What is job architecture? Job architecture is the published structure of work in an organisation: the functions or job families, the roles inside them, and the levels that describe increasing scope, impact and accountability. It is not the same as a list of job titles. Titles can vary by team or market; the architecture is the shared language underneath, so two “seniors” in different functions can be compared.

What is job levelling, and how do you build a framework? Job levelling (also called job leveling) is the written rule for placing a role at a level, agreed before anyone is mapped or promoted. For a mid-sized company, start with two or three job families, agree an individual-contributor track and a manager track, and write what changes from one level to the next. Use the same levels for calibration, promotion decisions and — separately — pay bands. A competency matrix then describes what good looks like inside a level; a talent bar is the performance standard at that seniority. There is no official “right” number of levels, and a consultancy job-evaluation method is not required to publish a first version.

This page is for People leaders and HR teams at employers of about 200 to 2,000 people whose titles, promotions and pay bands have grown by habit. It is a job architecture and levelling framework you can socialise with managers — not a student “career pathing” guide, not a game-levelling walkthrough, and not a consultancy job-evaluation pitch. If you already have a title list and a first cut of levels, you can book a 15-minute review with Ash Taherian, Product Sales Executive, to walk through it.

What job architecture is

Job architecture is the published structure of work in the organisation. In most mid-sized companies that means three things, written down in one place:

  • the functions or job families you actually have — for example engineering, finance or customer operations;
  • the roles inside each family; and
  • the levels that describe increasing scope, impact and accountability.

It is not a list of job titles. Titles can vary by team, market or hiring manager habit. One team’s “senior” is another team’s mid-level hire. The architecture is the shared language underneath those labels, so a promotion, a pay decision and a calibration can all point at the same level.

A job architecture framework is simply that structure plus the rules for using it: how a role is placed at a level, who owns the map, and how it is kept current. “Job family” here means a group of related roles, not US “job family services”.

Illustrative example only — not a real company, and not a recommended structure. In a fictional Product function you might publish three roles — Product Manager, Product Designer and Product Operations — and a short shared level set: Associate, Professional, Senior and Principal on the individual-contributor track, and Manager, Senior Manager and Director on the manager track. “Senior Product Manager” is a title. The architecture says which role and which level that title maps to. A fuller starting grid is in the template below. Rename the levels to match your own language.

What job levelling is

Job levelling — also called job leveling — is the set of rules that places a role at a level. Those rules should be written down before anyone is promoted or mapped. Titles describe what you call someone. Levelling describes the scope of the work.

A common design choice is two tracks: an individual-contributor (IC) track and a manager track. Some organisations also add a specialist or technical track where deep expertise grows without a people-management span. That can be a good fit later. It is not a requirement for a first published version.

There is no official “right” number of levels. Do not copy a bank’s or a Big Four firm’s long grade structure into a 400-person company, and do not treat a software product’s maximum number of seniorities as a market standard. Pick a small number of consistent criteria — for example scope, impact and decision rights — write what changes from one level to the next, and use the same criteria across families. You do not need a proprietary job-evaluation method (Hay, Mercer IPE, Korn Ferry, Radford or similar) to get a usable first version published. Those methods exist; they are not the Playbook method, and this page does not reproduce their factors or level names.

How this differs from a competency matrix and a talent bar

This is the split that keeps “ready for the next level” from becoming a vibe. The three artefacts do different jobs, and mixing them is how reviews lose their meaning.

How job architecture, a competency matrix and a talent bar fit together. The last two are already covered on their own Playbook pages.
ArtefactWhat it isWhat it is not
Job architectureWhich families, roles and levels exist, and how a role is placed at a level.A title list, a competency matrix, or the performance standard you score against.
Competency matrixWhat good looks like inside a level, in observable behaviours for that role.The levels themselves. The matrix hangs off the architecture; it does not invent it.
Talent barThe performance standard you score against at that seniority, so reviews are comparable.A career ladder, and not a substitute for writing what changes at the next level.

If you write a competency matrix and call it the architecture, you still will not know whether two “seniors” in different functions sit at the same level. If you publish a talent bar without a level set, you have a performance standard with nothing stable to hang it on. Build the skeleton first. Then use the existing matrix and talent-bar pages for the work inside each level.

How to build a framework in a 200–2,000-person company

This is a practical sequence, not a twelve-month grading programme. Start with two or three critical families rather than the whole company. A first cut you can socialise with managers is more useful than a perfect model no one has seen.

  1. List the job families you actually have. Use the work people do, not the org-chart labels you wish you had. Merge near-duplicates. Leave unusual or one-person titles to one side until the core families are stable.
  2. Agree an IC track and a manager track, and a short level set. Give each level a name you can say in a calibration. You can add a specialist track later if you need one. Do not invent a “correct” length of ladder.
  3. Write level descriptors that say what changes at the next level. Use the same criteria across families — scope, impact and decision rights are enough to start. Write the difference between adjacent levels in plain language a manager can use, not a paragraph of values statements.
  4. Map current titles to roles and levels. Do this with People and the function lead, not from memory in a spreadsheet no one owns. Record the exceptions and the reason. A title that sits in two levels is a signal to fix the title or the role, not a reason to blur the level.
  5. Publish, then use the levels in the next calibration and promotion cycle. A framework that lives only in a slide deck is not published. Managers need to see it before they write reviews or make a promotion case.
  6. Decide later how pay bands sit on the levels. Bands hang off the level; they should not be designed first and reverse-engineered into a ladder. Once the levels exist, the merit matrix is how ratings become pay. This page is not a salary-bands explainer.

The career conversation that uses the path sits on the employee development plan and career conversations guide. Critical roles in the same map belong in succession planning.

A starting levelling template

The grid below is an illustrative starting template a People team can copy and rename. It is not a market benchmark, not a recommended number of levels, and not a default taken from any software product. Invented salaries are omitted on purpose. Replace the level names with the language your organisation already uses, or with names you are prepared to defend in calibration.

Illustrative only. Scroll sideways on a small screen.

Starting levelling template — individual-contributor and manager tracks. Rename the levels. Do not treat this as a market standard.
LevelTrackTypical scopeDecision rightsWhat the next level adds
AssociateICOwn pieces of work inside a team, with regular guidance.Decides how to complete assigned work; escalates trade-offs.Professional: owns a whole workstream, not only assigned tasks.
ProfessionalICOwns a workstream or a clear area of the team’s delivery.Makes day-to-day decisions in that area; flags cross-team risk.Senior: sets approach for a domain and is trusted with harder trade-offs.
SeniorICLeads a domain or a complex area; influences adjacent teams.Decides approach in that domain; involves a manager on organisation-level calls.Principal: shapes direction across teams, not only inside one domain.
PrincipalICSets direction across several teams or a function-wide craft.Makes or frames decisions that other teams follow; accountable for the quality of that direction.If you add a further IC step, it should change organisation-level impact, not just the title.
ManagerManagerLeads a team; accountable for that team’s delivery and development.Decides team priorities and day-to-day people matters, within the function’s rules.Senior Manager: leads managers or a broader area, not only one team.
Senior ManagerManagerLeads a group or a broader area, often through other managers.Sets priorities across teams; owns harder people and resource trade-offs.Director: sets function or multi-team direction and is accountable for it.
DirectorManagerOwns a function or a major area; sets direction other managers execute.Makes or recommends organisation-level decisions for that area.If you add a further manager step, write the new scope and decision rights first.

Use the same level names across families wherever you can. If finance and engineering need different labels, keep the underlying criteria identical so a “senior” in one family can still be compared with a “senior” in the other. Map each current title to one cell in a grid like this, with People and the function lead, before the next promotion cycle starts.

How levels are used in ratings, promotions and pay

A published level is what a rating is scored at. The person is reviewed against the expectations of their current level, on your published performance rating scale, and the draft is tested in calibration (see also how to run a calibration meeting). “Exceeded” at one level is not the same decision as “ready for the next level”.

“Ready for promotion” means evidence against the next level’s descriptor, not a potential score on its own. Potential can still be discussed; it should not replace evidence of scope, impact and decision rights. That distinction is set out in the note on potential ratings and promotion bias.

Pay bands, if you use them, hang off the level. The merit matrix then sits on top, so a rating becomes a pay decision in a way you can explain. Whether reviews should be linked to pay at all is a separate evidence question — see should performance reviews be linked to pay. This page does not summarise UK or EU pay-transparency rules, and it does not try to rank for “salary bands” or “pay transparency” searches.

The same map should appear in career conversations and in succession plans: the path a person might grow into, and the critical roles you cannot afford to leave unmapped. The wider system those pieces sit in is the performance management framework.

Common mistakes

  • Treating job titles as levels. If the title is the only record of seniority, two “seniors” will not be comparable.
  • Inflating titles instead of changing scope. A new title with the same decision rights is not a promotion in the architecture.
  • A different ladder in every function. Local labels can vary; the criteria should not.
  • Writing a competency matrix and calling it the architecture. The matrix describes behaviour inside a level. It does not tell you which levels exist.
  • Copying a 15-level structure from a much larger firm into a mid-sized company. Start shorter. Add a level only when you can write what it changes.
  • Designing salary bands first and reverse-engineering the levels. Bands hang off the level, not the other way around.
  • Leaving the framework in a slide deck that managers never see. If it is not used in the next calibration, it is not the architecture yet.
  • Mapping people after a promotion round has already started. Write the rules, then map, then promote against the next level’s descriptor.

Where Revolut People fits

Revolut People does not have a module titled “job architecture”, “job levelling” or “career framework”. If you use the product, map your framework to what the documentation does describe. The points below are from Revolut People’s live documentation, checked on 15 November 2026, and are not a claim that the product ships this page’s template.

The product uses a functional structure of Function → Role → optional Specialisation. A Function is a high-level discipline (the docs give Engineering, Marketing and Sales as examples). A Role is a defined position within that function. A Specialisation is an optional, more specific variation of a role; it must be linked to a parent Role and cannot exist independently. Specialisations are not enabled by default. See Introduction to Roles and Introduction to Specialisations.

Job titles and Roles are not the same thing. A Job Title is a label for an employee’s position and is often external-facing; titles are optional. A Role defines responsibilities, required skills and competencies, and Roles are required to use the platform. The docs also distinguish a business structure (Company → Departments → Teams) from that functional structure. See Job Titles vs Roles.

Each role includes a competency matrix that outlines expectations at every seniority level. The same Roles introduction says roles connect to competency matrices, performance reviews, goal alignment, calibration and career paths. How to build a matrix is already on the competency matrix templates page.

Seniorities define an employee’s experience level within a role. The docs say that by default an organisation can have a maximum of eight seniority levels, and that seniority names can be renamed to match the organisation’s levelling framework. Eight is a product maximum, not a market benchmark and not a recommended number of levels. Custom seniority groups can support different progression paths, including technical ladders; each group also has a maximum of eight levels. See Managing Seniorities and Custom Seniority Groups.

Compensation bands, if you use them, can be set by a combination of attributes such as seniority. Each band stores a lower amount, an upper amount and a currency; there is no stored midpoint. How ratings become pay is the merit matrix page. See Introduction to Compensation Bands (checked 15 November 2026).

If you want a second look at a current title list and, if you have it, a first cut of levels, book a 15-minute review with Ash Taherian, Product Sales Executive. Soft fit-check only.

FAQ

What is job architecture? What is a job architecture framework?

Job architecture is the published structure of work: functions or job families, the roles inside them, and the levels that describe increasing scope, impact and accountability. A job architecture framework is that structure plus the rules for placing a role at a level and using those levels in calibration, promotions and — separately — pay. It is not a list of job titles.

What is job levelling, and how does it differ from job titles?

Job levelling (also called job leveling) is the written rule for placing a role at a level, agreed before anyone is mapped or promoted. A job title is a label, often external-facing, and it can vary by team or market. Two people can share a title and not sit at the same level, or sit at the same level with different titles.

Is a job title the same as a level?

No. The title is what you call the person. The level is the published scope, impact and decision rights of the work. Map titles to levels; do not treat the title as the level.

How do you build a job architecture or levelling framework in a mid-sized company?

Start with two or three job families. Agree an individual-contributor track and a manager track, and a short level set. Write what changes from one level to the next, using the same criteria across families. Map current titles with People and the function lead, publish the map, and use it in the next calibration and promotion cycle. Decide later how pay bands sit on the levels.

How do job architecture, a competency matrix and a talent bar fit together?

The architecture says which levels exist and how a role is placed. A competency matrix says what good looks like inside a level, in observable behaviours. A talent bar is the performance standard you score against at that seniority. Keep the three distinct so “ready for the next level” is evidence against the next descriptor, not a vibe.

How do levels connect to promotions, pay bands and performance ratings?

A rating is scored at the current published level, on your rating scale, and tested in calibration. Promotion is evidence against the next level’s descriptor. Pay bands, if you use them, hang off the level; the merit matrix then turns a rating into a pay decision.

Is there a starting template?

Yes. The starting levelling template on this page is an illustrative grid with an individual-contributor track and a manager track, plus columns for typical scope, decision rights and what the next level adds. Rename the levels. It is not a market benchmark.

Do we need a consultancy job-evaluation method first?

Not to get a usable first version published. A small set of consistent criteria, written descriptors, and a map of current titles is enough to start. Proprietary methods exist; they are not required for a mid-sized company to put a short, shared level set in front of managers.

Related guides

Talent bar by seniority · competency matrix templates · merit matrix · employee performance rating scales · performance calibration · how to run a calibration meeting · development plans and career conversations · succession planning · potential ratings and promotion bias · should performance reviews be linked to pay · performance management framework · free Playbook tools