Calibrate (verb)
mark (a gauge or instrument) with a standard scale of readings
Managers across corporate america have a yearly ritual of bench marking their reports according to standards set by Human Resources (HR) for their assigned job level and role. This process results in rewarding (bonus, additional equity) and or promoting employees to higher levels and roles.
A popular way of performing this calibration is by asking managers to place their assigned reports into a 9 box grid.
There are three levels of employee performance that is plotted on the
X axis:
- Low
- Solid
- High
Similarly, there are three levels of employee potential that is
plotted on the Y axis:
- At
- Moderate
- High
A graph is then constructed and each employee is plotted into their appropriate box.
This plotting is largely based on the managers subjective judgment. The manager is supposed to take into account the work performed by the employee during the year, and interactions (culture fit) with their peers and other leaders.
As can be inferred, most attractive spot on the grid is High performance, High potential (termed as STARS). This coveted ranking
signals that these employees are ready to take on challenges in a new
level.
Conversely, Low performance, At potential (termed as BAD HIRES)
employees are ready for a performance improvement plan (PIP).
The popularity of this calibration process is because it scales well as more employees are added to this grid.
Most companies reward top 20-25% (ones that place in any of the High boxes in the graph) of their employees well, keep the middle of road folks satisfied with cost of living adjustments, and build processes to remove bottom 5-10% by placing them on PIPs.
There are other nuances we can infer based on employee placement in the graph as well (you can read more about these techniques in the linked article above).
Fidelity loss
The obvious problem in this scheme is the loss of information that happens after an employee is placed on the grid. Adding more employees to the grid assumes that:
- All managers use exactly the same judgment criteria
- All employees had exactly the same access to opportunity
- All employee levels and roles are interchangeable
Since the entire placement exercise performed by the manager is subjective, these assumptions remain broken by default. However, no one from the CEO to HR to the employees complain of these obvious problems.
Some corporations and managers try to remove variance in subjectivity (Noise) by conducting popularity contests, such as crowdsourcing manager calibration judgments across a broader swath of committees, and hope that the consistency in popularity will lead to better performance model results.
However, this tends to skew peformance measurement towards culture fit, crowd pleasing, and extroversion and penalizes difference of opinions, pessimists, and whistle blowers. Notice that we’ve lost track of performance management totally.
Stack ranking
Typically, financial budgets are also constrained (there is only so much money to dole out), and managers and organization heads are asked to rank employees within the grid as well.
This ranking will then decide whether the employee retains their place
in any of the high boxes or falls into Solid performance, Moderate potential. This has no bearing on the Low performance, At potential
crew, all of whom are considered equally bad, irrespective of role or
level.
If all of this sounds ridiculous, it is because it is!
Can we do better?
- Could we (a) assign and (b) measure business impact within teams
and groups, and then use that to reward employees appropriately by
level?
- For example, Objective Key Results (OKRs) are a popular and durable technique to make measurable improvements in business.
- Could we perform succession planning by measuring effectiveness of
coordinating tasks?
- For example, leading scrums, project planning.
Long range projects
A reward scheme which depends on business metrics like OKRs works well for short term projects which can have a measurable impact in a quarter or two (remember that we are performing annual reviews).
However, programmers and other strategic professionals employed in the corporation may find flaw in this performance model because it relies on operational effectiveness, and it sounds terrible if they are penalized for out of control decisions made by their higher ups.
This implies that we need to strike a delicate balance between risk and reward. Long term projects are high risk, and consequently should be rewarded in a similar fashion as well.
Illustrative Example
A programmer is working on a new social product. The work will take atleast a year if not more in order to have a lasting positive impact to the top line.
A SRE is ensuring that the corporate web application remains in good health and always available to customers. The work is immediately useful, and requires constant upkeep to be in good shape.
Performance management
| Functional Role | Initiative | Objective | KR |
|---|---|---|---|
| Products | New subscription product | Increase users | +20% |
| Operations | Reliable site | Uptime | 99.99% |
Reward structure
| Role | Level | Compensation | Bonus (KR driven) | Equity (KR driven) | Vesting |
|---|---|---|---|---|---|
| Programmer | Principal | 110 | 10% | 50% | 3 years |
| SRE | Principal | 100 | 30% | 10% | 1 year |
Vulnerability to gamification
OKRs can be subject to sandbagging and reduce the effectiveness of the scheme. This structure only works well when the CEO, upper management and even the board is detail oriented and vigilant on what objectives move the top or bottom lines, by how much and when.
Conclusion
A nice upside of this proposal is that rewards tie into capitalistic outcomes, and succession planning is determined through performance of coordination tasks.
A lingering question remains: How will this scheme identify low performing employees?
Actually, it is quite straightforward, as when KRs are not being met, the signal is quite clear. It is the leader who takes the fall, so in this case it would either be the manager or senior IC level that come into performance discussions.
Accountability in a leadership position becomes straightforward. The hiring process automatically improves, and low leaf performers (bad hires) are managed out quickly, as otherwise the leaders risk their own jobs.
This scheme is dreamed up by an engineer programmer in a single day (and hence will have more rough edges and corner cases to contend with), and it is surprising that no one in the industry has come up with viable alternatives to an obviously broken model.
Corporations need people to succeed, and our reward structure is badly in need of modernization.