Continuous feedback and the annual review do two different jobs, and dropping either one breaks the system. Continuous feedback is real-time steering: it corrects the work while it's still in flight and builds the trust that makes candor safe. The periodic review is the aggregation layer: it rolls signal up across projects and leads, calibrates it fairly, and drives the decisions people care about: growth, ratings, promotion, pay. Kill the reviews and feedback never becomes a fair decision. Kill continuous feedback and the review becomes a once-a-year surprise built on recency and thin memory.
So the useful question is how to wire them together. The evidence cuts both ways: feedback given daily rather than annually makes people 3.6× more likely to be motivated to do outstanding work (Gallup, 2022; updated 2024), and firms that dropped ratings and reviews entirely saw performance fall about 10% and manager conversation quality drop 14% (CEB/Gartner, 2016). If your people are staffed across engagements under rotating leads, you need both: per-engagement feedback captured while it's fresh, then rolled up and calibrated in a cycle.
What job does each one do?
They solve different problems. Continuous feedback steers the work, and the review cycle decides. Confusing the two is why most firms do one badly and skip the other.
Continuous feedback is the ongoing, forward-looking, low-stakes exchange that happens in 1:1s, engagement debriefs and milestone moments. Its job is to correct course while the work can still change, to build skill, and to make candor normal. It's frequent and small, and it works best kept out of the pay decision. Engagement tracks with feedback given far more often than once a year (Gallup, 2022; updated 2024).
The review cycle is the periodic aggregation-and-decision layer. Its job is to gather signal from everyone who saw the work, calibrate it across leads so a rating means the same thing firm-wide, and produce the decisions people care about: grade, promotion, partner track, pay. Here a single lead's view falls short. About 62% of rating variance is the idiosyncratic rater rather than the person being rated, and only ~21% is actual performance (Scullen, Mount & Goff, Journal of Applied Psychology, 2000). The cycle exists to correct for that, and continuous feedback can't.
Why does annual-only feedback break down?
Feedback that comes once a year arrives as a post-mortem, too late to steer anything. If the only structured feedback you get is the annual review, four things go wrong.
- Recency bias. A review written in November mostly remembers October. Eleven months of engagements collapse into whatever happened last. Ratings research documents this as a known distortion, with no personal failing involved (DeNisi & Murphy, Journal of Applied Psychology, 2017).
- Thin memory across a year of projects. In a project-based firm, the person rating you may have staffed you for six weeks in Q1. Reconstructing that from memory in Q4 produces a vague, low-fidelity picture.
- Feedback arrives too late to act on. By review time the engagement is closed, the client has moved on, and the lesson is theoretical. Feedback only helps when there's still work to change.
- Surprise ratings. The fastest way to destroy trust in a review is to raise something at the rating moment that nobody raised during the work. A surprise at review time points to a management failure, whatever the person's performance.
None of this makes the review the problem. The review can't double as the steering wheel, though. Feedback that happens all year does that job, so the fix for a bad annual review is usually more feedback before it, and rarely a better form.
Why does continuous-only feedback break down?
Steering without a decision layer leaves you with lots of local signal and no fair way to act on it. "Just kill the annual review and give feedback all the time" sounds modern, but the firms that tried it got a warning. When organisations removed ratings and the formal review, employee performance fell about 10% on average, and employees rated the quality of their performance conversations with managers 14% lower than at firms that kept ratings. Managers also spent fewer hours on informal performance conversations (CEB/Gartner, 2016).
Continuous feedback alone has no aggregation and no calibration. Each lead's view stays siloed, and there's no mechanism to make one lead's "great" mean the same as another's. With no shared bar, promotion, partner-track and pay decisions go to whoever advocates loudest. That is the idiosyncratic-rater problem the cycle is built to contain (Scullen, Mount & Goff, Journal of Applied Psychology, 2000). Feedback can also misfire on its own. In a meta-analysis of 607 studies it raised performance modestly on average (d ≈ .41), and over a third of interventions made performance worse (Kluger & DeNisi, Psychological Bulletin, 1996). Without a system around it, a stream of feedback piles up and never turns into a fair outcome.
How do the two fit together without surprises?
Continuous feedback feeds the cycle, so the review contains nothing new. It aggregates and calibrates what people already heard, and that handoff is the whole design.
Picture two layers. The steering layer runs all year: per-engagement notes, 1:1 feedback and debriefs, all small, specific and mostly developmental. The deciding layer runs periodically: it pulls those documented signals together, calibrates across leads, and sets the rating and growth plan. You build the review from feedback the person already received, so nothing in it surprises them. The rating sums up what they already know.
- Keep a shared record that rolls up. Nobody can aggregate feedback that lives only in someone's head. A light, per-engagement note, captured when the work is fresh, lets the cycle assemble a fair, multi-lead picture instead of one manager's recency-weighted guess (DeNisi & Murphy, Journal of Applied Psychology, 2017).
- Decouple developmental feedback from the rating moment. Run growth feedback continuously and keep the pay/promotion decision to the calibrated cycle. This is why removing the cycle backfired and why cramming everything into the annual verdict backfires: honesty stays cheap only when today's candid note isn't tomorrow's comp risk (CEB/Gartner, 2016; Buckingham & Goodall, "The Feedback Fallacy," Harvard Business Review, 2019).
Two design rules make the handoff work: keep a shared record that rolls up, and decouple developmental feedback from the rating. Leading practice has converged on the same design: ongoing performance conversations running alongside a formal cycle (CIPD, Performance Management).
How do you combine them in a billable, multi-lead firm?
In an agency or boutique, "you need both" gets practical fast. Your people are staffed across several engagements a year under different leads, so both single-layer models fail harder here than anywhere. Annual-only asks one manager to rate work they barely saw. Continuous-only leaves each lead's feedback siloed, with no calibration for promotion or the partner track. Combining them makes ratings fair across projects. Here are six moves, each built to take minutes:
- Capture feedback per engagement, while memory is fresh. When a consultant rolls off, the lead who staffed them leaves a short, specific note then, instead of eleven months later. It's the highest-leverage habit you can build, because it feeds fair calibration.
- Attach feedback to work already happening. Put a standing five-minute feedback slot in 1:1s and make a two-way retro the default close of every engagement phase. You add no new meetings. Feedback rides the rhythm you already have, so utilization pressure can't erase it.
- Aggregate and calibrate in the cycle. The periodic review pulls the per-engagement notes together and calibrates across leads, so "strong" means the same thing whoever staffed you. That is the direct fix for the idiosyncratic-rater effect (Scullen, Mount & Goff, Journal of Applied Psychology, 2000).
- Keep the developmental track separate from the rating. Use continuous feedback for growth and the calibrated cycle for the grade and pay. Mixing them makes candor expensive and dries up the feedback (CEB/Gartner, 2016).
- Make client feedback a team ritual. Client-facing work throws off tough external feedback. Debrief it together, separate signal from sting, and feed the durable part into both layers.
- Have leads and partners model it. A partner who asks the team for feedback on how they ran an engagement, and acts on it, makes candor safe faster than any policy. Frequent, in-the-moment, strengths-anchored exchange beats saved-up critique (Buckingham & Goodall, "The Feedback Fallacy," Harvard Business Review, 2019).
A quick both/and self-check
Score your firm: one point per "yes". Six or more and your two layers are wired together. Four or fewer and you're running one layer and calling it a system.
- Feedback happens in 1:1s and engagement debriefs as well as at review time.
- Leads capture per-engagement feedback while it's fresh instead of reconstructing it at year-end.
- The review aggregates documented feedback rather than one lead's memory.
- Ratings are calibrated across leads, so "strong" means the same firm-wide.
- Nothing raised at the review is a surprise to the person hearing it.
- Developmental feedback is kept separate from the pay/promotion decision.
- You haven't "just dropped" the review: decisions still trace to a calibrated cycle.
- Client feedback is debriefed as a team and fed into both layers.
FAQ
Do you really need both continuous feedback and an annual review?
Yes, because they do different jobs. Continuous feedback steers the work in real time and builds the trust that makes candor safe. The periodic review aggregates signal across leads, calibrates it, and drives ratings, promotion and pay. Drop the reviews and feedback never becomes a fair decision. Drop continuous feedback and the review becomes a recency-driven surprise. The evidence points both ways: daily feedback makes people 3.6× more likely to be motivated to do outstanding work (Gallup, 2022; updated 2024), but firms that removed ratings entirely saw performance fall ~10% (CEB/Gartner, 2016).
Isn't the annual review dead?
No. The annual-only model is the broken part. When firms removed ratings and the formal cycle, performance dropped about 10% and manager conversation quality fell 14% (CEB/Gartner, 2016). Leading practice kept a calibrated cycle and added ongoing performance conversations alongside it (CIPD, Performance Management). Keep the review and drop the only.
How do we stop reviews from being full of surprises?
Feed the review from continuous feedback. If you've given someone specific, documented feedback throughout the year, the review sums up what they already know. The fastest way to destroy trust in a rating is to raise something at review time that nobody raised during the work. That is a management failure, whatever the person's performance (DeNisi & Murphy, Journal of Applied Psychology, 2017).
How does this work when people are staffed across many leads?
Capture short, specific feedback per engagement while it's fresh, then let the periodic cycle roll those notes up and calibrate across leads. That makes a rating fair when no single manager saw all the work. It is also the direct fix for the idiosyncratic-rater effect, where about 62% of rating variance is the rater rather than the person (Scullen, Mount & Goff, Journal of Applied Psychology, 2000).
Should developmental feedback affect someone's rating?
Keep them on separate tracks. Run growth feedback continuously and keep the pay/promotion decision to the calibrated cycle. Mixing them makes honesty costly (today's candid note becomes tomorrow's comp risk) and dries up the feedback you need (Buckingham & Goodall, "The Feedback Fallacy," Harvard Business Review, 2019; CEB/Gartner, 2016).