Continuous feedback vs annual reviews: what the data says

Annual reviews fail for one predictable reason: people cannot accurately remember a year of work. Research on memory and recency bias shows that evaluations written from memory skew toward the last few weeks. Continuous accomplishment tracking solves this by building a record as work happens, so periodic reviews are grounded in evidence instead of recollection.

The annual review has a bad reputation, and mostly it has earned it. But the problem is rarely the concept of reviewing performance. The problem is asking people to review a year of work from memory.

Why reviews written from memory are unreliable

Human memory is not a recording. When someone is asked to recall how a colleague performed over twelve months, they do not retrieve twelve months of data. They retrieve whatever is easiest to recall, which tends to be the most recent work, the most vivid work, or the work that fits a general impression they already formed.

Two effects do most of the damage:

• Recency bias: the last few weeks dominate the evaluation, drowning out the previous eleven months.
• Confirmation bias: once a reviewer has an overall sense of someone, they remember the moments that fit it and forget the ones that do not.

Recency bias is well documented in HR and management research. SHRM describes it as a manager placing too much weight on an employee's recent performance rather than their contributions across the whole review period, and warns that reviews can end up unfairly positive or negative "based on recent events rather than a comprehensive view." The bias exists because recent events are the ones easiest to recall. Asked to reconstruct a year, a reviewer reaches first for the work that happened recently, and that accessible handful of moments quietly stands in for the whole period.

The consistent recommendation from the same research is to stop reconstructing from memory. SHRM's guidance is to keep detailed records of performance as it happens rather than relying on recall, which is exactly what a running accomplishment record provides.

These are not character flaws. They are how memory works. The fix is not to tell reviewers to try harder. It's to stop relying on memory in the first place.

What continuous tracking changes

Continuous accomplishment tracking means every employee keeps a running log of what they accomplish as it happens. A developer logs the feature they shipped this week. A support lead logs the process they improved last month. Each entry takes seconds.

At review time, that log becomes the input to the review. The reviewer is no longer reconstructing a year from memory. They are reading a record that was created in the moment, across the whole period.

The result is a review that reflects the actual work: the wins from month three, the growth from month eight, and the projects that mattered even if they happened a long time ago. That is why accomplishment tracking is the input, and reviews are the output.

This does not mean abandoning periodic reviews

There is a version of this debate that argues for abandoning reviews entirely and relying only on continuous feedback. That misses what reviews are for.

Continuous feedback tells an employee how they are doing right now. It's excellent for course correction. But employees also need a periodic, structured moment where their performance across a period is evaluated, a rating is given, and compensation or promotion decisions are made.

The right model is both: continuous accomplishment tracking feeds a periodic review. The record is collected continuously; the evaluation happens on a cadence.

What this looks like in practice

Employees should not have to fight to be remembered. When the record exists, the review writes itself from evidence, and the employee walks in with the full picture rather than hoping their manager noticed.