Org Constitution Article 5: Perf

This continues The Organizational Constitution started here & previous Article 4: Managers here .

Intro

At its worst, performance management is a major source of inequity, angst, and wasted time. Improving objectivity in performance management can greatly enhance fairness, and few areas have as profound an impact on employee satisfaction as these decisions. For example, fairer promotion decisions have been shown to double the chances of employee engagement [1 ].

The right incentives shape the right outcomes, making performance management a crucial lever for enhancing both work quality and output quality. A fair and objective approach rewards substance over style and combats the politicky/salesy types who mask objectively worse means and ends.

A caveat: I’m not an HR professional. Based on my experience and feedback from others, even systems set up by HR experts at top companies like Google are unsatisfactory. This article is born from ample observations and represents hypotheses I’d like to try. I’ll keep experimenting till I find something that works well, because the solutions I’ve encountered so far are quite terrible.

Section 1: Evaluations

Instead of heavy performance cycles once or twice a year, aim for lightweight rolling performance evaluations, what I’m terming "360 drip evaluation." These are done throughout the year, ensuring people aren’t just optimizing for performance cycles. The inputs and outputs are designed to be super lightweight, minimizing burden and noise.

Inputs

1. How They Achieve: Deploy surveys to assess a person’s methods of achievements - collaboration, skills, and values. Honoring the right means is essential to upholding organizational values. Send primarily quantitative surveys weekly or monthly to a random sample of coworkers the subject has most engaged with, based on communication and meeting data. They should each take less than 2-5 minutes to fill out. Additionally, managers fill out surveys of their reports monthly to track their quantitative and optionally qualitative assessments over time.
2. What They Achieve: Evaluate two types of achievements: work artifacts and impact, both of which are submitted for review on a rolling basis. For work artifacts, evaluate the quality of their craft and usefulness. For impact, evaluate against expectations, the challenges of the circumstances, and relevant comparables. Generate scores for these achievements by sending quantitative surveys to randomly sampled, relevant employees.

Keeping Surveys Fair and Lightweight

Fairness - Algorithmically calibrate a feedback giver’s scores if they skew higher or lower than others, especially direct managers prone to bias. Normalize scores across the board to determine what constitutes good, average, or bad ratings. Randomly sample and anonymize surveys to ensure integrity, minimize bias, and protect relationships.

Lightweight - Use an easy multiple-choice format for various dimensions being evaluated. Opt for a 10-point scale for better variability in ratings compared to a 5-point scale. Favor quantitative feedback, which is easier to process and track over time. As a follow-up, managers can request qualitative feedback, which is summarized and anonymized via Gen AI.

Outputs

Composite Scores - Create a composite score from the inputs above. Components and weights vary by function and level. Continuously refine the components and weights based on what aligns best with your organizational values and goals. Individual component scores may be imperfect, but in aggregate, they provide a directional sense and show patterns over time.

Feedback Delivery - The system above provides managers with a near-automated source of rolling performance data, but feedback should be delivered to their reports quarterly, or monthly if there’s a major positive or negative shift. Be compassionate but firm, upholding quality and core values. Everyone deserves a second chance, but sustained underperformance should lead to significant changes.

Formulaic Promotions & Comp - The quantitative evaluation system above powers formulaic decisions on leveling and compensation (our next two sections) and precludes the need for burdensome calibration meetings, resulting in more objectivity, significantly reducing the time spent, and enabling a quarterly cadence.

Section 2: Leveling

Promotions

Promote employees who sustain performance above certain thresholds for 3-12 months. Instead of manufacturing new titles and layers, opt for 10 levels within each of the 3 layers—doers, functional managers, and general managers—with employees being promoted one level at a time every 2 years on average. Transition from doer to manager can happen before reaching level 10 but includes a provisional period to ensure aptitude in managerial functions.

Demotions and Dismissals

Demote employees who sustain performance below certain thresholds for 3-12 months. Remove the stigma of demotions and make it easy to recover from them. In a 10-level-per-layer system, down-leveling from, say, level 7 to 6 is more innocuous than being demoted to a new title or role. A system with more flexibility for down-leveling affords more flexibility to uplevel.

Two consecutive demotions should lead to a different role or dismissal, with an option for a performance improvement plan in borderline cases.

Section 3: Compensation

Increase or decrease compensation quarterly based on level and average weighted evaluation scores over the past 12 months. Bidirectionality creates more flexibility for compensation increases.

Remove managerial discretion from compensation decisions to improve fairness and reduce bias, allowing rare exceptions for practical purposes. Formulaic compensation can be shared transparently, further increasing fairness. The exceptional component can remain undisclosed to avoid any blowback.

Article 6: Hiring continues here .