Review methodology

How the RoleSprint Decision-First Score works

Version 1.0, last updated 2026-08-27. Seven dimensions, 100 points, one question. Published in full so you can disagree with the weighting rather than take the number on trust.

What the score measures, and what it does not

It answers one question: how well does this product help a candidate decide where their application effort should go?

It is explicitly not a measure of overall software quality, not a customer-satisfaction rating, not a safety or security certification, not a measure of interview success, and not an offer probability. It says nothing about how well a product serves employers. A product can score low here and still be excellent at what it was built for — several in this comparison are. It is also not independent: we publish it and we sell a competing product.

Commercial disclosure

RoleSprint publishes this score and sells a competing product. The rubric is built around one question — how well a tool helps you decide where your application effort should go — so it under-rewards products built for volume, which is a real thing to want. The criteria and weights are published in full, RoleSprint is scored under exactly the same rules and loses points in the same places, and every subscore states its reason so you can disagree with a specific number rather than the whole thing.

If your bottleneck is finding and reaching enough openings rather than choosing between them, read the Workflow utility subscore on its own and treat the total as close to irrelevant.

Every score, including ours

RoleSprint was scored first, before any competitor, so the number could not be reverse-engineered from whatever would have flattered us. RoleSprint loses points in public: half marks on Workflow utility, because it discovers no jobs and submits nothing.

Every figure below is a RoleSprint Decision-First Score out of 100. None of them is an overall product rating.

RoleSprint85 / 100
Jobscan61 / 100
Jobright46 / 100
JobCopilot42 / 100
AIApply37 / 100

RoleSprint scores highest, which is what you would expect from a rubric built around the problem RoleSprint was designed for. That is the honest reason to read the subscores rather than the totals: every competitor on this page beats RoleSprint on Workflow utility, one of them by double, and Jobscan comes within two points of us on Transparency.

The seven dimensions

Listed heaviest first. Integer points only — the evidence does not support decimals, and fake precision is its own kind of dishonesty.

Opportunity prioritization

25 pts

Does the tool help decide whether a specific opportunity deserves effort, before any further application work is done?

Full marks require a per-opportunity output that a candidate can act on as a decision — something that distinguishes roles worth pursuing from roles worth skipping, and says why. Partial marks for surfacing or ranking relevant roles without resolving the effort question. Low marks where the product's answer to every opportunity is effectively 'apply to it'.

Counts as evidence: An explicit per-role recommendation the candidate can act on
Stated reasoning tied to the specific role and the specific candidate
Ranking or filtering that visibly reduces, not expands, the set worth acting on
Does not count: Volume of jobs indexed or surfaced
A relevance or match percentage with no stated decision meaning
Speed of applying, which is a throughput property rather than a prioritization one
RoleSprint scores 22 / 25 on this dimension: The Job Fit Analyzer's whole output is a per-role Apply, Consider or Skip verdict with stated reasoning, which is this dimension's definition. Three points withheld because the candidate has to bring the role: RoleSprint does not surface opportunities, so it can only prioritise what you already found.

Evidence-based fit

20 pts

Does it evaluate the candidate's actual evidence against role requirements in context, rather than reducing fit to keyword overlap or automation heuristics?

Full marks require assessing whether specific requirements are actually evidenced by the candidate's history, and surfacing the ones that are not. Partial marks for structured comparison that remains primarily lexical. Low marks where fit is asserted without a stated basis, or where generated documents introduce claims the candidate's history does not support.

Counts as evidence: Identifying requirements the candidate does not evidence
A published account of how fit is determined
Reasoning that references context such as constraints or seniority
Does not count: Keyword or phrase overlap presented as fit
A tailored document as proof that fit was assessed
Any number with no published derivation
RoleSprint scores 17 / 20 on this dimension: Analysis works from parsed resume evidence against the specific posting, names requirements the resume does not support rather than papering over them, and weighs stated constraints such as relocation. Three points withheld: the judgement is model-produced and RoleSprint publishes no calibration or accuracy data for it.

Candidate control

15 pts

How much control and visibility does the candidate keep over what is sent and what is claimed on their behalf?

Full marks where nothing reaches an employer without the candidate seeing it first. High marks for automation gated behind a mandatory review step. Middling marks where a review step exists but is optional against an automatic default. Reduced marks where submission is autonomous with no review offered at all.

Counts as evidence: A review step the candidate must pass before anything is submitted
A published, opt-in review step, scored below a mandatory one
Stated constraints on which roles are acted on
Visibility of exactly what was sent and when
Does not count: A settings page offering constraints the product does not commit to honouring
The ability to cancel after submissions have already gone out
RoleSprint scores 15 / 15 on this dimension: Nothing is ever submitted on the candidate's behalf and no automation touches an employer's site, so there is no path by which something reaches an employer unseen. This is full marks on control and it is also why Workflow utility below is only half.

Transparency

15 pts

How clearly does the product expose its pricing, boundaries, methodology and limitations to someone who has not paid?

Full marks require public pricing, a public account of how the core output is produced, and stated limits on what that output means. Points are deducted for each of those a prospective buyer cannot find without creating an account. Publishing an unflattering limitation earns points; the rubric rewards disclosure, not favourable facts.

Counts as evidence: Prices readable without signing up
A published explanation of how the main score or match is computed
Publicly stated limits, caveats or failure modes
Does not count: Pricing visible only after account creation
A methodology page that restates marketing claims
Testimonials, badges or user counts
RoleSprint scores 13 / 15 on this dimension: Pricing is public, methodology pages are public, and output boundaries are stated on the pages themselves. Two points withheld: RoleSprint publishes this scoring rubric and sells against the products it scores, and some usage limits are only visible once inside the product.

Search-strategy clarity

10 pts

Does the product help assess whether the set of roles being targeted forms a coherent search?

Full marks require an output about the target set as a whole — whether the roles cohere, and where the targeting is too broad, too narrow or mismatched. Partial marks for per-role signals a candidate could aggregate themselves. Low marks where the product operates strictly one role at a time.

Counts as evidence: Analysis spanning several target roles at once
A stated finding about the shape of the search rather than a single job
Does not count: Saved searches, alerts or filters
A dashboard listing applications without assessing their coherence
RoleSprint scores 9 / 10 on this dimension: The Job Search MRI examines several target roles together and reports whether the targeting coheres, which is what this dimension asks for. One point withheld because it is a point-in-time diagnosis rather than something that tracks whether the search improves.

Workflow utility

10 pts

How much of the job-search workload does the tool actually complete?

Awarded for breadth of work genuinely done for the candidate: discovery, document production, submission, contact, tracking, interview preparation. This dimension deliberately rewards the throughput products. A tool that does one narrow thing well scores low here even if it scores high everywhere else.

Counts as evidence: Job discovery at scale
Document generation
Submission or submission assistance
Tracking, contact and interview preparation features
Does not count: Features announced but not available
Quality of the work produced, which is scored under Evidence-based fit
RoleSprint scores 5 / 10 on this dimension: Half marks, honestly. RoleSprint produces application kits, finds contacts and tracks applications, but it indexes no jobs, submits nothing, and generates no throughput. Against tools that surface millions of roles and apply automatically, this is a narrow slice of the workload.

Privacy and claim discipline

5 pts

How honestly are data handling and the meaning and limits of the product's outputs communicated?

Full marks require no unsourced outcome statistics and a stated boundary on what the output does not establish. Each distinct unsourced outcome multiplier or percentage costs a point. Usage counts are not outcome claims and are not penalised.

Counts as evidence: An explicit statement of what the output does not predict
Outcome figures accompanied by a published basis
Readable data-handling information
Does not count: Interview, offer or success-rate figures with no cited derivation
Any framing implying an application outcome can be guaranteed
RoleSprint scores 4 / 5 on this dimension: Outputs are labelled with what they are not, and RoleSprint publishes no interview or offer probability anywhere. One point withheld for the interview guarantee, which is an outcome-adjacent commercial promise even though its conditions are published.

Where the evidence comes from

1.The vendor's own live pages first. If a fact is not on the vendor's site, it does not become a fact here. The vendor's own blog and comparison articles count, but then the fact must say so and carry both dates — when the vendor published it, and when we read it.
2.Public pricing where it exists, recorded with the exact page and the date it was read. Where a vendor does not publish prices, we say so rather than repeating a figure from somewhere else. Where a vendor has moved its prices behind a sign-in since we last read them, we say that too, rather than presenting an older figure as a price we confirmed.
3.Published third-party ratings, only where the source and the review count could both be verified. Ratings, review counts and star distributions are the platform's own figures across all reviews. These are shown separately and never folded into our score.
4.Individual reviews, used only as illustration. We read the reviews visible on each vendor's public Trustpilot profile on one day — roughly the twenty most recent — without sorting, filtering or paging, and we did not count them. So these pages say "among the reviews we read" and never "users commonly complain": one read of one page cannot establish how frequent anything is. A reviewer's account is never treated as proof that a product behaves a particular way.
5.RoleSprint analysis, which is labelled as ours and is the only layer where judgement enters.

We have not used these products

We did not create a paid account for this review. What the product does, and what it costs, comes from the sources linked on this page rather than from using it ourselves. Where a fact is only visible behind a sign-up, we say we could not verify it instead of guessing.

This matters for reading the scores. Anything only observable from inside a paid account — how good the matching really is, whether generated documents hold up, whether limits bite — we have not seen. Where that gap affects a score, the subscore says so. Publishers who claim hands-on testing without evidence of it are a large part of why these queries are hard to research, and we would rather be less impressive than join them.

The vendor's own live pages, listed as sources on this page, each with the date we read it.
Current public pricing where the vendor publishes any, recorded with the date and the exact page it came from.
Published third-party ratings where the source and review count could be verified, shown separately from our own score.
The RoleSprint Decision-First Score methodology, published in full.

How these reviews are kept current

Pricing and features in this category move quickly. Every review records the date its facts were read, and each fact that could change carries its own date. When a vendor changes something material we re-check the whole record rather than patching one line, because a page mixing current and stale facts is harder to trust than one that is openly out of date. If you find a fact that has moved, the dated source on the page is there so you can check it yourself.