# 90n Performance Review Guide

This guide is designed for agency founders and senior leaders to think through, design, and run a high-impact end-of-year performance review cycle. It maps out why the cycle matters, what a decade of academic research actually says about what works (and what quietly backfires), the core practices to make it successful, a month-by-month preparation calendar, a practical example of a behavioral career growth matrix, a detailed strategy FAQ, and full details on who we are and how we help agencies scale their people operations without the corporate HR overhead.

A note before you start: this guide is deliberately thorough. It is not meant to be implemented alone, in one sitting, from a cold start. The agencies that get the most out of a review cycle treat it as a design problem, and they bring in someone who has run it before. That is exactly what we do. If at any point a section makes you think "this is the part I'm not sure about," that is the part to bring to **Alexey and Pauline at 90n** — book a call and we will work through it with you. The link is at the bottom of every major section for a reason.

---

## 01. Why the Cycle is as Much for You as for Your Team

Too often, a review cycle is treated as an HR formality—something to run once or twice a year, check off, and forget. The best leaders treat it as a chance to step back and ask a strategic question: **does the team have what it takes to reach company aspirations, and how can we support their development?**

Done well, it creates a culture where growth is planned, fair, and predictable, instead of being reserved for the loudest voices in the room.

A well-orchestrated and thought-out review cycle gives you visibility on three critical things:
1. **The Skill Map:** The skills you have in the organization and the ones you are currently missing.
2. **Informal Networks:** The hidden organizational chart demonstrating who carries critical knowledge, who might be a bottleneck, and which people work well together.
3. **Holistic Org Health:** A comprehensive view of expected coverage and performance distribution across levels.

### What the research adds to this

The instinct that a review is "for you, the leader, as much as for your team" is well supported. Decades of work on **organizational justice** (the field's foundational synthesis is Colquitt's 2001 meta-analytic review of 25 years of justice research, covering ~64,000 participants) shows that how people experience the *process* of being evaluated predicts their commitment, citizenship behavior, and intention to stay—often more strongly than the *outcome* they receive. In plain terms: a fair, transparent, well-run cycle buys you loyalty and discretionary effort. A sloppy or opaque one quietly erodes both, even when the ratings themselves are accurate.

This is also where most founders underestimate the stakes. Gallup's long-running engagement research attributes roughly **70% of the variance in team engagement to the manager**, and engaged employees are dramatically less likely to be job-hunting within the year. A review cycle is one of the highest-leverage moments you have to either build or burn that manager-employee relationship at scale. It is not a back-office task. It is org design.

> Most founders read this section and nod, then run the cycle the way they always have. The difference between "I understand why this matters" and "my cycle actually delivers org health" is design and facilitation. That is the conversation to have with **Alexey and Pauline** before your next cycle — [book a call](https://cal.com/pauline-bertry/people-ops).

---

## 02. If You're Going to Run Reviews, Make Them Worth It

Here are the six key practices that make the difference between a cycle that drains everyone's energy and one people look forward to.

| # | Practice / Focus | What We See (Mistake) | What to Do Instead (Best Practice) |
|---|---|---|---|
| **01** | **Assume people will use AI** | Even when you explicitly discourage it, people will use AI to prepare their self-reflections and growth memos. The risk is that without guidance, the output will be generic, leaving both sides confused: the person reading the self-reflection, and the person reading the feedback memo. | Give people guidance on using AI meaningfully: to help articulate things they would otherwise struggle to put into words, to summarize feedback conversations, and to prepare for difficult discussions. Set up a prompt, share an example of good output, and set the expectation clearly: AI should be used to prepare for human conversations, not to replace them. |
| **02** | **Keep feedback and reviews separate** | Feedback and the review cycle get collapsed into the same meeting. What should have been day-to-day feedback is delivered a year later, and what should have been a strategic, forward-looking conversation becomes a heavy backward-looking assessment. People feel judged on things they were never told about. | Keep them separate by design. Feedback is continuous: tied to real work, given quickly, specific to a moment. The review cycle is strategic: it looks at patterns over time. When people receive regular feedback throughout the year, the review conversation becomes lighter, more forward-looking, and easier for everyone in the room. |
| **03** | **Run the cycle for the whole company at once** | Reviews happen on individual start dates, or whenever there is time. Calibration becomes almost impossible. One person's "strong performer" is another's "meets expectations", and no one ever compares notes. | Pick two fixed moments a year and have everyone move through the cycle together. That makes ratings consistent, promotion decisions defensible, and the process something people trust. It also creates a shared moment when the organization pauses to reflect. |
| **04** | **Build and communicate shared criteria** | Promotions happen when someone asks loudly enough, when it feels like the right time, or, worse, when someone threatens to leave. There is no shared understanding of what "senior" means, and people who don't get promoted have no idea why. | A simple growth matrix, even if it's only three levels, gives everyone a shared reference point. It does not need to be perfect, but it does need to exist. When the criteria are explicit, growth conversations become easier to run, to hear, and to act on. |
| **05** | **Use structure to protect your culture** | When you have built something informal and human, adding process can feel like a threat, so it gets delayed. Meanwhile, the best people start wondering what their next chapter here looks like, and they eventually stop waiting once they're offered a better title elsewhere. | Build cultural expectations into the career matrix so that everyone can advocate for them: how people treat each other, how they communicate with clients, which information gets shared and which stays internal. The matrix is the best place to make culture explicit, especially if you want to protect it as you scale. |
| **06** | **Design for fairness** | Every review process has bias in it. When you run calibration alone, or without a shared framework, the people who get recognized are often those who are most visible, most vocal, or most similar to whoever is making the decisions, not necessarily the best people. | Name the biases before the cycle starts: recency, affinity bias, halo effect. Run calibration with more than one person in the room: different perspectives will help surface different things. Use a written memo to force reasoning into words. |

### The research behind each practice

**On assuming people will use AI.** The deeper issue is not that AI writes generic prose; it is that generic prose hides the specific, contextual stories that make feedback useful. The classic finding from Kluger and DeNisi's meta-analysis of feedback interventions (a review of 607 effect sizes and ~23,600 observations) is that feedback works *least* well when it pulls attention toward the self ("am I a good designer?") and *most* well when it stays on the task ("this layout buried the primary CTA"). Generic AI output drifts straight to the self-level abstraction. Your prompts should force the opposite: concrete moments, concrete decisions, concrete outcomes.

**On keeping feedback and reviews separate.** Kluger and DeNisi's most cited and most uncomfortable result is that **over a third of feedback interventions actually *decreased* performance**. Feedback is not automatically good; poorly timed or self-directed feedback can demotivate. The practical implication is that a once-a-year data dump of accumulated grievances is close to a worst case. Continuous, task-specific feedback throughout the year, with the formal cycle reserved for pattern-level and forward-looking conversation, is the design the evidence supports.

**On running the cycle all at once.** This is fundamentally about making **calibration** possible, and calibration is where the measurement problem gets fixed. Which brings us to the single most important number in this entire field.

**On shared criteria and fairness.** Bias in evaluation is not a character flaw in your managers; it is a predictable property of unstructured judgment. The behavioral-economics literature (Iris Bohnet's *What Works: Gender Equality by Design* is the accessible synthesis) makes the case that you cannot reliably de-bias people's minds, but you *can* de-bias the process around them—structured criteria, comparative evaluation, written reasoning, and more than one evaluator in the room. Every one of the six practices above is an instance of that principle.

> The six practices are easy to list and hard to install without disrupting the work. Sequencing them for *your* size, *your* services, and *your* culture is the design judgment we bring. **Alexey and Pauline** have built this for 100+ person teams and for ten-person studios — [book a call](https://cal.com/pauline-bertry/people-ops).

---

## 02b. The Measurement Problem: Why "Just Rate People" Doesn't Work

If you take one research finding into your next cycle, make it this one.

When Scullen, Mount, and Goff analyzed how managers actually rate people (two large datasets of ~2,350 and ~2,142 managers, each rated by multiple sources), they decomposed the variance in ratings into its sources. The result is one of the most replicated and most ignored findings in the field: **idiosyncratic rater effects accounted for roughly 50–62% of the variance in ratings.** The actual performance of the person being rated accounted for only about **21%.**

Read that again. When a manager gives someone a rating, more of that number reflects the *rater* — their personal standards, their leniency or strictness, what they happen to notice — than the *ratee*. Every rater carries their own internal yardstick. Some are tough graders; some are generous; most are wildly inconsistent across dimensions. This is not a training gap. It is a structural property of human judgment, and it shows up whether the rater is a boss, a peer, or a direct report.

This is why "we'll just have managers rate everyone" is not a plan. A single manager's raw rating is mostly noise about that manager. The entire architecture of a good review cycle — multiple inputs, shared criteria, written justification, and above all calibration — exists to extract the ~21% signal from the ~60% noise.

### What actually reduces the noise

Three design moves, all evidence-backed:

1. **Multiple raters, structured.** More perspectives dilute any single rater's idiosyncrasy — but only if their input is structured against shared criteria. Unstructured 360s often just multiply the noise.
2. **Behavioral anchors instead of trait scales.** Rating "communication: 3/5" invites every rater's private definition of communication. Rating against specific observable behaviors ("gives feedback that leaves the person knowing exactly what to change") narrows the target.
3. **Calibration meetings.** This is the heavy lifter. Research published in *The Accounting Review* found calibration reduced cross-manager rating variance by up to **40%**. When managers must defend a rating out loud against a shared standard, the idiosyncratic component gets argued down toward the real signal.

A caution that the research is equally clear about: calibration can *introduce* bias if it is run badly — the loudest or most senior voice anchors the room, and quieter managers defer. (Harvard Business Review has documented exactly this failure mode.) Structure protects you here too: written memos submitted *before* the discussion, a facilitator who is not the most senior person, and an explicit norm that every rating must be justified against the matrix, not against a personality impression.

> Calibration is the single highest-leverage hour in your cycle and the easiest to get wrong. A neutral, experienced facilitator is the difference between calibration that removes bias and calibration that launders it. This is the room where **Alexey and Pauline** add the most value — facilitating the conversation so your leadership aligns on the bar without anchoring on the loudest voice. [Book a call](https://cal.com/pauline-bertry/people-ops).

---

## 03. Month-by-Month Preparation & Implementation Calendar

To make sure your review cycle runs smoothly without taking over your calendar, start preparation early. Here is the month-by-month timeline:

- **April & May: Quiet Months**
  - Monitor day-to-day feedback loops. No active cycle actions needed.
  - *Research note:* this is the window where continuous feedback should be doing its work. The evidence that companies running real-time feedback outperform peers comes from the *quality of the year-round loop*, not the formal cycle. If feedback is only happening in the quiet months by accident, that is the first thing to fix.
- **June: Set Up the Cycle**
  - Decide your process (self-reflection, peer feedback, manager input).
  - Confirm your career framework, build a simple one if missing.
  - Identify who will run the growth conversations.
- **July: Build the Templates**
  - Prepare self-reflection memo + feedback form.
  - Run brief conversation on the process with key leaders.
  - *Design tip:* write the self-reflection prompts to demand specific stories, not summaries (see the AI FAQ below). The prompt design is where the cycle is won or lost.
- **August: Announce the Cycle**
  - Communicate the upcoming cycle to the team with 4 weeks' notice minimum.
  - Share self-reflection prompts, guidelines, and timeline.
  - *Justice note:* advance notice and transparency about the process are not courtesies — they are the procedural-justice levers that drive whether people trust the outcome.
- **September: Collect Input**
  - Kick-off self-reflection preparation.
  - Open feedback collection channels.
  - Collect exposure lists.
- **October: Submit & Calibrate**
  - Collect self-reflections and feedback summaries.
  - Run calibration discussions, including key leaders.
  - Prepare final growth memos.
  - *This is the load-bearing month.* See Section 02b — calibration is what converts noisy individual ratings into something defensible.
- **November: Hold Growth Conversations**
  - Run growth conversations (30–45 min).
  - Announce promotions and celebrate publicly.
  - Share key insights and success stories on all-hands.
- **December: Compensation & Close-out**
  - Share a written note with every person: this and next year total compensation and bonus amounts.
  - *Critical design choice:* this is deliberately separated from November's growth conversation. See Section 03b on why mixing the two damages both.

### 03b. Why compensation comes last, and on its own

The calendar above splits the *growth conversation* (November) from the *compensation note* (December) on purpose, and the research on this is unusually consistent.

When you mix "here is how you're growing and what to work on" with "here is your raise," the money dominates the room. Practitioners and neuroscience-informed work on performance management describe compensation talk as triggering a near-physiological threat response that shuts down the part of the conversation where actual learning happens — people stop hearing developmental feedback the moment a number is on the table. The widely cited practitioner guidance is to space the two conversations at least a month apart, which is exactly what the calendar does.

This is also part of the broader, well-documented industry shift. Deloitte's research on performance management redesign found a large share of companies moving toward continuous feedback, coaching, and the explicit *separation of feedback from pay* — and major firms (Microsoft, GE, Adobe) publicly abandoning forced-ranking "rank and yank" systems after concluding they damaged collaboration and morale. A recent study in *Management Science* on forced ranking in a multinational firm found it drove measurable underrecognition and talent loss. The direction of travel in the evidence is clear: rank people against a fixed bar of expectations, not against each other, and keep the pay conversation in its own room.

> The calendar looks simple on a page and gets chaotic in practice — input arrives late, calibration runs long, the comp note slips into the growth talk. Running the first cycle *with* you, on the real calendar, is exactly what we do so it becomes a habit you own. Talk to **Alexey and Pauline** about timing your cycle — [book a call](https://cal.com/pauline-bertry/people-ops).

---

## 04. Career Growth Matrix (Framework Design)

A growth matrix is a living document that defines what is expected at each level, what looks like progress, and how someone earns their next level.

### Core Design Rules:
1. **Describes behaviors, not traits or expectations:** A useful matrix describes what someone *does*, not what they *are*. "Communicates clearly with clients" is actionable; "good communicator" is not.
2. **Tied to how your agency actually operates:** Avoid copying a tech company's matrix. Ensure the dimensions and language reflect your services, clients, and value creation model.
3. **Built with the future in mind:** Reflect your team's day-to-day work while accommodating active strategy (e.g., incorporating AI workflows).

### Why "behaviors, not traits" is a research finding, not a style preference

This rule is doing more work than it looks. It is the practical fix for the two biggest evidence-based problems in evaluation at once.

First, the **idiosyncratic rater effect** (Section 02b): trait words like "good communicator" or "strong leader" invite every rater to apply their own private definition, which is precisely how 60% of rating variance ends up being about the rater. Observable behaviors give every rater the same target.

Second, **demographic bias in feedback.** This is one of the most striking research areas of the last decade. Analyses of real performance feedback (Textio's large-scale studies of workplace feedback, and Stanford sociologist Shelley Correll's work on the language of evaluation) found that women and people of color receive far more *vague, personality-based* feedback and far less *actionable, work-based* feedback. Textio found women were substantially more likely to be described as "emotional" or "unlikeable"; Correll's research showed men more often get specific, developmental, "do X to get promoted" feedback while women get vaguer critique tied to demeanor. One striking figure: for every piece of unactionable feedback a white man under 40 received, women over 40 received several times as much.

A behavioral matrix is the structural defense against this. When the criteria are concrete and observable, "abrasive" and "not a culture fit" are no longer admissible — the question becomes "which behavior at which level did or didn't happen," and that is a question you can answer fairly across everyone.

### Behavioral Matrix Snippet (Designer Track):

| Dimension | Mid Designer (L2) | Senior Designer (L3) | Lead Designer (L4) |
|---|---|---|---|
| **Craft** | • Applies interaction and visual design skills independently on assigned work.<br>• Uses AI tools to speed up early-stage exploration. | • Leads research, prototyping, and design phases.<br>• Knows design systems deeply.<br>• Applies AI and data tools to test assumptions faster. | • Sets design direction across multiple tracks.<br>• Guides the team on where AI improves workflows and where human judgement matters more. |
| **Client Impact** | • Sees how own work connects to the client goal.<br>• Delivers work that holds up to brand standards without heavy rework. | • Identifies the right problems to solve, including those beyond what is explicitly assigned.<br>• Makes sure output is tied to what the client actually needs. | • Knows the client's business well enough to see what they will need before they ask.<br>• Aims to be someone the client invites into conversations beyond the current project. |
| **Leadership** | • Shares useful things learned with peers.<br>• Helps new hires get productive by guiding them through tools and how the team works. | • Runs regular check-ins with juniors.<br>• Gives feedback that is specific and actionable: people leave the conversation knowing exactly what to work on. | • Sets clear growth expectations for each team member.<br>• Creates stretch opportunities.<br>• Team members visibly grow in skill and confidence over time. |

### Designing goals that actually move performance

Once the matrix tells people *what good looks like at their level*, the goals you set with them should tell them *what to push on this cycle*. Locke and Latham's goal-setting theory is one of the most robustly supported findings in all of organizational psychology, built on hundreds of studies over decades. The core results, in agency-usable form:

- **Specific, challenging goals beat "do your best."** In Locke's studies, specific and difficult-but-attainable goals outperformed vague encouragement roughly 90% of the time. "Do your best" gives people no target to calibrate effort against.
- **Difficulty helps — until it doesn't.** Performance rises with goal difficulty in a near-linear way *within the limits of ability*. Set goals beyond plausible reach and the correlation collapses; people disengage. The art is the stretch that is hard but believable.
- **Goals require feedback to work.** A difficult goal with no progress feedback is just pressure. The goal and the year-round feedback loop are a single system, not two separate things.

This is why the matrix and the goals and the continuous-feedback habit all have to be designed together. Pulled apart, each one underperforms.

> A matrix that reads well but doesn't fit how your studio makes money is worse than no matrix — it formalizes the wrong things. Pauline has built craft-leveling and designer-progression frameworks tested on 100+ person product teams; this is the part of the work she most enjoys doing *with* founders rather than handing over. [Book time with Pauline and Alexey](https://cal.com/pauline-bertry/people-ops).

---

## 05. About 90n: What We Do & Our Philosophy

We are **90n**, a boutique people operations and organizational design firm built specifically for professional service organizations—creative agencies, digital boutiques, design studios, and product consultancies.

### Why We Built 90n
Most people growth systems are designed for large corporations or venture-backed tech companies. They are heavy, bureaucratic, and rely on full-time HR departments to run. When agency leaders try to copy these frameworks, it feels unnatural, takes too much time, and drains the creative energy of the studio.

We do things differently. We build systems that are lightweight, lovable, and deeply aligned with how agencies actually make money and grow talent. We strip away corporate HR heritage and replace it with practical, high-impact workflows—but we keep the parts that the evidence says actually matter (calibration, behavioral criteria, procedural fairness, the feedback-pay separation) and throw out the parts that are corporate theater.

### Our Core Principles
1. **We build, not just advise:** We don't just hand you a 60-page PDF report and walk away. We co-design the system, run the first cycles with you, train your managers, and ensure you own the entire setup when we are done.
2. **We design for real life:** We build career frameworks that make sense for how your team operates day-to-day, mapping skills directly to client outcomes.
3. **We automate what we can:** Any administrative step that can run without manual effort, will. We make sure you spend time in meaningful human conversations, not drowning in spreadsheets.
4. **We start small on purpose:** We run pilot programs first so that your team builds trust and confidence before building a permanent habit.
5. **We design with the evidence:** Every recommendation we make is anchored in the research — feedback science, organizational justice, goal-setting theory, and the measurement literature — translated into something a ten-person studio can actually run.

### Our Core Services
* **Career Paths & Review Cycles:** Designing clear progression frameworks and structured review cycles that help your best talent understand their future with you, reducing expensive attrition.
* **Onboarding & People Engagement:** Designing premium first-week experiences and culture-aligned systems that make newcomers productive and engaged from day one.
* **Upskilling & Leadership Workshops:** Running highly practical management upskilling workshops for team leads and directors, designed specifically for the pace of agency life.
* **Calibration Facilitation:** Sitting in (or running) your calibration sessions as a neutral facilitator so the room aligns on the bar instead of deferring to the loudest voice.

---

## 06. Meet the Founders & Book a Discovery Call

90n was founded by two ex-McKinsey consultants who wanted to bring top-tier people growth practices to the agency world without the corporate fluff.

### Pauline Bertry (Product Growth & CX Design)
* **Background:** 10+ years leading product and design teams. Built from scratch and led Design Hubs at McKinsey Moscow and Budapest. Created career frameworks and growth systems tested with 100+ person cross-functional product teams.
* **How Pauline helps you:** Pauline specializes in craft leveling, designer progression, workspace operations, and translation of creative expertise into concrete career matrices.
* **Book a Discovery Call with Pauline:** [https://cal.com/pauline-bertry/people-ops](https://cal.com/pauline-bertry/people-ops)

### Alexey Lobachev (People Strategy & Engagement)
* **Background:** 15+ years inside top-tier professional services firms. At McKinsey, led Employee Engagement and Talent Management programs, including competencies reviews, region-wide DEI transformations, and Top Talent Retention Programs.
* **How Alexey helps you:** Alexey focuses on organizational health diagnosis, incentive designs, leadership upskilling, and custom review cycle calibration strategy.
* **Book a Discovery Call:** [Schedule a Call via Pauline's Booking Link](https://cal.com/pauline-bertry/people-ops)
* **Connect with Alexey on LinkedIn:** [https://www.linkedin.com/in/alexey-lobachev-tor/](https://www.linkedin.com/in/alexey-lobachev-tor/)

> Most of what is in this guide, Alexey and Pauline have built and run inside real organizations — including the calibration rooms, the bias conversations, and the matrices that survived contact with actual promotion decisions. If you are about to run a cycle, the highest-value thing you can do is spend 30 minutes with them first. [Book a call](https://cal.com/pauline-bertry/people-ops).

---

## 07. Strategy FAQ: The Founder's Guide to Agency Performance

### Q: What is the most interesting thing about performance reviews?
**A:** The most interesting (and overlooked) thing about a performance review cycle is that **it is not actually about evaluating the past. It is a diagnostic map of your agency's present and future organizational health.**

When you run a review cycle properly, you are tapping into the invisible currents of your studio. You gain visibility on three unexpected dimensions:
1. **The Skill Map Delta:** You see which skills you have in abundance and where you have gaps. If three people note that they feel weak in visual accessibility or brand storytelling, you have an immediate upskilling opportunity before a client notices a deficiency in your work.
2. **The "Underground Org Chart" (Informal Networks):** Through peer nominations and "exposure lists," you see who is actually holding the studio together. It is rarely just the loud managers. Often, it is a quiet Mid-level Designer or Developer whose peer feedback reveals they are unblocking five other people every week. This person is a high-leverage asset and a massive flight risk if they are overloaded and under-recognized.
3. **Calibration as Strategic Alignment:** The calibration meeting—where leaders sit together to align on ratings—is the most valuable hour in the entire process. When you debate whether someone's performance is "On-Track" or "Advanced" relative to L3 expectations, you are not just grading. You are aligning on your agency's creative bar, your expectations of quality, and your core values. It forces leadership to get on the same page about what "good" looks like in writing.

### Q: Is the move away from annual reviews just a fad?
**A:** No — but it is widely misunderstood. The research does not say "kill the annual cycle." It says **stop relying on the annual cycle to do the job of continuous feedback.** Kluger and DeNisi's meta-analysis showed feedback can hurt as easily as help, especially when it is delayed and self-directed; a once-a-year accumulation of feedback is close to that worst case. The companies that outperform are the ones running tight, task-specific feedback loops *all year*, and reserving the formal cycle for pattern-level and forward-looking decisions: leveling, promotion, and development planning. So you keep a formal cycle (you need it for calibration and fairness), but you stop pretending it is where feedback happens.

### Q: We're tempted to rank people on a curve / forced distribution. Should we?
**A:** Almost certainly not, especially at agency scale. The major firms that pioneered forced ranking — Microsoft, GE, Adobe — publicly walked away from it after concluding it damaged collaboration and morale, and recent peer-reviewed work (a *Management Science* study on forced ranking in a multinational) found it caused real underrecognition and talent loss. The deeper problem: forced ranking pits people against each other, which is poison in a studio where the work is collaborative and a single great senior can lift a whole team. Rank people against a **fixed bar of expectations** (your matrix), not against each other. This is exactly the kind of decision worth pressure-testing with Alexey and Pauline before you bake it into a cycle.

### Q: What opportunities do we get when we do a performance review for you?
**A:** When 90n partners with your agency to design and run a performance review, it is not an administrative exercise; it is a strategic intervention. Here are the core opportunities we unlock together:

1. **Stem Talent Attrition (Retaining Your Best People):** The #1 reason high performers leave agencies is a lack of growth clarity. They feel they've hit a ceiling or that promotions only go to the people who complain the loudest or threaten to resign. By establishing a clear, behavioral Growth Matrix, we give everyone a predictable path forward. They can see their next chapter within your agency, so they don't look for it elsewhere. (Gallup's research is blunt here: managers drive ~70% of the variance in engagement, and people who feel someone is invested in their development are far more likely to still be there a year later.)
2. **Upskill Your Managers (Developing True Leaders):** Most agency team leads are promoted because they are great individual contributors (great designers, writers, or engineers), not because they know how to coach people. Designing and calibrating a review cycle provides them with structured "training wheels" on how to deliver difficult feedback, guide growth conversations, and set objectives. We turn your senior practitioners into strong, respected leaders.
3. **Uncover Hidden Burnout & Resignation Risks:** We design peer exposure surveys that highlight who is holding up the team, who is currently working in silos, and who is carrying an unsustainable share of the studio's cognitive load. We spot the key contributors who are approaching burnout months before they hand in their resignation.
4. **Build a Defensible Compensation & Promotion Model:** When review results are calibrated against a clear matrix, promotion decisions become objective and defensible. You protect your leadership team from awkward negotiations and "fairness" complaints, establishing deep cultural trust. This is the organizational-justice payoff: procedural fairness predicts commitment and retention, often more strongly than the raw outcome.
5. **Zero Friction & High Efficiency:** We handle the system integration, templating, and schedule management. We ensure your cycle runs smoothly and finishes on time, taking the administrative burden completely off your plate so you can focus on what matters most—having rich growth conversations.

### Q: How simple should a growth framework be for a small agency?
**A:** Extremely simple. If you have under 50 people, do not build a complex, multi-track matrix with dozens of dimensions. A simple 3x3 matrix—three core levels (e.g., Associate, Mid, Senior) across three dimensions (e.g., Craft, Client Impact, Leadership)—is more than enough. It should fit on a single page or screen. The goal is clarity and alignment, not complexity. The evidence on goal-setting and justice both point the same way: people need a target they can actually see and understand, not an exhaustive taxonomy. Start with the 3x3, run a cycle, and let it earn its complexity.

### Q: How do we make sure people don't use generic AI-generated reflections?
**A:** By changing the prompts. If you ask generic questions like "What did you accomplish this year?" people will copy-paste bullet points into ChatGPT and get corporate buzzwords in return. Instead, ask for specific, contextual stories: *"Describe a moment this quarter where you had to make a trade-off between speed and craft quality. What was the impact on the client?"* AI cannot invent your real studio stories. There is also a research reason this works: feedback (and self-reflection) is most useful when it stays at the task level rather than drifting to self-evaluation — exactly the level where generic AI prose lives. Also, instruct your team explicitly on *how* to use AI: as a sparring partner to help refine their thoughts, not a replacement for their voice.

### Q: Why should we run the cycle for the whole company at once rather than on individual work anniversaries?
**A:** Because running reviews on individual work anniversaries makes **calibration and comparison impossible**. You cannot compare a designer's growth in February with another's in October. And without calibration, you are stuck with raw individual ratings — which, per Scullen and Mount, are roughly 60% about the rater and only ~20% about the person. Running everyone through the cycle together twice a year creates a shared moment of organizational reflection, lets you calibrate the noise out, makes promotion budgets manageable, and builds trust because the whole team sees that everyone is evaluated against the same bar at the same time.

### Q: How do we keep bias out of the process?
**A:** You don't keep it out of people's heads — the behavioral-economics evidence (Bohnet and others) is clear that debiasing individual minds is extremely hard. What works is debiasing the *process*: behavioral (not trait) criteria, written justification submitted before discussion, comparative rather than one-at-a-time evaluation, more than one evaluator in the room, and a calibration facilitator who isn't the most senior person. Name the specific biases out loud before you start — recency (over-weighting the last month), halo (one strength coloring everything), affinity (rating people like yourself higher), leniency and central-tendency. The demographic-bias research (Textio, Correll) is a reminder of the stakes: vague personality feedback lands disproportionately on women and people of color, and a behavioral matrix is your structural defense. Running a clean calibration is genuinely hard the first time — it is the thing founders most often ask Alexey to facilitate.

### Q: Should performance ratings drive pay directly?
**A:** Be careful. The growth conversation and the compensation conversation should be **separated in time** — practitioner and neuroscience-informed guidance is to leave at least a month between them, because money triggers a threat response that shuts down the developmental conversation. Pay obviously *relates* to performance, but the moment you fuse "here's how to grow" with "here's your number," people stop hearing the growth half. Calibrate and discuss development first; communicate compensation as a separate, written, clear note afterward (see the December step in the calendar).

---

## 08. Glossary: The Biases to Name Before You Start

Naming biases out loud at the top of calibration is one of the cheapest, most effective interventions you can make. Here are the ones worth a sentence each before any rating gets discussed:

- **Idiosyncratic rater effect:** the big one. Each rater's personal yardstick accounts for more rating variance than actual performance does. The reason calibration and shared criteria exist.
- **Recency bias:** the last month overshadows the other eleven. Counter it with year-round notes and exposure lists, not memory.
- **Halo (and horns) effect:** one strong (or weak) trait colors every other rating for that person. Counter it by rating each dimension independently against the matrix.
- **Affinity bias:** rating people who resemble you — background, style, communication — more favorably. Counter it with multiple evaluators and written justification.
- **Leniency / severity bias:** some raters run hot, some run cold, regardless of the ratee. Calibration normalizes across them.
- **Central-tendency bias:** clustering everyone at "meets expectations" to avoid hard conversations. Counter it by requiring evidence for the middle rating too, not just the extremes.
- **Gendered / racialized language bias:** vague, personality-based feedback ("abrasive," "emotional," "not a culture fit") landing disproportionately on women and people of color. Counter it by demanding behavioral, work-anchored language.

---

## 09. Ask Your AI (Or Ask 90n) — Collaborative Diagnostics

To kick-start a strategic conversation, copy this guide's content and paste it into your AI assistant (Claude, Gemini, or ChatGPT) along with one of these high-yield diagnostic prompts:

> **PROMPT 1: Agency Growth Diagnostic**
> *"I have uploaded the 90n Performance Review Guide. Based on their philosophy and principles, help me audit my agency's current state. We are a [number of people] person [agency type] agency. Our current challenge is [e.g., junior burnout, promotion negotiations, high attrition, vague roles]. Help me brainstorm what a lightweight, lovable people ops system would look like for us."*

> **PROMPT 2: Simple Growth Matrix Draft**
> *"Based on the Designer matrix snippet in Section 04, help me draft a simple, 3-level growth framework for our [insert role, e.g., Project Managers, Copywriters, or Developers]. What are the behaviors, not traits, that define progress from Level 1 to Level 3 in an agency context?"*

> **PROMPT 3: Calibration Meeting Simulator**
> *"Act as Pauline and Alexey from 90n. I am a founder preparing for a calibration meeting. Let's roleplay. I will present a specific employee performance scenario where a lead manager wants to promote a highly talented but sometimes abrasive designer, and you will guide me on how to calibrate this fairly and behaviorally against our growth criteria."*

> **PROMPT 4: Bias Pre-Mortem**
> *"Based on Section 08 of the 90n guide, run a pre-mortem on my upcoming calibration meeting. Here is my team and a one-line note on each person. Flag where recency, halo, affinity, and gendered-language bias are most likely to creep into our ratings, and give me a script for naming these at the start of the meeting."*

A real diagnostic, though, is a conversation, not a prompt. The AI can help you draft; it cannot facilitate your calibration room, read your team's informal network, or tell you which of your leads is one cycle away from quietly burning out. That is what Alexey and Pauline do.

---

*This guide was generated by 90n People Growth and grounded in the last decade-plus of research on feedback, evaluation, goal-setting, and organizational justice. We help creative and digital agencies scale their talent without corporate overhead. To start co-designing your growth frameworks or review systems, book a call with Pauline and Alexey — that is genuinely the best next step after reading this.*

* **Schedule a Discovery Call:** [https://cal.com/pauline-bertry/people-ops](https://cal.com/pauline-bertry/people-ops)
* **Visit Our Website:** [https://90n.ai](https://90n.ai)

---

### Selected research referenced in this guide
*Light-touch pointers for the curious — none of this needs to be read to run a great cycle, but it's where the claims come from.*

- Kluger, A. N., & DeNisi, A. (1996). *The effects of feedback interventions on performance: a historical review, a meta-analysis, and a preliminary feedback intervention theory.* Psychological Bulletin — the finding that ~1/3 of feedback interventions reduce performance, and that task-level feedback beats self-level.
- Scullen, S. E., Mount, M. K., & Goff, M. (2000). *Understanding the latent structure of job performance ratings.* Journal of Applied Psychology — the idiosyncratic rater effect (~50–62% of rating variance is the rater).
- Colquitt, J. A. (2001). *Justice at the millennium: a meta-analytic review of 25 years of organizational justice research.* Journal of Applied Psychology — distributive, procedural, and interactional justice and their link to commitment and performance.
- Locke, E. A., & Latham, G. P. *Goal-setting theory* (1990 and subsequent reviews) — specific, challenging goals outperform "do your best."
- Bohnet, I. (2016). *What Works: Gender Equality by Design* — debiasing the process, not the person; structured, comparative evaluation.
- Textio (2022) workplace-feedback analyses and Shelley Correll's research on the language of evaluation — demographic skew toward vague, personality-based feedback for women and people of color.
- Research in *The Accounting Review* on calibration reducing cross-rater variance (~up to 40%); HBR on how calibration can also introduce bias if run badly.
- Deloitte performance-management research and a *Management Science* study on forced ranking — the industry move away from rank-and-yank and toward separating feedback from pay.