowance for leakage and future nozzles, giving roughly 470 litres per minute as the design figure. A receiver sized for the instantaneous draw rather than the average is what keeps the pressure stable during a simultaneous actuation; sizing only the compressor to the average is the classic mistake. These numbers are illustrative only — the real figures must come from the machine specification sheets.
Utility interface | Design value to write down (example, not a specification) | Owner | Acceptance evidence |
|---|---|---|---|
Compressed air at line inlet | 6 bar dynamic at inlet, dew point suitable for the specified instruments, oil class stated | Utilities contractor | Pressure and flow logged with the line at peak simultaneous demand |
Incoming power | Three-phase supply within plus or minus 10 percent of nominal, single-phase within plus or minus 10 percent, separate protective earth | Customer | Voltage log over a full shift, earth resistance measurement, breaker coordination sheet |
Reflow and wave exhaust | Extract volume stated per oven, make-up air balanced so the hall stays at slight positive pressure | Utilities contractor | Hand-held anemometer readings at each hood, room pressure differential record |
Floor loading | Point load per machine foot and crate weight per square meter during installation | Customer | Structural note from the building owner, crate weight list from the supplier |
Network drop | One drop per machine cell plus one for the line controller, on a defined VLAN | MES/IT | Link test from each machine to the MES endpoint, firewall rule reference |
Site access | Door width and height clear of crate dimensions, forklift capacity above crate weight | Customer | Route survey with dimensions, lifting plan signed before delivery |
A machine that is connected to a supply is not the same as a machine that has a supply it can run on, and the difference is entirely a matter of who measured what. Voltage within tolerance at the panel is not the same as voltage within tolerance at the machine terminals under load, and a neutral that is shared with other equipment can put the line's control electronics on a different reference from the one the supplier tested against. The matrix row should name the party who measures, the instrument, the duration and the point of measurement, because "we have 380 volts" is a statement about a multimeter reading, not about a supply.
The earth arrangement is the second half of the same row and is frequently skipped. Whether the building provides a separate protective earth, whether the neutral is bonded at the transformer, and whether the line's own earth bar is connected to the building's by a conductor sized for the fault current are three questions that need one owner and one piece of evidence. In many jurisdictions the local wiring regulations will also require the installation to be inspected before energisation, which turns this from an electrical question into a permit question, and that permit is on a different clock from the equipment delivery.
Reflow ovens and wave solder machines extract process air, and every cubic meter extracted has to be replaced by make-up air from somewhere. If the replacement comes from outside untreated, the hall loses temperature and humidity control and the paste printing process starts to drift. If it comes from adjacent spaces, the hall goes negative and pulls contamination in through every gap. The interface therefore has two owners and one shared number, which is exactly the shape the matrix is designed to expose: the contractor owns the extract and the make-up provision, the customer owns the setpoint decision on room pressure and temperature, and the equipment supplier owns the extract volume per oven.
As an assumption for planning only, a two-oven line extracting 500 cubic feet per minute per oven needs 1,000 cubic feet per minute of make-up air, which is roughly 28 cubic meters per minute. Supplying that unconditioned in a humid climate will overwhelm a small precision air conditioner, so the make-up air usually needs tempering, and tempering needs capacity that nobody put in the budget. The point of the matrix is not to solve the engineering — it is to make sure the person who discovers this is the person who owns it, on a date when there is still time to order the unit.
The network row is the one most often written as "IT will handle it", which is not a decision, because IT has no way to know that a placement machine needs a fixed address on the same segment as the line controller, or that the MES collector polls every machine on a schedule that cannot tolerate a proxy. The row needs an owner in IT, an input from the equipment supplier listing every address the line consumes and every port it opens, and an approver in security who can accept or reject the rule set rather than silently rewriting it three weeks later.
Write the row before the purchase order, not after, because the equipment supplier's connectivity documentation is part of what you are buying. The right moment to demand it is the same moment you define the scope of supply that the RFQ will be measured against, since an interface list that arrives after the order is a change request, and one that arrives before it is a specification. The evidence column should say "link test from each machine to the MES endpoint, with the firewall rule set attached", because that is a result somebody can verify rather than an intention somebody can restate.
The data layer is where the matrix is most often incomplete at the start and most expensive to complete at the end, because the parties involved do not share a vocabulary. Operations talks about traceability, IT talks about interfaces and schemas, the equipment supplier talks about what its machine can export, and the customer's quality team talks about what an auditor will ask to see. All four are describing the same requirement from different ends. The matrix has to translate, and the translation is a short list of concrete deliverables: a data model, a tag list, a protocol, a retention period and a named person who accepts each one. Where that list is missing, the project buys equipment with an undefined data obligation, and the obligation is only discovered during commissioning, when the last remaining lever is manual work and a request for more headcount.
The customer's MES or IT function owns the data model, because the model has to serve reporting, traceability and audit requirements that only the customer knows. What the customer does not own is the machine's ability to produce a field, and this is where projects lose weeks: the tag list is issued, the machine cannot supply three of the fields, and the discussion about substitutes starts after the equipment is already on the water. The supplier owns a document that says exactly what the machine publishes, in what format, with what update rate and what identifier for the board — and that document should be a deliverable in the purchase order, not a favour after installation.
One rule makes this workable: the data model is defined against the fields that exist, not against the fields that would be nice. If a machine cannot report a nozzle identifier per placement, the traceability design must not depend on it. Discovering that dependency after the acceptance test is the most expensive possible moment, because the fallback is usually manual scanning, which changes labour content and sometimes the headcount the business case was built on.
Three layers need to be separated in the matrix. The first is machine-to-machine: the conveyor handshake that lets one machine release a board to the next. The SMEMA interface standard exists precisely to reduce that connection to a small set of discrete signals and a defined mechanical interface, and where both machines comply, that row is nearly free. The second is machine-to-MES: the protocol and the payload, which may be a standard such as IPC-CFX, a vendor-specific interface, or a plain log file collected by an agent. The third is line-to-business: the reporting layer, which is the customer's problem and should be marked as such.
Placement performance data sits in the second layer and deserves a caution. A machine's declared placement rate is measured under a defined set of conditions, and the standards that define those conditions — IPC-9850 is the usual reference for placement equipment — describe the test, not the production line. A matrix row that says "supplier provides CPH data" without saying "under the conditions of the declared test" will produce a number that nobody can defend in a review, and the argument will consume more time than the data was worth.
Retention is a decision with a cost attached and it belongs to the customer, not to IT and not to the supplier. How long serial-level records must be kept is driven by the customer's contracts, the product's safety classification and the local record-keeping rules, and each of those can override the others. The matrix row should state the retention period in months, the storage location, the export format and the person who can prove the export works — because an audit does not ask whether records exist, it asks for a specific board's history in a specific format.
The distinction between a new build and a retrofit also changes who owns the data rows, and it is worth checking how planning changes when the project runs beside an existing line before assigning them. In a brownfield project the retention question often answers itself, because an existing system already holds the records and the new line has to fit its schema. In a greenfield project the schema is being designed at the same time as the line is being specified, which means the matrix has two moving parts instead of one and needs a freeze date that both sides accept.
Logistics is the part of the matrix most likely to be written by the commercial team and then forgotten by everyone else, because a delivery term looks like it settles the question. It does not. A delivery term allocates cost and risk for transport between two named points; it says nothing about whether the destination has a forklift, whether the gate is wide enough for the crate, or who pays for the container while it waits. Those are three separate rows with three separate owners, and each of them can stop the project while the equipment sits in a yard fully paid for and completely useless.
Under most delivered-duty arrangements the seller's obligation ends when the goods are made available at the named place, and the act of unloading is the buyer's. That single sentence generates a predictable set of failures: a container arrives, nobody has a machine capable of lifting the crate, the driver waits, and the waiting is charged. The matrix row should name the party who provides the lifting equipment, its rated capacity against the heaviest crate, the person who confirms the route from gate to final position, and the fallback if the site cannot take a delivery on the day it arrives.
As an illustration only, assume two 40-foot containers, a demurrage and detention exposure of roughly USD 120 per container per day after the free period, and a five-day delay caused by a missing forklift and a locked gate. That is about USD 1,200 of direct charge, which is trivial next to the real cost: five days of an installation crew standing by, which at an assumed USD 2,400 per day for a four-person team is USD 12,000. A matrix row costing one phone call is protecting a five-figure exposure, and it is the kind of row that gets skipped because it looks like purchasing administration.
Physical access is measurable, which makes it the easiest interface to close properly and the most embarrassing to get wrong. Four numbers decide it: the crate's outside dimensions, the heaviest single crate weight, the narrowest door or corridor on the route, and the load capacity of the floor along that route. All four should be exchanged before the shipping date, because the supplier can usually modify a crate and the building cannot usually modify a door. When those numbers are exchanged late, the solution is a crane through a roof opening, which is a different project with a different budget.
Floor loading needs its own row because the number people quote is usually the wrong one. A machine's static weight spread over its footprint is not the same as the point load at a levelling foot, and neither is the same as the dynamic load during unpacking when a crate is set down on a pallet truck. Assume a machine of 3,000 kilograms on four feet: the average load is 750 kilograms per foot, but if the weight is unevenly distributed one foot can carry 1,400 kilograms, and the floor has to be checked against the higher number. Every one of those figures is an assumption for illustration, and the actual values must come from the machine's installation drawing.
Customs clearance is an interface with a clock that runs independently of everything else, and the matrix needs an owner for the paperwork before the equipment ships rather than after it arrives. The documents that decide the timeline are usually the commercial invoice and packing list, the certificate of origin, the product documentation required by the destination authority, and any conformity marking the destination requires. Which of those apply is a question for the local advisor, not for the supplier and not for the freight forwarder, because the forwarder will clear whatever documents it is given and will not tell you that a certificate is missing until the shipment is held.
There is a second, quieter question that belongs in the same row: what the destination considers local content or locally supplied scope. Some projects carry conditions about local subcontracting, local testing or local documentation that only become visible when the permit application is filed. The matrix should carry a row called "local content conditions" with the advisor as owner and the customer as approver, even if the answer turns out to be "not applicable" — a row that says not applicable is a decision, and a missing row is a discovery waiting to happen.
The local regulatory advisor is the role most often left out of the matrix and most often blamed when the schedule slips, which is unfair because the failures attributed to it are almost always failures to ask early enough. An advisor cannot make a permit arrive faster than the authority allows. What an advisor can do is tell you, three months before you need it, that the authority will require a specific test, a specific drawing or a specific local signature — and that single piece of information is usually worth more than every other line in the document, because it is the only one that cannot be recovered by spending money late. That asymmetry — cheap to ask early, impossible to buy late — is the reason the advisor belongs in the matrix at kickoff rather than at the permit stage.
The advisor's job is prediction, not representation. A good one will read the project's equipment list and layout and tell you which permits apply, which inspections will be scheduled, which documents will be requested and roughly how long each stage takes in that jurisdiction. A less useful one will tell you that everything is fine until it is not. The matrix row should therefore specify the deliverable in terms the customer can check: a permit list with lead times, a document checklist mapped to each permit, and a named contact for each authority — delivered before the purchase order, not after.
The interface with the equipment supplier matters here in a specific way. Most permits require documentation that only the manufacturer can produce: electrical schematics, a declaration of conformity, a noise or emission figure, sometimes a local-language manual. Those documents are cheap to produce at the factory and expensive to reconstruct from a photograph of a nameplate. The matrix should list them as deliverables with a due date tied to the permit application rather than to the shipment, because a manual that arrives with the machine is a manual that arrives after the permit window has already closed.
Permits are the clearest example of an interface where the owner has no ability to accelerate the other party. The authority's timetable is fixed, the inspection has to be booked, and the booking usually depends on a document that depends on a test that depends on the line being energised. That chain is worth drawing as a sequence in the matrix rather than as a list, because the dependency is what causes the slip, not any individual step. The typical sequence runs from building or alteration consent, through electrical inspection and fire or environmental review where applicable, to the operating permit that lets the line run for sale.
Line size changes the list in a way that is easy to overlook. A semi-automatic line with manual loading, no in-line inspection and a single oven carries fewer utility interfaces than a fully automatic line, and a shorter utility list usually means fewer inspections — an example of that smaller package is a semi-automatic SMT line with a reduced interface count, and the reduction is a permitting advantage as much as a cost one. The matrix should record which permits were assessed and dismissed, with the reason, because the authority will ask why a given permit was not applied for and the answer needs to be in writing.
Permit or inspection stage | Trigger | Owner | Lead time assumption |
|---|---|---|---|
Building or alteration consent | Layout changes, new openings, floor penetrations | Customer | 4 to 12 weeks, jurisdiction dependent |
Electrical inspection before energisation | New supply, new distribution board, earth arrangement | Utilities contractor | 2 to 6 weeks after installation complete |
Fire and life safety review | Solvent storage, coating room, additional occupancy load | Local advisor | 3 to 8 weeks |
Environmental or emission review | Reflow and wave exhaust, coating solvents, waste streams | Local advisor | 4 to 10 weeks |
Operating permit | Line ready to run for production | Customer | 2 to 6 weeks after final inspection |
The table is a planning shape, not a rule book: every lead time in it is an assumption and every jurisdiction differs. What transfers between jurisdictions is the structure — the fact that the electrical inspection gates the energisation, that the energisation gates the line test, and that the line test gates the operating permit. Once the chain is written down, the matrix has a critical path, and the critical path is what tells you which of the six roles you actually need in the next meeting.
The document set is where the matrix stops being a management tool and becomes a checklist that either exists or does not. Typical items include the equipment list with model and serial numbers, layout drawings with dimensions and clearances, electrical single-line diagrams, the earth and bonding arrangement, exhaust and make-up air calculations, noise data, the operating manual in the local language where required, and evidence of the operator training that the authority may expect before production starts. Each item has an author, and the author is rarely the same party twice.
Assigning authors is the whole exercise. The electrical diagrams come from the supplier; the single-line diagram for the installation comes from the contractor; the layout usually comes from the customer with input from both; the local-language manual may need translation done by the customer; the training record comes from whoever ran the training. Writing six names against six documents in one table takes twenty minutes, and it is the cheapest part of the entire permitting process.
A matrix without an escalation rule is a description of intent, because every unresolved cell eventually becomes an argument and arguments need a time limit to stay cheap. The rule does not need to be sophisticated. It needs to say who decides at which level, how long each level has before the item moves up by default, and what happens to the schedule while the item waits. Projects that write this down before kickoff resolve interface disputes in days; projects that leave it to the weekly meeting resolve them in weeks, and the difference compounds across eighteen rows with a dependency chain behind each one. The rule also protects the people in the middle, because an engineer who knows that a question moves up automatically after two days does not have to choose between chasing a decision and doing the work.
Three tiers are enough for a project of this size. Tier one is the working level: the engineer or coordinator on each side, with a two-working-day limit and authority to settle anything that does not change scope, cost or acceptance criteria. Tier two is the project manager level, with a five-working-day limit and authority to accept a change within an agreed contingency. Tier three is the sponsor or director level, unlimited in authority but expensive in time, and it should be reached only for scope, contract or payment questions — because the moment a technical question reaches tier three, it stops being a technical question.
The important detail is the default. Every tier should have a rule for what happens when the time limit expires: the owner's proposal stands, or the item is formally recorded as a risk against the schedule, or both. A ladder without a default is a suggestion, and suggestions are how a two-day decision becomes a three-week silence. Recording the lapse as a risk is usually the better default, because it puts the delay into the project report where somebody with authority will see it, rather than leaving it in a mailbox.
Tier | Who decides | Authority | Time limit | Default if the limit expires |
|---|---|---|---|---|
1 — working level | Engineer or coordinator from each party | Any interface question with no scope, cost or acceptance impact | 2 working days | Owner's proposal stands, logged as a decision |
2 — project management | Project manager from customer and supplier | Changes within the agreed contingency, schedule shifts | 5 working days | Item recorded as a dated schedule risk |
3 — sponsor | Director or contract owner from each side | Scope, price, contract terms, termination of a dependency | As scheduled, with minutes | Written notice required; no silent default |
Every project accumulates changes that nobody remembers approving, and the matrix is where they become visible. A log with five fields is sufficient: date, interface row affected, what changed, who approved it, and what the consequence is for cost, schedule or acceptance. The discipline that matters is that no cell changes owner without a log entry — because the most expensive change on a factory project is not a design change, it is a quiet transfer of responsibility from the party that was paid for it to the party that was not.
The log also protects the supplier relationship, which is worth saying plainly. When an interface transfers mid-project, both parties usually believe the transfer was agreed, and both are usually right in their own recollection of a corridor conversation. A dated line in a shared document converts that into a fact that can be checked without anyone losing face, and it is the reason the log should be visible to all six roles rather than held by the project manager as a private record.
Handover is not the day the line runs. It is the day the last row of the matrix has evidence attached and a signature against it, and those two events are usually a week or two apart. The evidence is what was written in the acceptance column six months earlier: the pressure log taken at peak demand, the voltage record over a full shift, the link test with the firewall rule attached, the training record with names. A handover meeting that reviews signatures without reviewing evidence is a meeting that will be repeated in three months with the same agenda.
The number of rows to close at handover scales with the size of the package being installed, which is one practical argument for choosing a smaller package when the site team is thin. A fast-changeover line with fewer stations and fewer joints — of the kind described in a flexible semi-automatic SMD line built for frequent product changes — carries a shorter interface list and therefore a shorter close-out, and that difference should be visible in the matrix before the choice is made rather than after. Whatever the package, the close-out list should be issued at the same time as the installation schedule, so both sides know from the first week what the final week will require.
Bring the matrix, not just the equipment list
If you are planning an SMT or through-hole line and the interface list is still in six people's heads, send us the layout, the utility constraints and the roles you already have named. We will map the interfaces we normally own against the ones your contractor and IT team will own, and flag the rows that typically go unassigned on projects of this size. It costs nothing and it usually shortens the kickoff meeting by a full day, because the argument about who owns the compressed air happens once, on paper, before anyone books a flight. You do not need a finished document to start — a list of what is already named, plus the two or three interfaces you suspect nobody owns, is enough. What comes back is a marked-up version of your own list rather than a generic template, so it can go straight into the kickoff pack and be argued with by the people who have to live with it.
The customer's project manager should own it, because the customer is the only party with a contractual relationship to all six roles. The supplier can draft the interface rows — it usually knows the equipment side best — but it cannot assign responsibilities to the utilities contractor, the IT function or the local advisor, and a matrix issued by the supplier will be read as the supplier's view of the project rather than the project's own record. The practical arrangement is that the supplier drafts, the customer owns, and both sign each revision.
Most single-line projects settle between fourteen and twenty rows once utilities, data and documentation are counted honestly. A dozen covers the physical and commercial essentials: power, air, exhaust, floor loading, access, network, conveyor handshake, MES feed, marking, spares, crate removal and permits. The rows that push the count higher are usually the documentation rows — the drawings, certificates and manuals that a permit or an acceptance test depends on. If a draft matrix has fewer than twelve rows for a project with through-hole or coating in scope, something is missing.
Usually not, and it is better to say so explicitly than to leave the impression that it is included. The supplier is normally responsible for the machine's internal wiring, the specification of what it consumes, and the interface point where the customer's services connect. The physical installation of air, power and exhaust up to that point belongs to the contractor or the customer. What the supplier should always own is the document that states the consumption figures, because everything the contractor sizes depends on it, and that document should be a named deliverable with a due date.
Before, at least in draft. The matrix is one of the cheapest ways to discover that the scope you are about to buy does not include something you assumed, and that discovery is worth far more before signature than after it. A draft issued with the RFQ also improves the quotations, because suppliers price against a defined interface boundary instead of guessing where their responsibility stops. The matrix is then revised at kickoff, when the contractor and IT names are known, and frozen at the point where the installation schedule is agreed.
Split it into two rows rather than sharing one. A shared cell is not a compromise, it is an unmade decision, and it will surface during installation as two incompatible results. The usual split separates design from installation, or supply from acceptance: one party owns the design value and another owns the physical delivery, each with its own evidence and its own signature. Where the split is not obvious, the test is to ask who would be called first if the interface failed on a Saturday night — that party owns the row, and the other party supplies or approves it.
Something a third party could re-check without asking a question. For utilities that means a logged measurement taken under load, not a nominal value. For data interfaces it means a successful link test with the firewall rule reference attached. For documentation it means the document itself, in the required language, with a revision number. For training it means a record with names and dates. The rule that keeps this honest is that the evidence is decided when the row is written, not when the row is closed, because evidence chosen at close-out is evidence chosen to be easy.
No, and using it as a substitute causes trouble in both directions. A scope of supply document defines what is being bought and sold, in commercial language, with prices and delivery terms. A RACI chart describes roles inside one organisation. The matrix sits between them: it takes the commercial boundary and converts it into operational decisions across parties who do not share a reporting line. Where the matrix and the contract disagree, the contract governs, which is why the matrix should be reviewed against the commercial documents before it is issued rather than after the first dispute.