Associates stop giving honest upward feedback when they conclude that someone inside the firm can trace it back to them. Survey Research Associates (SRA) has run confidential upward reviews for US law firms since 1991, and the pattern is consistent: candor is not a personality trait or a cultural achievement. It is a function of who holds the data. Change that one variable and the same associates who wrote three bland sentences last year write something a partner can act on.
Why do associates give diplomatic feedback instead of honest feedback?
Because being diplomatic is the rational choice, given what they are actually being asked to do.
Picture a fourth-year associate asked to assess the partner who staffs their matters. The review runs on a platform their firm licensed. Their firm's HR team administers it. Their firm's IT function holds the credentials. The partner in question decides which matters they work on, signs off their hours and will be in the room when partnership is discussed.
The associate is not weighing whether the feedback is true. They are weighing what happens if it is read by the wrong person. Against a career, the value of a candid answer to question seven is close to zero.
So they write that the partner is "generally responsive" and "could provide clearer direction at times." Both statements are technically accurate and completely unusable. The firm files a report showing its partnership is performing adequately, and eighteen months later a strong mid-level associate leaves for a competitor.
What does the attrition data suggest is happening?
The associate attrition rate at US law firms was 19% in 2025, and 83% of those departures occurred within five years of hire, an all-time high, according to the NALP Foundation's Update on Associate Attrition covering calendar year 2025 across 141 firms.
Two things are worth pulling out of that.
First, the five-year concentration. People who leave that early are not leaving because they outgrew the work or reached a natural transition. Something happened, and it happened close enough to the start of their career that the firm had time to notice.
Second, the smaller-firm gap. NALP reports 24% attrition at firms of 100 or fewer attorneys against 16% to 18% at larger firms. Smaller firms have less anonymity to offer in any internal feedback process, because the practice groups are smaller and the reviewer pool is thinner.
Neither figure proves that feedback quality causes attrition. But if a firm's own upward review data says supervision is fine while a fifth of its associates leave every year, at least one of those two things is not telling the truth.
Does anonymity in an internal system actually work?
Rarely, and the reason is structural rather than technical.
Most platforms genuinely do anonymize responses. The problem is that the associate has no way to verify it, and several reasons to doubt it. They know the system is administered by colleagues. They know IT has database access. They may have heard about a firm where a comment was traced, whether or not the story was true.
Anonymity is not a technical state. It is a belief the respondent holds, and it survives only as long as nothing in the arrangement invites doubt. An internal system asks associates to trust that the people they are criticizing have built a process that protects them from the consequences.
There is also a mathematical problem. In a twelve-person tax group, a partner reading feedback about themselves can often work out who said what, whatever the platform promises. In a four-person group it is not even a puzzle. Anonymity that fails at small group sizes fails exactly where associates are most exposed.
Want to see how the mechanics work in practice? Our law firm upward review service sets out the eligibility mapping, response thresholds and comment review process we apply to every cycle.
What changes when a third party holds the data?
The calculation the associate is running changes, which is the only thing that matters.
When responses go to an outside firm and never enter the client's systems in raw form, the honest answer to "who will see this" stops being "people here." That is not a reassurance offered in a communications email. It is a fact about where the data physically sits, and associates can reason about it.
Three controls make the promise credible rather than rhetorical:
A minimum response threshold. No partner receives a report unless at least [CONFIRM: n] associates responded. Below that, results are suppressed entirely rather than reported with a warning, because a warning does not stop a partner counting who staffed them this year.
Comment review before release. Written comments are read and, where necessary, redacted before they reach the partner. A comment identifying its author by matter, secondment, practice group or a distinctive turn of phrase is edited or withheld.
Aggregation for leadership. Firm-level reporting shows patterns across the partnership. It does not show who said what about whom, and leadership cannot request that.
How can a firm tell whether its feedback is already diplomatic?
Four signals, none of which require a new survey to spot. Look at your last cycle.
Participation. If it sits below 60%, treat that as the finding rather than an inconvenience. Associates who distrust a process do not tell you so, they decline to take part.
Score compression. If almost every partner lands within a narrow band near the top of the scale, the instrument is not discriminating. Real partnerships vary more than that.
Comment length. Diplomatic feedback is short. Candid feedback, positive or critical, tends to be specific and therefore longer.
Absence of anything negative. A cycle that produces no critical comments at all has not discovered that your partnership is uniformly excellent. It has discovered that nobody felt safe.
What should a firm actually do about it?
Start by looking honestly at last cycle's participation rate and score distribution against the four signals above. That costs nothing and usually settles the question.
If the signals are there, the fix is not a better questionnaire or a stronger message from the managing partner. Associates do not distrust the wording. They distrust the destination. Moving the data outside the firm is the change that alters their reasoning, and everything else follows from it.
Frequently asked questions
Will associates believe a third party any more than they believe our HR team? They believe the structure rather than the promise. What changes their reasoning is that responses physically leave the firm and that reports are suppressed below a response threshold. Both are verifiable facts about the process rather than assurances.
What participation rate should we expect? [CONFIRM: SRA average]. As a general rule, below 60% the result should be treated with caution, because low participation is itself a signal about trust.
Should upward review scores affect partner compensation? We advise against it and will explain why at the design stage. Once associates believe their feedback affects partner pay, they manage their answers and the data stops being worth collecting.
Our partners say they already get honest feedback. How do we test that? Check score compression and comment length in your last cycle. Partnerships that genuinely receive candid feedback show real variance in both.
Can we run this alongside our existing annual review process? Yes, and most firms do. Upward reviews measure supervision quality. They are not a substitute for the evaluation record.
See what honest upward feedback looks like
SRA has designed and run confidential upward reviews exclusively for US law firms since 1991. We will walk you through a live cycle and share an anonymized sample report, so you can see the deliverable before committing to anything.
Contact us · Survey Research Associates, Inc. · 30 Wall Street, 8th Floor, New York, NY 10005 · 800-523-8350
Related reading
- Why Does Your Law Firm Need Upward Reviews?
- Downward Reviews vs Upward Reviews: What's the Difference?
.png){kind=small}