// methodology
How we score every EOR
Scores come from a transparent rating system, not hand-picked opinions. Every provider runs the same weighted model, so a Select Score of 88 means the same thing across the entire dataset. The system was built and is maintained by Robbin Schuchmann.
The scoring model
How many countries the provider can employ staff in.
Published EOR price per employee per month. Lower scores higher; providers that do not publish a price are scored below those that do, because price transparency is part of the rating.
Security certifications and review-verified trust signals, weighted by how much evidence backs them.
Integration breadth and product depth.
Support channels, responsiveness and review sentiment.
Community sentiment from verified Reddit discussion, human-screened for relevance.
The process
- 01
Collect
We pull provider data into one normalized schema: pricing, coverage, capabilities and certifications.
- 02
Verify
Third-party review ratings are scraped weekly from G2, Capterra and Trustpilot, with sanity and drift checks.
- 03
Score
The scoring algorithm applies the same weighted model to every provider, totalling a Select Score out of 100.
- 04
Re-check
Scores recompute on a weekly schedule, and on demand, so updated data flows through within a week. Material changes are logged in the changelog.
// evidence and confidence
Why a score is not just an average
A rating is only as trustworthy as the evidence behind it. We build that into the score in two ways, so a provider cannot top the ranking on claims alone.
Review-volume confidence
A 5.0 from sixteen reviews is weaker evidence than a 4.7 from six thousand. Before a third-party rating counts toward Compliance or Support, we weight it by how many reviews stand behind it, so a thin sample barely moves a score.
The evidence gate
The final Select Score is pulled toward a neutral baseline when little independent data backs a provider. A well-reviewed provider keeps its full score; an under-evidenced one is held closer to neutral until the data catches up. This is why a total can sit below the simple average of its category scores.
Not yet rated
When we do not yet have the data a dimension needs, we mark it “not yet rated” and score it at the neutral baseline, rather than the bottom. A missing data point is our gap, not a fault of the provider, so it never drags the score down. Each profile shows how many of the six dimensions are data-backed, so you can see how complete the assessment is.
// reputation
How the Reputation score works
Reputation is the sixth dimension. It captures what operators actually say about a provider on Reddit, a signal the star ratings on review sites miss. The pipeline is deliberately human-gated, and the math that turns it into a number is fixed and documented.
- 01
Source
We search Reddit for genuine discussion of each provider’s EOR service. A deterministic filter drops job ads, provider marketing and best-of listicles before anything reaches review.
- 02
Human relevance gate
A real person reviews every candidate and approves only mentions that are genuine discussion of that specific provider. This step exists because a keyword match is not always about the product: a brand name can appear all over Reddit in unrelated contexts, and some posts are competitors or marketing agencies talking their own book. We screen those out. We verify relevance, not the truth of any individual comment.
- 03
AI sentiment
The rating system labels each approved mention positive, neutral or negative toward that provider, with a one-line reason. Any label can be corrected by hand, and a hand-set label is never changed by a later run.
- 04
Aggregation
Approved, labelled mentions become a score from 0 to 100. Recent and upvoted mentions carry more weight, a first-hand account counts more than a second-hand remark, and small samples are pulled toward neutral so a handful of posts cannot swing a provider.
- 05
Author checks
We look at who is talking, not just what they say. Accounts that read as promotional, such as brand-new profiles with no history or accounts praising the same provider across many threads, keep their mentions visible but carry almost no weight in the score. Each profile states how many mentions were down-weighted this way.
A provider with fewer than ten approved mentions is scored at a neutral 60, never penalised for being under-discussed. Reputation is 15% of the Select Score. Star-rating sentiment stays in Compliance and Support, so nothing is counted twice. There are no invented reviewer personas: sentiment is machine-labelled on human-screened posts, and every provider is scored by exactly the same rule.
// fee transparency index
How the fee transparency score works
The fee transparency index is a separate, simpler score, and it is deliberately not about price. It measures only how much of its real cost a provider is willing to put in writing. A provider that publishes a high fee ranks above one that hides the same fee, because what we reward here is disclosure, not being cheap. Undisclosed is the default and scores worst. It never feeds the Select Score.
The six costs we check
EOR base price
The per-employee monthly EOR price, published with a link to the provider’s own pricing page.
FX / currency markup
The margin added when converting salary into local currency, a fee providers rarely put in writing.
Security deposit
Any upfront cash deposit or salary buffer held before the first payroll run.
Setup fee
A one-time onboarding or implementation charge to start service.
Offboarding fee
A charge to terminate or offboard an employee at the end of the engagement.
Minimum term
Any minimum contract length or commitment before you can leave without penalty.
The three states
Disclosed with source
The provider publishes the actual figure on its own site, and we link to it. Full credit. For hidden fees this means a fact we verified against the provider’s own domain, human-approved before it counts.
Partially disclosed
The figure exists in our data but cannot be independently confirmed on the provider’s public site, for example a price with no source, or a fee that only appears in a contract. Half credit.
Undisclosed
Nothing public, including “contact sales” with no number. This is the honest default, and it scores zero. Most providers sit here for most fees.
We never infer a fee from marketing language or from memory. If we cannot point to the source, it does not count. As our hidden-fee research verifies more sourced facts, transparency scores rise.
We always
Apply the same model to every provider
Verify review data against primary sources
Screen community mentions by hand before they count
Publish how each score is built
Log material changes in the changelog
We never
Take placement fees to rank a provider
Let a provider buy a higher score
Invent reviewer or expert personas
Hide our scoring method
Present opinions as verified facts