Methodology
How the LanceRank Score is calculated.
One number, six factors, zero mystery. Every input is independently verified and the weights are public. One strong platform is enough — multi-platform is a bonus, not a requirement.
Algorithm v6.6 — effective 2026-07-10
At a glance
Six factors. One hundred points. One career stage.
Each factor is weighted by how reliably it signals genuine trust. Verified track record and consistent ratings dominate because they are the hardest to fake.
Each factor explained
What earns the points.
Track Record
up to 35 ptsAnchored on your work VOLUME (jobs/contracts/reviews), with earnings + hours as bonuses, plus a Bayesian rating-quality lift.
The single hardest thing to fake. Track Record is anchored on your work VOLUME — total verified jobs/contracts/reviews (log-scaled to 100 jobs = full marks). Tracked earnings (log-scaled to $100K) and hours logged (log-scaled to 2,000 hours) are BONUSES on top: they can only raise your score, never lower it. The combine is max-dominant (70% your strongest signal, 30% the average), so one maxed-out signal is never dragged down by a low-but-real value in another. Missing fields never penalize — a Fiverr profile without an "hours" field, or a fixed-price Upwork profile with very few logged hours, is scored on the strength of the signals it DOES have. A platform that doesn't expose hours or earnings never costs you points. A tenure ramp (years active) reaches full multiplier in 2 years. Volume also counts verified direct reviews + client-confirmed offline projects toward "jobs". Vouched projects count by quality, not headcount (v6.5): a vouch from an aged, independently-verifiable corporate domain counts at full weight; a vouch from a free webmail address or a young, thin domain counts at 0.4×, and only one counted vouch per client domain per 12 months. With no connected platform, vouches contribute at most 12 of the 35 Track-Record points — enough to matter, never enough to fake a full history — and vouches from free webmail alone never unlock a public number. A Bayesian-smoothed rating across all sources is multiplied in as a quality lift (a 4.9 across 250 reviews beats a 5.0 across 3).
Recency & Tenure
up to 20 ptsActive recency (16 pts) + years of history (4 pts).
Sixteen points for recency (18-month half-life — a senior between contracts isn't punished) and four for tenure (5+ years = full marks). The score reflects both freshness and depth. Tenure counts your platform-stamped join date — including on an elite platform that hides its public counts: a Toptal "Verified Expert" profile shows no rating or job count, but its member-since join date still counts toward your years of history (the same place its badge already earns Platform Credibility). Recency (last-active freshness) is the opposite — it only credits real verified activity, never just the date you onboarded. And "activity" means an externally-verified event (v6.5): an extension re-sync, a scrape-confirmed increase in your job or review count, a payment-verified direct review, or a client-confirmed vouch. Re-uploading a screenshot never refreshes your recency — a fresh screenshot is not fresh work.
Claimant Credibility
up to 20 ptsIdentity verification + OAuth proofs + extension-confirmed platform ownership.
Identity verified via Stripe Identity (a government ID), LinkedIn OAuth, GitHub OAuth, or the LanceRank Chrome extension confirming you own the platform profiles you connected. The more proofs you carry, the higher this factor — non-platform-derived, so single-platform freelancers can still build a strong score. Verification means verification: any confirmed real-world identity — a government ID or a LinkedIn login whose name matches your profile — earns the identity credit once, and it never stacks (holding both proves the same identity twice, not two identities). A LinkedIn login whose name we could not check against anything keeps its login credit but not the identity credit, and a LinkedIn login whose name does not match the profile, or that is already linked to another account, earns nothing here. The factor still caps at 20 points no matter how many proofs you stack, and the verification ledger on your profile is tiered — a LinkedIn login is labelled "LinkedIn login" in the ledger, not "Identity verified" (a LinkedIn login whose name matches the profile still counts as a verified identity for the Trusted Leader gate).
Platform Credibility
up to 10 ptsBest badge held — across all your platforms.
MAX across platforms — the strongest credential wins. One Top Rated Plus on Upwork is full credit; you don't need it on three platforms. Elite badges (Top Rated Plus, Expert-Vetted, Toptal Verified Expert / Top Talent / Top 3%, Fiverr Top Rated Seller, Fiverr Pro, Contra Top/Featured Independent, PeoplePerHour Top Certified, 99designs Top Level) → 10 pts; high-tier (Top Rated, Level 2 Seller, Preferred Freelancer, PeoplePerHour Certified, 99designs Top-Mid Level) → ~8; mid-tier (Rising Talent, Level 1 Seller, 99designs Mid Level) → ~5; strong Upwork Job Success Score without a badge → ~7. An elite badge counts even on a platform that doesn't publish counts — some elite networks (e.g. Toptal) show only the badge, no rating or job count, and both the badge AND that platform's member-since join date are scored (full Platform Credibility here, plus tenure under Recency & Tenure); for high-tier and mid-tier badges we still want to see some verified volume. Paid membership tiers (e.g. Guru's plan tiers) are not credentials — they grant no points. One platform is enough. New in v6.5, freelancers who work direct (no marketplace badge) can still earn this factor: 3+ direct client reviews backed by payment-rail or platform-transaction proof add +5 here, and 8+ add +7. The proof has to be a real payment trail — an uploaded invoice PDF is self-reported and never qualifies for this rung.
LinkedIn Endorsements
up to 10 ptsSkill endorsements + LinkedIn profile-signal credibility.
A 2026-05-22 addition. Pulls skill endorsement counts and LinkedIn profile signals (tenure, listed services, recommendations) — imported through the LanceRank Chrome extension when you sync your own LinkedIn profile. Only your own profile, only with your active click — never scraped from outside your browser. In v6.5, endorsements themselves are capped at 6 of the 10 points and scored concavely: the first endorsements on a skill count for a lot, but past roughly 15 on the same skill each extra one adds almost nothing (200 endorsements is not 200× the trust of 15). The remaining points come from your LinkedIn tenure, recommendations, and listed services — so a wall of easy-to-trade skill endorsements can no longer dominate the factor.
Skills
up to 5 ptsPassed skill quizzes — admin-set point value per quiz.
Optional. Take a LanceRank skill quiz; pass it; the score-impact value (set per quiz by the admin team) lands here. Caps at 5 points across all passed quizzes combined.
Career stages
The number, translated.
The 0–100 number gives the precision. The stage gives the verdict at a glance.
Top tier on LanceRank — proven across volume, quality, and consistency. Nobody enters the 90s without a verified real-world identity: a government ID, a name-matched LinkedIn login, or extension-proven platform ownership.
Trusted by clients across platforms with a deep, independently-corroborated history.
Proven track record. Established freelancer worth shortlisting. Nothing above 69 renders without corroborated evidence.
Building momentum. Has the credentials, building the corroborated volume.
Just getting started. Score grows fast as verified work accumulates.
The top band is a requirement, not a coincidence
Nobody enters the 90s without a verified real-world identity: government-ID verification, a LinkedIn login whose name matches the profile, or cryptographically-proven platform ownership together with corroborated work evidence. It is a hard requirement, not a side effect of the weights — an unconfirmed name can never reach Trusted Leader.
The displayed number
How the raw score becomes the number you see.
The engine computes a raw 0–100 score from the six factors. A single, published, frozen curve then maps that raw score to the number you see, so the steepest gains land where the typical member actually sits — early verified work moves the displayed number a lot, and the top of the range is deliberately hard to reach. The curve only changes what is DISPLAYED; it never changes the underlying computation or your rank against anyone else.
Published curve — frozen knots
| Raw score (engine) | Displayed |
|---|---|
| 0 | 0 |
| 10 | 18 |
| 25 | 40 |
| 40 | 60 |
| 52 | 71 |
| 63 | 82 |
| 72 | 90 |
| 85 | 96 |
| 100 | 100 |
LanceRank Scores run from 0 to 100 and are shown as a bare number welded to a career-stage word — never as a fraction. The bands are defined by criteria, not by a fixed percentile, so the number means the same thing this year and next.
The display curve is frozen. We only ever re-fit it if a published distribution alarm is breached, never more than once every two quarters, and never without 90 days' public notice. Existing profiles keep their current displayed number until their next material change, and every certificate carries the scale version it was issued under — so a future re-anchor makes old numbers self-dating, never contradictory.
Worked examples
| Profile | Raw → Displayed | Stage |
|---|---|---|
| Empty / spam stub (no verified work)Nothing to score. A synthetic account cannot manufacture a number. | raw 2 → 4 | Building |
| Median freshly-claimed profileHonestly weak until real work is connected — the point of the floor. | raw 9 → 16 | Building |
| Claimed, top-10% of new claimersSome proof, not yet enough to cross the public number floor. | raw 19 → 31 | Building |
| Identity + first verified work signalThe public numeric floor: the first point a number is shown at all. | raw 25 → 40 | Rising |
| Thin but real newcomer, one platformA single core verified action moves the displayed number 6+ points here. | raw 41 → 61 | Rising |
| Day-one maxed single-platform pro, ownership proven (E3)One strong, corroborated platform is enough to reach Verified Pro. | raw 63 → 82 | Verified Pro |
| Trusted-Leader entry, gov-ID + corroborated evidenceThe 90s are hard-gated on strong identity — never reachable by weights alone. | raw 72 → 90 | Trusted Leader |
| Deep multi-year, multi-platform veteranThe very top of the range is reserved for the most corroborated careers. | raw 92 → 98 | Trusted Leader |
Band shares — dated observation
| Trusted Leader (90+) | ~0.2%–1.5% of verified members |
| Verified Pro (80–89) | ~2%–8% of verified members |
| Established (70–79) | a modest slice above the corroboration line |
| Rising (60–69) | the broad middle of active verified members |
| Building (below 60) | most freshly-claimed and thin profiles |
Bands are defined by criteria, not by a fixed percentile. These shares are a snapshot, refreshed nightly — they describe the population, they do not set the thresholds.
Evidence tiers
How strong the proof is decides how high the number can go.
Any profile showing 70 or above carries ownership-proven or independently-corroborated evidence. Screenshots alone never cross 70.
E3 — Verified
VerifiedFull range, up to 100.
Ownership is proven or the numbers are independently corroborated: the LanceRank extension confirmed you own the platform profile, OR an anonymous scrape saw the same rating, counts, badge, and earnings bracket your screenshot claims, OR a government ID sits alongside payment-verified client reviews. Only E3 evidence can render Established, Verified Pro, or Trusted Leader.
E2 — Reported, pending cross-check
Reported — pending cross-checkDisplay capped at 69 (top of Rising).
A screenshot a human reviewer accepted as plausible, but that nothing independent has confirmed yet — including the normal case where a platform is login-walled and an anonymous scrape cannot see the numbers. It counts, but it can never alone cross 70. Screenshots alone never render the Verified Pro credential a client trusts. Re-uploading a fresher screenshot never lifts the cap or refreshes recency — a new screenshot is not new work.
E1 — Unclaimed
Not scoredNo number at all.
A profile we built from public data that no verified member has claimed. It shows the verified facts we found, not a LanceRank Score.
Three ways to move from "Reported" to "Verified"
- Install the LanceRank Chrome extension and sync the platform — proof of ownership is instant.
- Wait for automatic cross-check — we re-attempt an independent read of the platform on a 30-day cycle; a fetch or parse failure never counts against you.
- Book a supervised live-session verification with our team — reviewed within the published dispute SLA below. This is the extension-free path.
What the extension can see: The LanceRank Chrome extension only reads the public stats already visible on YOUR OWN platform profiles, and only in the moment YOU click sync inside your own logged-in browser. It confirms you control the account. It never reads private messages, client lists, financial pages, or anyone else's profile, and it never scrapes a platform from outside your browser.
When a number shows — and when it doesn't
An honest weak page beats a mockable number.
The public numeric floor
A public number requires three things: a verified identity, at least one real verified work signal, and a displayed score of 40 or more. Below that line a profile shows what it has honestly earned — "Identity verified · Stage 1: Building · N of 8 verified signals" — but no number, because a low two-digit number invites more misreading than it resolves. The owner still sees their exact number and their ladder on their own dashboard.
Unclaimed profiles
Unclaimed profiles we built from public data are not scored at all. They read "This profile isn't scored — LanceRank Scores are calculated only for verified members," and show the verified facts we found (for example "★ 4.9 across 212 jobs on Upwork · Top Rated Plus · member since 2019 — source: Upwork, checked 12 days ago"). Every stub carries "This is me — claim or correct" and "Remove my profile," and removal is instant and automated.
Reading it against your platform stats
Why is my LanceRank lower than my platform stats?
Because LanceRank runs harder than any single platform's averages. A 96% Job Success Score reflects one platform's clients being polite; your LanceRank aggregates independently-verified evidence across your whole career and discounts anything it cannot corroborate. A raw 96% platform stat routinely maps to a LanceRank in the 80s — which still places you in the top few percent of verified members. The number is deliberately a harder currency than a platform average, which is exactly why a client trusts it more.
Something wrong? Dispute it.
Think a fact or a score is wrong? Every profile and certificate carries a dispute link. A human reviews every dispute within 5 business days — no bots, no editorial layer.
For new freelancers
Your credential shows from day one — the number arrives once there is something real behind it.
A public number requires three things: a verified identity, at least one real verified work signal, and a displayed score of 40 or more. Below that line a profile shows what it has honestly earned — "Identity verified · Stage 1: Building · N of 8 verified signals" — but no number, because a low two-digit number invites more misreading than it resolves. The owner still sees their exact number and their ladder on their own dashboard.
A verified work signal is: a connected platform with real jobs, a verified direct review, a client-confirmed vouched project, or a passed skill quiz. Every new verified item makes the score steadier — the algorithm puts more weight on factors that have more signal behind them — and the "backed by N verified data points" counter beside your card is a shareable artifact from the very first one, before any number exists.
Why you can trust the number
No editorial layer. No pay-to-rank.
Every input is verified
Reviews require proof of work. Offline projects require the client to confirm by email. Platform data is scraped from publicly visible profiles only.
You cannot pay for a higher score
Pro never affects your score, your rank, or organic search order. Featured placement is paid, capped at two clearly-labeled slots per page, and never mixed into organic results. Free and Pro are scored by the identical published function. The only non-score ordering adjustments in search are verification signals open to every account: claimed profiles sort with a +10 term and extension-verified profiles with a +5 term — neither can be bought.
Recalculated continuously
Every new verified review, vouched project, or platform sync triggers a fresh score. There is no manual approval and no editorial intervention.
Methodology updates
What changed, and why.
We publish every change to how the score is computed. Version history below, newest first.
v7 — Scale recalibration (display only)
2026-07-02We recalibrated how the score is DISPLAYED. The computation did not change in this release — the factor tightening shipped separately in v6.5 (see the entry below), so your rank against every other freelancer is identical. A published, frozen curve maps the raw score to a decade-band scale (Trusted Leader 90+, Verified Pro 80s, Established 70s, Rising 60s, Building below 60), uncorroborated screenshots are capped below 70, and a public number now requires identity plus one verified work signal. Thin and unverified profiles still read weak — that is the point.
- Published display curve — a single frozen curve maps the raw score to the displayed number, with the steepest gains where the typical member sits. The raw score and every rank are unchanged.
- Decade bands on the grade-schema joints — bands move to 90 / 80 / 70 / 60 so the number reads the same to a client whether or not they mentally add "out of 100."
- Screenshot evidence capped below 70 — an uncorroborated screenshot can never alone render Established or above — Verified Pro requires proven ownership or independent corroboration.
- Public numeric floor at 40 — a number shows only with a verified identity and at least one real work signal; below that a profile shows its honest credential and signal count instead.
- Unclaimed profiles are not scored — stubs show the verified facts we found and a claim/remove control, never a synthetic number.
v6.6 — Identity parity
2026-07-10Verification means verification. A LinkedIn login whose name matches the profile now earns the same identity credit as a government-ID check — any confirmed real-world identity earns the identity credit once, and it never stacks. At the same time, a LinkedIn login that failed its name check, or that was already linked to another account, no longer earns the LinkedIn sign-in credit. Factor weights and the 20-point identity cap are unchanged.
- Full identity credit for a confirmed LinkedIn identity — the Trusted Leader gate has always accepted a LinkedIn login whose name matches the profile as a verified real-world identity; the points now agree. Any confirmed real-world identity — a government ID or a LinkedIn login whose name matches your profile — earns the identity credit once; it never stacks, so verifying both proves the same identity twice without double-counting.
- Failed name checks lose the LinkedIn sign-in credit — a LinkedIn login whose name does not match the profile, or that is already linked to another account, no longer earns the LinkedIn sign-in credit. A small number of scores go down as a result — that is what verification means.
Per-persona before / after (identity points within Claimant Credibility)
| Test persona | Before | After |
|---|---|---|
| Confirmed LinkedIn identity, no government ID | 5 pts | 10 pts |
| Both LinkedIn and government ID | 10 pts | 10 pts (unchanged) |
| LinkedIn login that failed its name check | 5 pts | 0 pts |
v6.5 — Factor integrity
2026-07-02The six factor weights did not change (35 / 20 / 20 / 10 / 10 / 5). We tightened what earns points inside them so the score is harder to inflate with forgeable or low-effort signals. Ranks may shift slightly for a small number of profiles; each affected freelancer receives a plain-English, line-item explanation of what moved and how to restore it.
- LinkedIn endorsements cap and curve — skill endorsements now count for at most 6 of the 10 LinkedIn points and score concavely (past ~15 on a skill they flatten), so a wall of easy-to-trade endorsements can no longer dominate the factor.
- Government-ID identity reweighted up — a Stripe government-ID check was reweighted to score the same as a LinkedIn login (5 points, up from 3) and more than a GitHub sign-in (3). The 20-point identity cap was unchanged. The identity-signal hierarchy this introduced was superseded by v6.6 (see the entry above): any confirmed real-world identity now earns the identity credit once.
- Direct-branch platform-credibility rung — freelancers who work direct can earn +5 (3+ reviews) or +7 (8+) from client reviews backed by real payment-rail proof — uploaded invoice PDFs never qualify.
- Vouch quality weighting — vouches from aged, independently-verifiable corporate domains count in full; free-webmail or young-domain vouches count at 0.4×, at most one per client domain per year, capped at 12 of 35 Track-Record points with no platform, and free-webmail-only vouches never unlock a public number.
- Recency counts verified events only — only an extension re-sync, a scrape-confirmed count increase, a payment-verified review, or a client-confirmed vouch refresh recency — re-uploading a screenshot does nothing.
- Trusted Leader hard-gated on identity — no profile displays 88+ without a verified real-world identity — a government ID or a LinkedIn login whose name matches the profile (each clears the gate alone), OR extension-proven platform ownership together with corroborated work evidence. It is a published requirement, not an arithmetic side effect.
Per-persona before / after
| Test persona | Before | After |
|---|---|---|
| Per-persona figures are being finalised from the v6.5 recalculation and will be published here. | ||
See your own score.
Free forever. Two minutes to claim. Your reputation, finally portable.