Inside
Use it in the next meeting.
Write the roles
Set the six shared boundaries, then add the craft and decision rights each local role needs.
Collect evidence during the work
Record outcomes, context and observable behavior while people can still correct the facts.
Run calibration
Prepare each case, work through disagreement and leave with a recorded decision.
Editable templates included
Company setup, performance record, evidence log, self-assessment, calibration scorecard, promotion case and transfer record, plus a worked example.
Where to start
Complete Company setup with your partners. Agree the role expectations and working standards, then use the evidence log and grading examples before calibration.
“Renn’s model puts evidence, judgment and decision in the same conversation, where research and design can change an outcome instead of decorating one.”
Jenny Lo · Head of User Research, Harvey AI
Across the three workbooks
Run the cycle with People and Finance teams.
Functional leaders
Write the roles, bring the evidence and make room for the next assignment.
People
Keep the standard consistent and give disputed evidence somewhere to go.
Finance
Set the limits, check the cost and approve the final numbers.
One case through all three books
Follow a fictional team through calibration, an over-budget pay proposal, an approved timing change and a development plan with protected time.
The evidence stays put
If a pay proposal exceeds the budget, change the amount or its timing. Do not rewrite the assessment.
“Renn understands that clarity and humanity are not opposites. This system makes scope, pay and growth discussable without reducing creative people—or their ambition—to boxes on a ladder.”
Li-Anne Dias · Executive Creative Director, Brex
Written by Renato Valdés-Olmos
Built from running the cycle.
I have run leveling, compensation and development cycles at companies of 25, 250 and 2500 people. These workbooks turn those mechanics into documents you can use with your team. The shared six-level framework is new in v1.
More about my work
One complete chapter · No email required
Read section two.
02The shared core: six levels, six axes
The shared core covers engineering, product, research, design and marketing. A level records the scope and consequence of a role’s judgment. The discipline guide records the craft needed to carry it. This gives equal level numbers a common meaning while preserving the standards of each field. Local roles and operating records add the company-specific detail.
Craft is required, but it cannot be the common scale. An excellent systems architect, product strategist, researcher and brand designer demonstrate different expertise, and the useful craft list changes as the work changes. Each role therefore names a primary specialism, any supporting specialisms and the depth required. Level the role first from its durable mandate; then assess the person from evidence of the judgment and craft the role requires.
The model has four layers
| Layer | What it contains | Why it is separate |
|---|---|---|
| 1 · Shared core | Six decision thresholds and six evidence axes used across every covered discipline. | Makes levels comparable across functions without forcing the work into one generic competency matrix. |
| 2 · Discipline capability | Engineering, Product, Research, Design or Marketing craft, professional standards and the depth the role requires. | Protects expertise, assurance and quality. A shared level never excuses weak craft. |
| 3 · Role profile | The actual mandate, specialism, context, IC or management lane, local title and any within-level step. | Preserves useful company differences and makes clear which opportunities and authority the role genuinely provides. |
| 4 · Operating record | Dated evidence, counter-evidence, calibration rationale, framework version, review route and change log. | Turns the framework into a living decision system rather than a poster or a checklist. |
Do not collapse these layers. A title is not a level. A craft matrix is not a promotion score. A good outcome is not proof that one person owned the decision. A shared threshold is useful only when the role profile and discipline standard make it concrete.
| Level | Decides | In one sentence |
|---|---|---|
| 1 | How | Chooses the approach inside a problem somebody else framed and scoped. |
| 2 | What | Can be handed a goal instead of a spec, and decides what to build against it. |
| 3 | Which problem | Decides what is worth solving, and says so before anybody asks. |
| 4 | What we refuse | Holds a position with a commercial cost attached, in the room where money is allocated. |
| 5 | What outlives us | Encodes a standard so it operates without the person who holds it. |
| 6 | What we will not do | Sets the constraints everyone else works inside, in the company's own terms. |
Why six levels rather than nine
Six is the default here because I can name and test six changes in responsibility: how the work is done, what gets made, which problem matters, what the company refuses, which standard should outlast its author and which constraints govern the company. Titles, pay steps and craft maturity can add detail inside those boundaries.
A company can still use titles such as Senior, Staff, Principal, Director, Senior Director and VP. Several titles may map to one shared level when their decision boundary is the same, while the discipline, management lane, organizational scope and pay band preserve the distinctions that matter. This avoids manufacturing a new level to reward tenure, recognize scarce craft or make room for another title.
Use nine levels if you can show nine durable boundaries. Test each extra level with one sentence: “Unlike the level below, this person may decide ___ without further approval, and we will accept ___ as evidence.” Try the answer on more than one role and one real calibration case. If the boundary disappears, keep the distinction as a title, craft depth, management scope or position in band.
What the comparison changed
| First assumption | Rule after testing it |
|---|---|
| Six levels are the model | Six is the shared translation layer. Local titles, sub-levels and within-level steps can remain when they make a useful distinction. |
| Decision rights define a person | Decision responsibility anchors the role. It includes final calls, and also authority over questions, evidence standards, technical constraints and professional assurance. |
| Craft is separate | The disciplines share a level ruler and keep their own craft gates. Placement requires both the role’s decision responsibility and the craft needed to carry it. |
| Observed authority is capability | Level the role's mandate first, then place the person from evidence. When authority or opportunity was withheld, mark the axis not gradeable and fix the role or assignment. |
| IC and management levels match | Parity is a design requirement: equal authority thresholds with different work. Test the actual mandate; the same number does not guarantee equal pay or an available same-level role. |
| Six marks produce a level | The axes structure the evidence; they are not six equal points in a score. The calibration room makes and records the judgment instead of averaging the marks. |
What this model inherits from the strongest public frameworks
A public framework earns a place here by doing more than naming a good idea. Its source must show the problem behind the choice, or the same lesson must appear in several working systems. I also need to have seen the failure in a real cycle and be able to turn the lesson into a decision, behavior or control another company can test.
| Lesson retained | Where it is visible | How it changes this model |
|---|---|---|
| Career progression is an organizational design problem. | Peter Merholz and Kristin Skinner, Org Design for Design Orgs. | The level describes a role’s responsibility in the organization, not just its craft skills. Ownership is an explicit axis; section 5 makes it observable in the work. |
| A common structure needs specialist matrices. | GitLab, Dropbox and the UK Government Digital and Data framework. | The shared core supplies comparability; each discipline supplies its own capability and assurance gate. Section 3 maps the threshold and section 6 records the required craft depth. |
| Progression is a pattern, not a checklist. | Monzo, Wise and Intercom. | The six axes organize evidence but are never averaged. Repeated next-level evidence, craft and counter-evidence are judged together in calibration. |
| People should be judged on work they could influence. | Intercom Product Design. | Every axis can be marked not gradeable. The role and manager must provide the mandate, opportunity and support before absence of evidence can count against a person. |
| Large jumps and growth inside a level are different. | Buffer. | A level changes only when the durable mandate changes. Titles, craft depth, ownership milestones and position in band can recognize growth inside a level. |
| Senior IC work has more than one shape. | StaffEng archetypes, GitLab and Intercom. | Section 6 records primary and supporting specialisms; the management fork does not become the only route to authority, influence or pay progression. |
| Management is a different craft and the move is reversible. | Intercom Research and Data Science and Wise. | The fork begins at level 3. Equivalent numbers require an equivalent mandate, while management evidence is assessed separately from individual craft output. |
| A framework must be maintained in public view. | GitLab's living handbook, the UK Government framework's dated updates and Intercom's revisions after employee feedback. | Every cycle names an owner, version and effective date. Managers record ambiguous boundaries and missing expectations; cross-functional owners review them after the cycle and publish changes before the next evidence period. |
When you borrow a row, record the problem it solves, where it sits in the four-layer model and who owns it. Then try it on a real case. Keep company values, titles, salary data and role boundaries local. Anything that fails the case stays in the references.
Every level is described against the same six axes. A case is ready for calibration only when it contains relevant evidence on decision rights and at least three other axes. Promotion requires repeated next-level evidence on decision responsibility and across several other axes, the required craft, and an honest account of counter-evidence. The room records the judgment and rationale; it does not average six marks. A case argued on one axis is a retitle, not a promotion.
| Axis | The question it asks |
|---|---|
| Decision rights | What does this person decide alone, what with one named other person, and what do they escalate? Written as a list of decisions, never as a scope statement. |
| Judgment density | Does adding them to a room add throughput, or add decisions? On a stuck project you are almost never short of capacity. You are short of people able to end an argument. |
| Opinions held | What contestable positions do they hold about what should exist, and have they defended one when defending it cost something? |
| Range | Which decisions are they trusted to make in a lane that is not theirs? |
| What survives them | What sits in a component, a schema, a definition of done, a constraint on what a model may generate? |
| Ownership | Informed, involved, invested. Carried over intact. |
Decision rights over agents
Agent authority belongs in the decision-rights row. Record what the person may delegate to an agent, what they must verify and what still needs human approval. Boris Cherny's original post and replies describe prototyper, builder, sweeper, grower and maintainer as temperaments that cut across job functions. Editor, scout, orchestrator and risk steward extend the idea to people and signals. These describe how somebody works at a level. They do not create new levels.
| Level | May let an agent decide | May not |
|---|---|---|
| 1 | How, within the task, with the output reviewed. | Anything that changes the task. |
| 2 | What to build against the goal, on their own surface, with the trail kept. | What ships to a customer without a person having read it. |
| 3 | Which options to generate and which to discard, inside a problem the person framed. | Which problem. The framing is the level. |
| 4 | Whatever the team's written refusals permit, without a person in the loop. | Anything the refusals cover. Writing the refusals is the level. |
| 5 | Anything inside the encoded standard. | The standard. That the constraint holds in the tooling, not in the review, is the level. |
| 6 | — | What a model may generate at all, in the company's terms. This is a level 6 decision and nobody else's. |
Why decision responsibility anchors the level
Scope is the word every ladder uses, and it is the reason promotions stop meaning anything. "Owns a larger surface" can be true of somebody who decides nothing on it. Decision responsibility forces the question to be answerable: name the calls, questions, standards or constraints this role now owns that the level below does not. Then test whether the person has exercised that mandate with the required craft. If the role never granted the authority or opportunity, the case is not evidence of failure; it is not gradeable yet.
Set the company context before you grade
Functional leaders, People and Finance agree the setup before the cycle opens. Record the business priorities, covered roles, cycle dates, evidence period and named decision owners in the Company setup template. Record which external framework elements are being retained, adapted or rejected, and why. The examples in this workbook cover engineering, product, research, design and marketing. Other job families need their own role mapping and calibration examples before adoption.
| Owner | Responsibility |
|---|---|
| Functional leadership | Define role scope and expected work, supply evidence and own development commitments. |
| People | Maintain the standard, support managers, facilitate calibration and arrange an independent review of disputed evidence. |
| Finance | Agree the financial envelope and cost assumptions before the cycle. Cost and approve pay proposals after calibration; keep funding constraints separate from assessment. |
Four questions, four recorded answers
A level describes scope and decision rights. Performance describes delivery during a defined period. Working standards describe how the work was done. Compensation records the funded action under the pay policy. Keep all four answers visible; a level or a count of above marks cannot answer all four.
| Question | Evidence to record |
|---|---|
| What level is the work? | Decisions owned, scope, six-axis evidence and the calibration rationale. |
| What was delivered? | Agreed outcome, baseline, result, quality and delivery context. Record dependencies and changed priorities before judging the result. |
| How was it delivered? | A company value translated into an observable behavior, a dated example and its effect on colleagues or customers. |
| What action is funded? | Proposed and approved pay, effective dates, budget impact, approvers and any deferred action with a review date. |
Use the Performance record alongside the calibration scorecard. Agree expectations at the start of the period and record changes when they happen. The person can add context and correct factual errors before calibration. People records the reviewer, decision, rationale and response date when a material disagreement remains. Do not treat a missing opportunity or a manager's withheld authority as a lack of ability.
Company values as working standards
Write behavior that can be observed in the work. For ownership: name the decision owner, surface a material risk early and close the commitment. For collaboration: invite the affected discipline before deciding and record how its evidence changed the choice. Use examples across different roles and working styles. Familiarity, confidence, social similarity and hours visibly spent at work are not substitutes for evidence.
Publish the examples with People before the period starts. Assess both results and working standards, with context; do not invent a universal weighted score. An unresolved conduct or policy issue follows the company's established process. The six axes are a guide to calibration, not an automatic employment decision.
Try it
Stress-test a level with one real decision.
Choose the highest decision boundary the person already owns, then test whether the evidence would survive calibration. The result is a copyable calibration note.
Highest boundary demonstrated
Evidence test
Your entries stay in this browser.
Choose a volume
Start with the decision in front of you.
Leveling
For EPD leaders who need Engineering, Product, Research, Design and Marketing to use the same level structure. People gets a standard it can run and audit.
Compensation
For CFOs and Finance teams setting pay across Engineering, Product, Research, Design and Marketing. Functional leaders bring the role and market evidence.
Development
For EPD and Marketing leaders working with People. Turn review evidence into live work, protected time and a manager commitment for the next 90 days.
The new standard
Leveling, compensation and development in the age of AI.
One standard for levels, pay and growth. The set includes three printable workbooks, 24 editable templates and reference sheets, and three worked examples. Save $18 compared with buying separately.
$117$99
Team license
All three workbooks and their templates for one leadership team or function at one company. Already bought a single copy? Its price is credited toward the team license. License details.
Enterprise license
Company-wide, multi-division or portfolio-company use, with the licensed entities and rollout scope agreed before purchase. Suitable when procurement, legal, security or implementation support needs to be part of the work. License details.
From $2,500
Contact about enterpriseThe thinking behind the tools
Read the related essays.
Before you buy
A few practical questions.
What do I receive?
You can read the workbook on this site and download a printable PDF and editable .xlsx templates. The set includes all three. No physical book is shipped.
Do I need the other workbooks?
Start here if you are introducing the system. Compensation and Development use the levels and six axes defined in this workbook.
Will this work with our HRIS and AI tools?
Yes. Map the spreadsheet fields into your HR system and use a stable employee ID. Keep private notes out of the approved record. An AI client or MCP connection can read the exported JSON through your own file or drive connector, check for missing fields and prepare drafts. A person still approves every level, pay decision and development commitment.
Why not use the framework built into our HR package?
You may be able to. HR software handles forms, permissions, workflow and storage well. This set helps the leadership team decide what the fields mean, which evidence counts and who may make an exception. If the package can preserve those choices and keep assessment, pay and development separate, configure the process there.
Why use this instead of a framework from an investor, COO or People team?
Begin with theirs. It may save months and fit your HR software already. Then test it against work your company has done: which decisions were delegated, what craft the role required and what settled the disputed cases? Keep every part that answers those questions. Use this workbook for the gaps.
Why six levels rather than nine?
Six gives this system enough room to distinguish how work is done, what gets made, which problem matters, what the company refuses, what should survive its author and which constraints govern the company. Add another level when you can name the new decision, separate it from both neighboring levels and find the same boundary in more than one role.
How do I get access?
After checkout, an email goes to the address you paid with. Its private reader link opens the complete workbook. You can also open the reader from this page and enter the access key from the email. The sample remains a separate public chapter.
Can I use it with my team?
A single purchase covers one person, with sharing permitted with your own manager and managers who report to you while running a cycle. The $499 team license covers one leadership team or function at one company. Company-wide, multi-division or portfolio-company use needs an enterprise license. A single purchase can be credited toward the team license. Read the license.
What if we need an enterprise rollout?
Enterprise licenses start at $2,500 for company-wide, multi-division or portfolio-company use, procurement requirements and an agreed rollout scope. Contact me with the organizations, approximate audience, countries and timing for a quote.
Can I get a refund?
Read the complete sample chapter before buying. Because the workbook, PDF and editable files are delivered immediately, checkout asks you to acknowledge that the purchase becomes final when digital access begins. Duplicate charges, persistent access failures, material differences from the description and mandatory legal remedies remain covered. Read the access and refund terms.
Are updates included?
Every revision of v1 is included and sent to the address you paid with. Your purchase does not subscribe you to the essay newsletter.
Can I get a company invoice?
Stripe sends a receipt after purchase. Reply to it with your company name and tax number if you need a company invoice. Any applicable tax is shown at checkout.
Build career levels your team can actually use.
Give EPD and Marketing leaders one standard for roles, performance and promotion. People gets the evidence trail and change controls to keep it consistent across teams.
One payment · Immediate digital access · v1 updates included
Read the sample before buying · Access and refund terms
© 2026 Renato Valdés-Olmos