You've hit a wall you weren't expecting. The company is growing. Headcount is up, revenue is climbing. But somewhere around 50 people, the way you've been handling performance stopped working.
Managers are inconsistent. Feedback is random. Your top performers don't know where they stand. And nobody wants to have the hard conversation about the employee who's clearly not cutting it anymore.
This is the scaling PM problem. You need more rigor than a startup's gut-feel approach, but you're not a 2,000-person enterprise with a full HROD function. This guide covers exactly that gap.
When to formalize performance management
Most companies formalize too late. They wait until the informal system breaks visibly (usually a bad attrition event or a legal near-miss) before building structure.
Here's what to watch for:
Size signals
- 50+ employees: One person can no longer observe everyone directly. You need documented reviews.
- 75+ employees: Managers have their own managers. Consistency becomes a real problem.
- 150+ employees: Dunbar's number. Individual familiarity breaks down completely. Process must carry the load that relationships used to.
Behavior signals (more reliable than headcount)
- You've had a compensation decision that felt arbitrary. Two similar performers got different raises, and you had no good explanation
- A manager gave someone low ratings that felt retaliatory, and you had no documentation to push back
- You lost a high performer who cited lack of clarity about their growth trajectory
- You're making a promotion decision based on gut feel and realizing you can't explain it to the person who didn't get promoted
If you're seeing two or more of these, formalization is already overdue.
What "formalized" actually means at this stage
It doesn't mean quarterly 360-degree reviews with competency matrices and nine-box grids. It means:
- Everyone gets written feedback at least twice a year
- Ratings (or equivalent) are calibrated across managers, not just assigned in isolation
- There's a documented connection between performance and comp decisions
- Managers have a consistent framework for performance conversations (even if they run the meetings themselves)
That's the minimum viable PM system for a 50-100 person company.
Building your first calibration process
Calibration is where most scaling companies stumble. You run your first formal reviews, collect manager ratings, and then realize that your engineering VP rates everyone a 4 out of 5 while your sales VP rates everyone a 2. Neither of them is wrong given how they're interpreting the scale.
Without calibration, your performance data means nothing.
The pre-calibration question
Before you run calibration sessions, you need to answer: calibrating what?
Options:
- Rating distributions: Are managers using the scale consistently? (Most common starting point)
- Promotion readiness: Is this person actually ready to move up, or is this manager advocating based on relationship?
- High performer identification: Who are the top 10-15% we need to retain at all costs?
- Performance concerns: Who is struggling, and does every manager have the same threshold for flagging concerns?
Pick one focus for your first calibration. Trying to do all four at once produces a four-hour meeting where nothing gets decided.
Running a calibration session
A basic calibration for a 50-100 person company looks like this:
Before the session
- Collect manager ratings and written justifications (not just numbers)
- Build a simple spreadsheet: name, manager, role, rating, one-sentence justification
- Flag outliers: anyone two or more steps from their peer group's average
During the session (60-90 minutes)
- Start with the outliers, not the middle
- For each outlier: manager presents rationale, group asks questions, group decides whether to adjust
- No individual discussions about people not in the outlier list (this is where sessions go long)
- HR or whoever is facilitating tracks all adjustments in real time
After the session
- Update the final ratings before they go to employees
- Document the calibration decisions (not publicly, but as an internal record)
- Brief managers on any adjustments to their ratings before employees see them
Common calibration mistakes at this stage
Skipping written justifications. Numbers without context are calibratable. But they produce bad calibration. Managers argue about feelings rather than evidence. Requiring one paragraph per rating adds maybe 20 minutes per manager and makes calibration twice as useful.
Calibrating every single person. You have 80 employees and 8 managers. If each manager brings 10 people to calibration and you discuss each one for 5 minutes, that's 400 minutes. Don't do this. Focus calibration on people who are being rated differently than their peers, people being considered for promotion, and people on performance plans.
No follow-through. Calibration decisions have to actually make it into the final ratings. If managers feel like calibration can be overridden or ignored, they'll stop taking it seriously.
Choosing performance management software
At 50-100 employees, you probably don't need dedicated PM software yet. A well-designed Google Form + spreadsheet system can work.
At 150+ employees, you need software. The manual overhead becomes unsustainable, and the data integrity issues start causing real problems.
What to actually evaluate:
The four things that matter
1. Review administration overhead How much work does it take to launch a review cycle? Some platforms require days of configuration per cycle. Some take 30 minutes. If HR is a team of one, this matters enormously.
2. Manager experience Your reviews are only as good as your managers' willingness to complete them. Platforms that are confusing or require too many clicks will produce rushed, low-quality responses. Run a manager test before you buy.
3. Calibration tooling Basic platforms don't have this. Mid-tier platforms have a view that shows all ratings side by side. Good platforms let you run a calibration session in the tool, track adjustments, and maintain an audit trail.
4. Data you can actually use After you run a cycle, can you answer: Who are your top performers by department? What's the distribution of ratings by manager? How have individual ratings changed over time? If the answer is "we'd have to export to a spreadsheet," you're going to use the data about once a year.
What to skip at this stage
- 360-degree feedback collection unless you have a specific need. Multi-rater feedback adds significant administrative complexity and is hard to action without training. Most scaling companies don't have the manager capacity to run 360s well.
- OKR/goal-tracking integration if your goal-setting process isn't already consistent. Don't solve two problems at once.
- HRIS integration as a buying requirement. Nice to have, not worth paying significantly more for until you're 250+ employees.
The honest sizing guide
| Company size | What you probably need |
|---|---|
| 50-100 employees | Structured templates, consistent process, no special software required |
| 100-200 employees | Dedicated PM software for review administration and calibration |
| 200-500 employees | PM software with calibration tooling, analytics, and some talent workflow support |
| 500+ employees | Enterprise platform with HRIS integration and advanced analytics |
Scaling reviews across departments
At 75 employees, you might have 3-4 functional areas. At 300 employees, you have 10-15. The challenge isn't just logistics. Different functions have fundamentally different ideas about what good performance looks like.
The consistency vs. relevance tension
A software engineer and an account manager are not doing the same job. Evaluating them on the same competencies with the same rating scale leads to meaningless comparisons.
But if every department has completely different criteria, you can't compare people across departments for calibration or promotion decisions.
The answer is a two-layer model:
Company-wide criteria (2-3 things) These apply to everyone, regardless of role. Typically: results against goals, how they work with others, and some version of values alignment. Keep this layer thin.
Role-specific criteria (3-5 things) These vary by function or job family. Sales has pipeline management and close rate. Engineering has technical contribution and code quality. Marketing has campaign execution and data literacy.
This gives you comparable data at the company level while making the feedback useful at the individual level.
Building the process before the software
One of the most common mistakes: companies buy PM software and then try to design their process inside the tool.
Design the process first, on paper:
- How many times per year? (Start with two.)
- Who reviews whom? (Manager-only to start; add peers later.)
- What questions will managers answer?
- What questions will employees self-assess?
- What happens after reviews are submitted? (Calibration, comp decisions, development conversations?)
Then buy software that fits the process. Not the other way around.
Rollout sequence across departments
Don't try to launch a new PM process across the whole company at once. The failure mode: three of your ten managers don't complete reviews on time, which breaks the calibration session, which delays comp decisions, which causes a mess.
Sequence by department size and manager readiness:
- Start with your most organized department and a manager who wants this to work. Run the first cycle with them.
- Run a retrospective. What broke? What was confusing?
- Fix those things. Then add two more departments in the second cycle.
- By cycle three, you're company-wide.
This is slower than doing everything at once. It produces a working system instead of a chaotic launch you spend two years recovering from.
The manager problem
Every PM system design conversation eventually hits the same wall: the process is only as good as the managers running it.
At scaling companies, this is especially acute. You've promoted your best individual contributors into management. They're competent at the work; they're often not equipped to have performance conversations.
Three conversations managers avoid
The "you're below expectations" conversation. Most managers avoid this until the situation is critical. By then, the employee is usually blindsided, and the documentation to support a performance plan often doesn't exist.
The "you're doing fine, but not great" conversation. Easy to avoid by giving everyone a 3 or "meets expectations" and saying nothing. This destroys the meaning of the rating scale over time.
The "here's why you didn't get promoted" conversation. Managers hate delivering this, so they give vague answers, which leaves employees confused and sometimes bitter.
None of this gets fixed by better software or cleaner process design. It gets fixed by training managers on these specific conversations and holding them accountable for doing them.
Minimum viable manager training for PM
You don't need a two-day workshop. You need:
- One session before each review cycle (45 minutes): What are we measuring, how to write useful feedback, common rating errors to avoid
- A few written examples of good feedback vs. bad feedback for each role level
- A debrief after calibration for any manager whose ratings were significantly adjusted
The investment is small. The difference in review quality is large.
Common questions at this stage
Should we use ratings at all?
Yes, if you're making any compensation decisions based on performance. The "no ratings" movement makes sense in companies where reviews are entirely developmental. If comp is involved, you need some kind of calibratable measure, and ratings are the most practical one.
How often should we run reviews?
Start with twice a year. Annual reviews don't provide enough data points for good calibration and feel high-stakes to employees. Quarterly reviews at 100+ employees create so much administrative load that quality drops. Twice a year (typically January and July, or February and August) is the sweet spot.
What about continuous feedback tools?
Useful supplements. Not replacements for structured reviews. Continuous feedback tools work best after you have a solid review process, not before. The data from continuous feedback is hard to act on without the structure of a periodic review cycle.
When should we add peer feedback?
Add it when two conditions are true: managers have enough time to actually read and synthesize peer input, and employees have enough context about each other's work for peer feedback to be meaningful. At 75 employees with no defined peer groups, peer feedback often produces low-information observations. At 150+ employees with clear team structures, it adds real signal.
What good looks like at different stages
50-100 employees
- Two structured review cycles per year
- Manager ratings + calibration session (even informal)
- Written documentation for all performance decisions
- Clear criteria for what "meets" vs. "exceeds" vs. "below" expectations means
100-200 employees
Everything above, plus:
- PM software handling review administration
- Formal calibration session with decision documentation
- Performance data connected to compensation decisions (transparent criteria, not just outcomes)
- Manager training before each cycle
200-500 employees
Everything above, plus:
- Two-tier competency model (company-wide + role-specific)
- Talent segmentation (who are the people you'd fight to keep?)
- Development plans tied to review outcomes
- Regular reporting on performance distribution by department and manager
Getting started
The biggest obstacle at scaling companies isn't not knowing what to do. It's organizational inertia. The current system is imperfect but familiar, and changing it requires energy that's already spread thin.
The practical first step: pick one thing to formalize in your next review cycle that you haven't formalized before. Just one. If you have no calibration at all, run a single 60-minute calibration session after the next cycle. If your ratings are inconsistent, define what each level means before the next cycle starts.
Build one layer at a time. You'll have a real system by the time you hit 300 people—without a massive redesign project.
Confirm helps scaling companies run performance management that's fair, consistent, and actually defensible when comp decisions get questioned. If your next review cycle is coming up, see how it works.
