A leaked TikTok doc reveals how managers are told to score employee performance — and what not to talk about

Business Insider

a day ago

A leaked TikTok document reveals the company's guidance to managers on conducting performance reviews — and what they shouldn't tell their teams.
The company and its owner, ByteDance, conduct reviews twice a year, and executives set targets for the distribution of team scores based on business priorities. But if you're a TikTok manager, you shouldn't talk about the rating curve with employees or call it a "forced distribution," according to the document.
"The messaging to management or to the employee is that we do not have forced distribution," one staffer said, but "every cycle we do get 'guidance" on scoring.
The document, shared with managers of TikTok's e-commerce division around the time of its last review cycle, provides a framework for its employee scoring system. It instructs those managers to assign the three highest scores to no more than 5% of team members in total and limit the top four ratings to 10% or fewer employees.
TikTok did not respond to requests for comment.
TikTok's next performance reviews are expected to kick off next month, and some employees tell Business Insider they're worried that low scores could lead to a new wave of performance-improvement plans (PIPs) or exit offers.
The last cycle in March led to a stream of low ratings and cuts to teams like e-commerce. The company's leadership was disappointed with TikTok Shop's US results last year, and staffers previously told BI that they felt the reviews were used to trim head count.
Performance reviews have become dreaded by many across the tech industry, as companies like Meta and Microsoft have recently overhauled their processes to weed out low performers. TikTok and ByteDance adjusted their review process last year to call out more low and high performers, blending self-evaluations, coworker feedback, and ultimately manager discretion to assess team members and meet top-down performance distribution goals.
How TikTok scores workers
TikTok managers score employees based on a mix of their work output, their alignment with core cultural tenets, internally called ByteStyle, and their leadership principles if they are a manager.
BI earlier published a version of the company's scoring rubric, which ranges from "F" for failed to "O" for outstanding. A score of "M-," reserved for employees who sometimes fall below expectations, or "I," for staffers who often do, can lead to a PIP or an offer to leave via a mutual separation agreement.
Staffers receive an overall score that considers each factor, including peer reviews. Ultimately, there's no standard calculation that decides the overall score, according to the manager guide.
How managers should avoid bias
The guidelines document said that while managers are expected to weigh different performance factors in reviews, they ultimately should make their own call about each overall score.
"There is no specific formula to calculate the Overall Rating, which means leaders need to make comprehensive management discretion," the company said in the document.
Because managers have a large degree of discretion in setting individual scores, the company's guidelines focus on what to avoid when making decisions.
For example, one slide in the document lists various biases to watch out for.
Those included:
Recency bias: Overvaluing recent events instead of considering the full performance cycle.
Choosing a single positive or negative workplace trait for a worker and scoring based on just that.
The "central tendency," or rating everyone near the mid-point to avoid conflict with workers.
Showing preference to workers who act similarly to a manager, or share some of their workplace habits (e.g., a team member is high performing because they also look at their email outside work hours).
The "contrast effect," where reviewing one candidate who is very strong or weak makes another candidate seem comparatively better or worse.
The "assimilation effect," when team members are reviewed back-to-back and appear too similar to differentiate.