Reevue
Every applicant leaves with something useful.
Reevue embeds into a company's application portal and returns tailored, specific feedback to candidates — including the ones who were not selected. It reads the application against the role the company actually defined, and explains the distance between them in language a person can act on.
- Interface
- Portal SDK · REST · Webhooks
- Unit of work
- Application ↔ criterion pair
- Latency target
- Async, minutes
- Human control
- Review before send
- Built for
- Employers, universities, grant and fellowship programmes
Not a screenshot.
Hover a highlighted passage to see which criterion it answers. Toggle Reevue off to read what the same applicant would otherwise receive.
Submitted application — excerpt
During my final year I built dashboards in Power BI for the student union, pulling numbers from a shared spreadsheet each week. I taught myself basic SQL through an online course and have used it on sample datasets. My dissertation used a regression to test whether attendance predicted grades, which I presented to my department.
Criterion alignment
gap — below the bar for this role
gap — below the bar for this role
Alignment is computed against criteria the hiring team wrote. It is an explanation of the fit, not a ranking — Reevue never selects.
Returned to applicant
generated · reviewedThe SQL evidence in your application is coursework on sample datasets. This role's first criterion is querying production data, where the difficulty is unfamiliar schemas and dirty joins rather than syntax. A public dataset with a genuinely messy schema, written up as a short case study, would close most of this distance.
Nothing in the application shows a project you carried from question to delivered decision. The dashboard work is described as a recurring task set by someone else. Reviewers look for one instance where you chose the question, and can say what changed because of the answer.
Your strongest signal. The dashboard work and the departmental presentation both show translation for a non-technical audience — keep this prominent in future applications.
Four steps, none of them hidden.
Each stage produces an artefact a human can read, contest and overrule. That constraint shapes the architecture more than any model choice does.
Read the criteria the team wrote
The role definition is parsed into weighted criteria. Anything vague is surfaced back to the hiring team before the first application is processed, because unclear criteria produce unusable feedback.
Anchor evidence in the application
Each criterion is matched to the passages that speak to it. A criterion with no supporting passage is a gap, and the absence itself is the evidence.
Draft the feedback
One or two gaps, named plainly, with the specific next step that would close them — plus what the applicant did well, so the strong signal is not lost in the critique.
Team review, then send
Nothing leaves without the configured level of human review. Teams tune tone once, spot-check continuously, and can block any individual message.
What it actually does.
Portal-native
01Drops into an existing applicant tracking system or careers portal. Candidates never leave the company's own flow.
Role-grounded evaluation
02Feedback is derived from the criteria the hiring team wrote down, not from a generic model of what a good CV looks like.
Specific, not soothing
03Named gaps, evidence from the application itself, and the concrete next step — instead of 'we went with other candidates'.
Auditable by the employer
04Every piece of feedback is traceable to a criterion and a passage. Teams review, tune, and approve tone before anything is sent.
Bias surface monitoring
05Feedback distributions are monitored across cohorts so a systematic skew shows up as a measurement, not as a complaint.
No decision authority
06Reevue explains outcomes. It does not rank, score, or select candidates — that stays with the hiring team.
A rejection is the only contact most applicants ever have with a company. Making it useful is not a nice-to-have; it is the whole product.
- Reevue does not rank, score or select candidates — it explains a decision already made.
- It does not send anything the hiring team has not authorised.
- It does not use protected attributes, and it monitors its own output for skew across cohorts.
Limits are part of the specification. They are enforced in the system, not in the marketing.
Want Reevue against your own data? Pilots start with a short scoping call and a sample of the real thing.