The case against annual performance reviews has been made many times and is largely correct. Once-yearly feedback is too infrequent to be actionable. The recency bias in annual reviews is severe — events from January are functionally invisible by December. The cognitive load of summarizing a full year in a single document distorts what gets included and what gets omitted.
So organizations switched to continuous feedback. Quarterly cycles, monthly check-ins, always-on feedback tools. And then discovered that the outcomes hadn't improved as much as expected. Engagement surveys still showed that engineers found reviews unhelpful. Managers still felt unprepared for calibration sessions. Promotion decisions were still made without adequate evidence.
The problem was not the frequency. The problem was the structure — or rather, the absence of it.
The Case Against Annual Reviews — and Why It's Partially Wrong
The critique of annual reviews conflates two separate problems: the problem of frequency and the problem of structure. Annual reviews are bad for both reasons, and the frequency argument is more emotionally resonant, so it dominated the reform agenda. The structure problem went largely unaddressed.
The frequency problem is real: a gap of 12 months between formal feedback instances is too long for the feedback to be actionable. Behavioral patterns that could have been corrected in February are now entrenched in October. Development opportunities that were visible mid-year are now missed.
But the structure problem is equally real: annual reviews are typically unstructured essays produced by one person (the direct manager) based on a full year of imperfect memory, with no consistent framework for what to observe, no defined evidence sources to draw on, and no calibration process to ensure that "excellent" means the same thing in Team A as it does in Team B.
Moving to quarterly reviews that have the same structural problems produces quarterly bad feedback instead of annual bad feedback. This is not an improvement worth the organizational cost of the transition.
What the Evidence Says About Frequency
The research on feedback frequency and performance improvement is more nuanced than the simple "more is better" narrative that drove most continuous feedback adoptions. The finding is not that frequent feedback is better than infrequent feedback; it is that timely feedback is better than delayed feedback.
Timely means connected to specific, recent events. An engineer completes a difficult architecture review; feedback about that review is most useful within a few days, when the experience is fresh and the specific behaviors are available for reflection. Waiting three months for a quarterly cycle means the feedback has to reconstruct what happened from memory — and the specific behavioral details that make feedback actionable have largely faded.
The practical implication is that the right cadence varies by event, not by calendar. Some behaviors warrant immediate feedback. Others benefit from accumulation before being surfaced — patterns are more useful than one-off observations, and patterns take time to establish.
The Real Variable: Evidence Anchoring
The variable that most strongly predicts the usefulness of a feedback instance is not its timing — it's whether it is anchored to specific, observable evidence.
"Good communication skills" is unanchored feedback. It describes a conclusion, not an observation. An engineer reading it cannot understand what specific behavior produced that assessment, cannot identify what she would need to do differently to improve it, and cannot use it as evidence in a promotion conversation.
"In the Q3 launch planning meeting, explained the database migration risk in terms the non-technical stakeholders could act on, which enabled a faster decision than we'd typically see in that group" is anchored feedback. It is tied to a specific event, describes a specific behavior, and is usable as evidence. It is also more useful for the reviewer — writing it forces the clarity of observation that produces good feedback.
"Moving from annual to continuous feedback without fixing the structure produces the same low-quality observations at a higher frequency. This is not progress — it's churn."
Why Most Implementations Fail
Most continuous feedback implementations fail because they focus on the tooling and cadence without addressing the competency gap in feedback writing. Most engineers and managers have never been taught how to write evidence-anchored feedback. They have been taught that feedback should be specific and actionable, which is correct but insufficient — it tells you the destination without describing the route.
The result is that feedback tools get adopted, cycles get established, and the feedback that flows through the new system is structurally identical to the feedback that came through the old system — unanchored, conclusion-focused, and largely useless as evidence.
The second failure mode is feedback fatigue. Asking engineers and managers to provide frequent structured feedback without reducing the administrative burden of doing so well results in compliance without quality. People complete the cycles because it's required, but they complete them quickly, which means the output is low effort and low value.
What Structured Continuous Feedback Looks Like
Structured continuous feedback has three design requirements that most implementations don't satisfy simultaneously.
Observation prompts, not open fields. Rather than asking "How is this person doing?", structured feedback asks questions that force specificity: "Describe a situation in the last 30 days where you observed this person's impact on the team." The prompt shapes the output without constraining the content.
Skills dimension anchoring. Feedback that is tied to a skills taxonomy dimension becomes comparable across cycles and across reviewers. When every instance of "system design" feedback references the same taxonomy dimension and assessment criteria, accumulation across multiple cycles becomes possible. You can track trajectory, not just snapshots.
AI-assisted drafting that improves quality, not just speed. The most practical way to improve feedback quality at scale is to use AI writing assistance that prompts for specificity — not as a replacement for the reviewer's judgment, but as a co-pilot that asks clarifying questions until the feedback is grounded in observable evidence. Organizations that implement this see both quality improvement and time reduction in the feedback writing process.
Key Takeaways