Monday, ten in the morning. A QA engineer from an external vendor receives three messages within fifteen minutes. The head of the quality department asks him to move all test cases to the department’s new template by Friday, because an audit is coming up next week. The project manager for the warehouse system rollout writes that the regression suite has to be ready by Wednesday, because the release date has moved. The vendor’s delivery manager reminds him that from Thursday he is taking over part of a colleague’s work on another project at the same client. Each of them is right from where they sit. None of them knows about the other two requests.
On Wednesday the regression suite is not ready. The project manager escalates to the vendor: your engineer is not delivering. The vendor replies that the engineer did what the client asked. The head of quality is surprised, since her request seemed so small. In the follow-up meeting it turns out that nobody can say who actually decides what this engineer works on in a given week.
This article is for COOs, PMO leads and anyone running a portfolio of projects in a matrix organisation that relies on external specialists. The thesis is this: a matrix that employees can live with falls apart once a third dimension, the vendor, is added. An external specialist must not end up with a third boss. Whether the company runs internally as a weak, balanced or strong matrix, towards the vendor’s people it should behave like a strong matrix: one point that assigns work on the project side, standards from the departments, personnel matters with the vendor. A capable functional head may well disagree, because this takes away direct influence over people working in their area. I will argue that losing that influence costs less than the alternative.
What is a matrix organisation and why does it exist?
A matrix organisation is a structure in which people report along two lines at once. The first is functional: engineering, quality, finance, operations. It owns skills, standards, people development and resource allocation. The second is the project or product line: a project manager or product owner who owns the goal, scope, deadline and budget of a specific piece of work. Each person therefore has two bosses, and the organisation accepts the tension between them on purpose.
The matrix is not a recent fashion. Jay Galbraith described it in 1971 in “Matrix Organization Designs: How to combine functional and project forms” in Business Horizons, as a way of joining two orders: the functional one, which is good at building expertise, and the project one, which is good at delivering results. A purely functional company has excellent specialists, but projects that cut across departments stall at the boundaries. A purely project-based company delivers fast, but every project rebuilds expertise from scratch and scatters knowledge. The matrix tries to have both.
The price of that compromise has been known for a long time. Stanley Davis and Paul Lawrence, in “Problems of matrix organizations” in Harvard Business Review in 1978, described the pathologies a matrix develops when there are no rules for settling disputes. Christopher Bartlett and Sumantra Ghoshal, in the same journal in 1990, titled their piece “Matrix Management: Not a Structure, a Frame of Mind” and argued that the problem rarely lies in the org chart and more often in an organisation’s ability to operate according to it. Both points matter here. A matrix works when the people in it know who decides what and have the habit of resolving conflicts quickly. An external vendor has neither, because it is not part of that culture.
It is also worth noting what a matrix does not say. It does not say that two bosses have equal rights over every decision. Every working matrix has a split: the department decides how work is done well, the project decides what has to be done and when. Trouble starts where that split is implicit and lives only in the heads of people who have been at the company for years.
How do weak, balanced and strong matrices differ?
The distinction between weak, balanced and strong matrices is popularised by PMI’s PMBOK Guide. The difference is not in the drawing but in who holds real authority over people and money. This matters, because it determines who will in practice assign work to an external specialist if nobody decides otherwise.
| Characteristic | Weak matrix | Balanced matrix | Strong matrix |
|---|---|---|---|
| Who prevails in decisions about people | Functional manager | Shared, negotiated | Project manager |
| Project manager’s role | Coordinator, often part-time | Manager with limited authority | Full-time manager with real authority |
| Who controls the project budget | Mostly the department | Jointly, by agreement | Mostly the project |
| Typical conflict | The project cannot enforce its priorities | Every decision needs negotiation | Departments lose influence over standards |
| Who assigns work to an external specialist by default | Functional manager | Whoever writes first | Project manager |
| Risk for the vendor relationship | Specialist works “for the department”, not the project | Conflicting instructions from both sides | Lowest, provided departments set standards |
The last two rows are not from PMBOK; they are this article’s interpretation. They show something easy to miss. In a weak matrix, an external specialist formally hired “for the project” effectively becomes a departmental resource: the functional manager holds the power, so they hand out tasks and the project manager has to ask for time. In a balanced matrix, the specialist takes instructions from both sides and has to choose whose matter more. That is the worst place an outsider can be, because they do not know the company’s informal hierarchy and have no standing to resolve it.
There is no single right matrix for a whole company. Departments with long traditions, strong standards and few projects naturally lean towards a weak matrix. Companies that run most of their work as projects and products lean towards a strong one. This article does not argue which matrix your company should have for its employees. It argues how the company should behave towards the vendor’s people, and there the answer is the same in every case.
What changes when a vendor joins the matrix?
An employee in a matrix has two bosses, but both work for the same company. They share a senior manager, a priority calendar, tools and history. When the functional head and the project manager disagree, the dispute has somewhere to go: a director, a steering committee, the PMO. The employee can say “please sort it out between you” and risks nothing more than brief awkwardness.
An external specialist has three “bosses”, one of whom sits in a different company. The client’s department sets standards and would often like to control their time. The client’s project sets tasks and deadlines. The vendor is responsible for their availability, appraisal, development, leave, cover and replacement. The three parties have different interests and different communication channels, and a dispute between them has no natural place to be settled. The specialist will not tell the client’s head of quality to “check with the project manager”, because they worry about being seen as difficult, and the client’s opinion decides whether they stay on the engagement.
The result is predictable. The specialist does what the nearest or loudest person asks. A priority conflict that a director would settle inside the company is settled by an outsider with the least information about what matters to the business. Later, when something is not delivered, accountability lands on the vendor, even though the vendor assigned none of the tasks.
There is a second, less visible effect. A vendor that sees its people pulled in three directions starts protecting itself with reporting: weekly timesheets broken down by “requester”, emails asking for priorities to be confirmed, meeting minutes. That is not bad faith; it is a rational response to missing rules. But the client is paying for the specialist’s time, and some of that time now goes into proving who asked for what.
The problem is sharpest where the client directs the vendor’s people, as in staff augmentation and time-and-materials work. In business process outsourcing and managed services the vendor owns the outcome, and a service manager on the vendor side sits between the client’s departments and the vendor’s staff. That layer protects the relationship only while client departments do not bypass it. Once they start messaging the vendor’s people directly, the service quietly turns into staff augmentation without a staff augmentation contract. We compare these models in the article on BPO, shared service centres and IT outsourcing.
Finally, there is a legal effect. The more people on the client side give the specialist day-to-day instructions, decide their hours and appraise them like department staff, the more the relationship starts to look like employment by the client. I return to this in a separate section, because the rules differ by country and some of them changed recently.
Why shouldn’t an external specialist report to a functional head?
This is the central claim and it needs defending, because intuition points the other way. The functional head knows the discipline best. The head of quality knows how test cases are written at the company, what the audit requires and which areas are historically weak. Why shouldn’t she direct an external QA engineer?
First: an external specialist is bought for a specific purpose, and that purpose has an owner. The vendor contract, the purchase order and the budget are usually attached to a project or product. If a functional head assigns work outside that purpose, the company pays from the project budget for departmental work and the project manager does not know. In management accounts it looks like a project cost overrun; in reality it is a hidden transfer of resources between departments.
Second: accountability follows decision rights. The same mechanism appears in the article on who owns what in a project with an external team, through the lens of delivery managers and project managers. If the project manager is to be accountable for the deadline, they must decide what the team works on this week. If a functional head can override that at any moment, the project manager is accountable for something they do not control.
Third: departments have better levers than direct instructions. Standards, templates, quality reviews, acceptance criteria, a seat at retrospectives, a veto at acceptance if work falls short. The head of quality does not need to tell the engineer to rewrite test cases. It is enough to put the new template into the Definition of Done and have the project manager plan time for it. The change then passes through the person who knows the deadlines instead of going around them.
Fourth: an external specialist has no way to push back on too many requests. An employee can go to their manager and say they are overloaded. An external specialist would have to go to the vendor, and the vendor would have to go to the client and say its people are receiving conflicting instructions. In many organisations that conversation is heard as a complaint, so it never happens.
The counter-argument is serious and deserves a fair hearing. Sometimes an external specialist does work for a department rather than a project: system maintenance, administration, support for an analytics team. Then the department is the project, and the functional head is the right point of assignment. The thesis is not “never the department” but “always one point”. If the specialist works for a department, nobody from a project should assign them work without going through the functional head. The problem is not which line assigns the work. It is that both do.
Who may give instructions to an external specialist, and who may not?
The word “instruction” hides three different things. It helps to separate them before deciding who owns what.
What needs to be done. Tasks, priorities, sequence, deadlines. These are decisions about the purpose of work. Where the client directs the team under its contract with the vendor, these decisions sit with the client, but they should come from one point: the project manager, the product owner, or the functional head if the specialist works for a department. Why that point must exist and must have time for decisions is the subject of the article on the client-side product owner.
What the result must look like. Standards, templates, quality requirements, architecture, security, compliance. These are decisions about criteria, not about time. They may come from functional departments, but they should reach the team as requirements, not as one-off requests to a particular person.
Personnel matters. Availability, hours, leave, absence, appraisal, development, pay, discipline, cover and replacement. These are decisions about the person, not the work. For an external specialist they should sit with the vendor that employs or contracts them.
| Type of decision | Example | Who decides | Who is consulted | What to avoid |
|---|---|---|---|---|
| Tasks and priorities | Sprint order, regression deadline | One point on the client side (project or department, never both) | Vendor, tech lead | Side requests from other departments |
| Standards and criteria | Test case template, audit requirements | Client’s functional department | Project, team | Rolling out standards through requests to one person |
| Acceptance of work | Sign-off of a task or release | Client’s project | Department (compliance with standards) | Acceptance by someone who did not assign the work |
| Availability and hours | Weekend work before a release, on-call | Vendor, at the client’s request via the single point | Project | Agreeing hours directly with the specialist |
| Leave and absence | Two weeks off in August | Vendor, with notice to the client | Project (planning) | Approving leave in the client’s HR system |
| Appraisal and development | Annual review, training | Vendor | Project (feedback) | Enrolling the specialist in the client’s performance cycle |
| Replacement | Swapping a person who is not a fit | Vendor, at the client’s request | Project, department | The client “dismissing” the specialist directly |
The table is not theory. It answers the question nobody asked in the Monday scenario: who had the right to tell the QA engineer to rewrite test cases this week? According to the table, nobody in the quality department. The head of quality could introduce the new template as a standard and ask the project manager to plan time for it. The vendor’s delivery manager could propose that the engineer take on work from another project, but whether he had capacity for it was the call of the project manager who pays for that capacity.
Where does misclassification risk begin, and what has changed recently?
This section is not legal advice. It describes rules a COO should know before setting up how the organisation works with external specialists. Have an employment lawyer review your specific contracts and practices, in each country where you engage people.
The common thread across jurisdictions is that facts beat labels. The contract can say “services”, but if the client actually controls how, when and where a person works, integrates them into its organisation and manages them like staff, tax and labour authorities may treat the relationship differently. What follows are three examples relevant to companies that engage external specialists, including nearshore teams from Poland.
United Kingdom. Since 6 April 2021, under the off-payroll working rules (often called IR35), medium and large clients in the private sector are in most cases responsible for determining the employment status of a contractor who works through an intermediary, and must give a status determination statement with reasons. HMRC offers the CEST tool to help with the assessment. How these rules apply to engineers based in Poland is covered in our article on IR35 and engineers from Poland.
United States. The IRS applies a common-law test with three categories: behavioural control (does the company control or have the right to control what the worker does and how), financial control (are the business aspects of the job controlled by the payer) and the type of relationship (written contracts, employee-type benefits). The IRS stresses looking at the entire relationship and the extent of the right to direct and control the worker. State laws and other federal agencies apply their own tests, which is another reason for local advice.
Poland. Article 22 § 1 of the Polish Labour Code defines employment as work performed for the employer, under its direction, at a place and time it sets; § 1¹ says such work is employment regardless of what the contract is called, and § 1² prohibits replacing an employment contract with a civil-law contract while keeping those conditions. An amendment to the National Labour Inspectorate Act (Journal of Laws 2026, item 473), in force since 8 July 2026, allows a district labour inspector to declare by administrative decision that an employment relationship exists, once an earlier order to remedy the breach has not been complied with. The same amendment brings businesses that are not employers, but for which individuals including the self-employed perform work, within the scope of these inspections, and lets companies request an individual interpretation from the Chief Labour Inspector for a PLN 40 fee, issued within 30 days. Poland also regulates temporary agency work separately: an agency may assign the same temporary worker to one user employer for no more than 18 months within any 36-month period. Staff augmentation and temporary agency work are different arrangements, as explained in the glossary entry on the differences between body leasing and temporary work, and the more the relationship looks like a user employer directing an agency worker, the harder it is to argue it is a service.
What does this have to do with a matrix? A great deal. With an external specialist, the risk does not come from someone on the client side assigning tasks. That is the essence of staff augmentation and is described in the vendor contract. It comes from more and more people inside the company treating the specialist like a member of their department: setting their hours, approving their leave, appraising them in the performance cycle, putting them on the department’s on-call rota, applying the staff handbook to them. A matrix without rules multiplies this behaviour, because each line wants the same control over the specialist that it has over its own people. Our article on legal pitfalls of body leasing compared with employment contracts goes deeper into the Polish rules.
The practical conclusion for a matrix organisation: one point of assignment and personnel matters with the vendor are not just tidy organisation. They are also the simplest way to make sure the actual relationship matches what the contract says.
How do you build a RACI across three parties: department, project and vendor?
A RACI for a project with an external team usually has client and vendor columns. In a matrix organisation that is not enough, because the client side has two lines that can contradict each other. The missing column is the functional head. Without it the RACI looks clean, while in practice the department steps into the team’s work around the table.
A reminder of the letters: R does the work, A is accountable for the result and gives final approval, C is consulted before a decision, I is informed afterwards. The one-A-per-row rule matters especially here, because the two client-side lines in a matrix naturally want an A in the same rows.
| Decision or activity | Sponsor | Project manager | Functional head | Vendor delivery manager | External specialist |
|---|---|---|---|---|---|
| Project goal and scope | A | R | C | I | I |
| Priorities and tasks for the coming period | I | A/R | C | I | C |
| Quality standards, templates, architecture | I | C | A/R | I | C |
| Definition of Done | I | A | R (department requirements) | I | C |
| Acceptance of work | I | A/R | C | I | I |
| Request for work outside the project | I | A (approval) | R (makes the request) | I | I |
| Hours, on-call, non-standard working time | I | C | I | A/R | C |
| Leave and absence | — | I | — | A/R | R (notifies) |
| Appraisal, development, pay | — | C | C | A/R | I |
| Replacing the specialist | I | C (request) | C | A/R | I |
| Project vs department priority dispute | A | R | R | I | — |
| Contract-level dispute | A | R | I | R | — |
Three rows make most of the difference. “Request for work outside the project” turns a functional head’s informal instruction into a formal request, approved by the person responsible for the specialist’s time. “Project vs department priority dispute” moves the decision from the specialist to the sponsor. “Hours, on-call” states plainly that the specialist does not agree these directly with client departments.
This table complements the RACI in the article on delivery managers and project managers rather than replacing it. That article splits roles between client and vendor. This one adds the second client-side line and shows where it may and may not step in. In a project-based company the functional head’s column will be almost empty. In a weak matrix it will be the most important column in the table, which is exactly why it has to be written down.
How do you design escalation so priority conflicts don’t land on the specialist?
A matrix without escalation pushes disputes down to the lowest level. Inside the company that means an employee trying to please two managers. With an external team it means a person from another company with the least authority and the least information. A good escalation path has three levels and clear time limits.
Level one: the point of assignment. Any request to the specialist outside their current work goes to the person who assigns their work. The specialist has one answer to every such request: “happy to help, please agree it with the project manager”. That answer has to be written into the working agreement, so the specialist does not risk being seen as unhelpful. The vendor should brief the specialist on it at the start.
Level two: the project manager and the functional head. If the project manager does not agree to a department’s request and the functional head considers it essential, the two talk directly. A time limit helps: two working days. After that the dispute goes up, and until it is resolved the specialist keeps working to the existing priorities.
Level three: the sponsor or steering committee. It settles the dispute and records the decision. The vendor is informed of the outcome, not asked to decide. The vendor should not decide whether its person works for the project or the department: it does not know the company’s priorities and has an obvious conflict of interest, since any decision pleases one internal client and annoys another.
Escalations about the vendor itself follow a separate path: the quality of the specialist’s work, their availability, conduct or the need for replacement. These go from the project manager to the vendor’s delivery manager or account manager, not directly to the specialist. A functional head with concerns about the specialist’s work passes them on through the project manager or, if they concern standards, raises them as a quality issue.
The PMO’s role here is simple. It makes sure an escalation path is written down for every project involving a vendor and checks each quarter how many project-versus-department disputes reached level three. If many, the company’s matrix has a problem with how authority is split, not with vendors. If none, it is worth checking whether specialists are quietly settling disputes themselves.
Who appraises an external specialist and how should feedback flow?
Appraisal is where a matrix most quickly starts to resemble employment. The company has a performance cycle, functional heads are used to reviewing everyone in their area, and the external specialist sits in the same meetings and uses the same tools. Including them in the cycle feels natural. It is a mistake for two reasons.
The first reason is legal, as described above. A formal appraisal in the client’s HR system, with annual objectives and a development conversation, is a feature of an employment relationship. The client buys a service from the vendor. It has every right to judge the quality of that service, but appraising the person as an employee belongs to whoever employs them.
The second reason is practical. A functional head who appraises an external specialist does not know the terms of their engagement with the vendor and does not decide their pay or development. The appraisal has no consequence other than teaching the specialist to treat the functional head as a second boss, which is exactly the effect we want to avoid.
So how should feedback flow? The client evaluates the service; the vendor evaluates the person. In practice: once a month the project manager gives the vendor feedback on the team’s work, and functional departments pass their comments to the project manager, who consolidates them. Day-to-day feedback on a specific piece of work, in code review, at acceptance or in a retrospective, is a normal part of teamwork and needs no intermediaries. The line runs between “this test case does not cover the boundary condition” and “this quarter I rate your engagement three out of five”.
The vendor, for its part, should put that feedback to use. A useful question when selecting a vendor: how do you collect client feedback and what do you do with it? If the answer is “we run a satisfaction survey once a year”, the project manager’s input goes nowhere and the client will start appraising people itself. Structured feedback also helps predict departures. The metrics described in the article on specialist attrition in outsourced projects start with someone regularly asking how the work is going.
How should a specialist be shared across projects, and what does it really cost?
A matrix encourages sharing people. A functional head has a pool of specialists and spreads their time across several projects, because that uses resources best. With employees it makes sense: the pool is stable and employees know all the department’s projects. With external specialists the temptation is even greater: we pay by the hour, so the hours can be split any way we like.
Sharing an external specialist across projects has three costs that do not show up in the spreadsheet. The cost of switching: each project has its own context, tools and people, and a specialist working on three projects loses part of their time remembering where they left off. I will not quote a percentage, because the commonly cited estimates rest on observations from the 1990s and vary in method. The direction is obvious to anyone who has worked on several projects at once.
The accounting cost: if a specialist works for several projects, someone has to allocate their hours to budgets. In a matrix this is usually the functional head or the specialist. Either way, project managers lose control over what they pay for. The accountability cost: when the specialist misses something on one project because another had an urgent need, who is accountable? Every project manager will say their priority was agreed.
A practical rule: if an external specialist is to work on several projects, one person sets the split, usually the functional head or the PMO, and does it in fixed blocks (say, three days on project A and two on project B) for at least a month. Changes to the split go through the same person and are communicated to the vendor. This is the one case where the functional head becomes the point of assignment: when deciding how time is divided between projects, not what happens within each.
A different engagement model is often the better answer. If a department needs a flexible pool of skills for several projects, a vendor team on a subscription basis, with its own backlog and a single owner on the client side, fits better than individual specialists assigned to many projects at once. We compare billing models in the article on which IT vendor billing model actually shares the risk.
What should go into the vendor contract and the team’s kick-off document?
The vendor contract usually knows nothing about the client’s matrix. It covers the service, rates, notice period, confidentiality and intellectual property. It rarely says who on the client side may assign work to the vendor’s people. Yet that is the information that decides whether the engagement runs the way the contract describes.
The contract or an operational annex should cover five things:
- The person authorised to assign work for each order or team, with a deputy. Name the role, not just the individual, so a change of person does not require an amendment.
- The single-channel rule. Requests from other client departments go to the authorised person, and the specialist is not obliged to act on them without that person’s approval.
- The split of personnel matters. Availability, working hours, leave, absence, appraisal and development sit with the vendor. The client communicates its needs through the authorised person.
- An escalation path with time limits: who on each side talks at which level, and how quickly.
- How feedback on service quality is given: frequency, format, and who on the vendor side receives it.
Then there is the team’s kick-off document: one or two pages prepared by the project manager in the first week. It contains the three-party RACI, a list of departments the team will deal with and the standards from those departments that apply. The specialist receives it together with their access rights. From day one they know to whom they say “yes” and to whom they say “please agree it with the project manager”.
It is also worth planning the end of the engagement. In a matrix, an external specialist’s knowledge is spread between the project and the department, and when they leave it is easy to hand it over to only one side. We cover how to plan this in the article on offboarding a contract specialist. One point is enough here: the knowledge transfer plan should have a department column in the RACI, not just a project one.
At ARDURA Consulting, the standard model is a B2B services agreement between the client and ARDURA, with the specialist working as our subcontractor or employee. Employment matters, invoicing and continuity of staffing stay with us: a two-week free replacement guarantee at the start, a replacement within five business days if a specialist leaves, and a 30-day notice period. That is precisely the column of the table where the client should not hold the A.
Which signs show that accountability has already blurred?
Blurred accountability rarely looks like a crisis. It looks like ordinary friction. So it is worth knowing the signs that tend to recur in matrix organisations working with vendors, and checking them every quarter.
The specialist has meetings with more than one department that are unrelated to the project. Department stand-ups, department reviews, department training they were never invited to through the project manager.
The project manager learns about their specialist’s work from the timesheet. If tasks appear in the hours report that the project manager did not assign, someone else is assigning work around them.
The vendor asks whose priorities matter more. That question means its people are receiving conflicting instructions and have no rule to resolve them.
The specialist books leave in the client’s system. Or appears on the department’s on-call rota, or has objectives in the client’s performance tool. This is an organisational and a legal warning at the same time.
Work is accepted by someone other than the person who assigned it. The project manager assigns, and the functional head rejects at acceptance because it does not meet a standard the team never heard of.
Escalations end at the vendor. The sponsor or steering committee never settles a project-versus-department dispute, while the vendor regularly hears that its people “are not delivering”.
Nobody can name the person authorised to assign work. The simplest test: ask three people on the project who may assign a task to a particular external specialist. If you get three different answers, accountability has already blurred.
The PMO can turn these signs into a short survey for project managers and vendors. The point is not control but finding the places where the matrix has started treating the vendor’s people as its own staff, with all the consequences that brings.
What are the key takeaways?
A matrix organisation is a compromise a company makes deliberately: better expertise from departments, better delivery from projects, in exchange for two bosses and constant negotiation. Employees have the means to live with it. An external specialist does not, and when the vendor becomes a third boss, accountability dissolves across three parties.
Remember six things. First, whether your company runs internally as a weak, balanced or strong matrix, towards the vendor’s people it should behave like a strong matrix: one point of assignment, tied to the goal that pays for the work. Second, separate three kinds of instruction: what to do (one point on the client side), what the result must look like (departments, through standards) and personnel matters (the vendor). Third, add a functional head column to your RACI, with a row for “request for work outside the project” and a row for “project vs department dispute” with the sponsor as A. Fourth, design an escalation path with time limits, so disputes are not settled by the specialist or the vendor. Fifth, evaluate the service and leave appraisal of the person to the vendor. Sixth, remember that the actual working relationship counts for more than the contract in the UK, the US and Poland alike, and Poland’s labour inspectorate gained new powers in July 2026.
If you are planning to bring external specialists into teams in a matrix organisation, see how ARDURA’s staff augmentation services work: a contract with ARDURA, employment and continuity of staffing on our side, and the specialist working in your projects to your standards. If you want to talk through how to split roles on a specific project, get in touch.