WhichPayroll Research Methodology

WhichPayroll Research Methodology

Source hierarchy, claim taxonomy, scoring rules, freshness policy, and correction process.

Version 1.0 • May 2026

Every claim on WhichPayroll is scored against a five-tier source hierarchy, with vendor marketing ranked last and statutory sources ranked first. Claims are tagged as verified fact, sourced data, or editorial judgement, and editorial judgement is capped at 30% of any page. Freshness policy forces a review on price, coverage, and regulatory claims at fixed intervals, and corrections are published with a dated note rather than silently amended.

---

What Does This Methodology Cover?

WhichPayroll Research collects, scores, and publishes market intelligence about payroll providers, EOR platforms, global payroll services, and UK payroll software. Every published finding links to its source. Every statistic states its sample size and observation date. Every estimate is labelled as an estimate.

This page tells you the rules governing what WhichPayroll Research publishes, how it scores evidence, how it handles missing data, and how you can submit a correction.

How Do We Rank Our Sources?

WhichPayroll Research ranks evidence sources in five tiers. Higher tiers override lower tiers when sources conflict. When you see a source reference on a research page, it maps to one of these five tiers.

1. Official government or statutory source: legislation, official regulatory guidance, government agency publications. Used for country obligation data (minimum wage, payroll deadlines, severance rules).
2. Provider pricing page or official documentation: the provider’s own published pricing page, official product documentation, or press release. Used for pricing, feature, and coverage claims.
3. Provider-published data sheet or contract template: downloadable resources published by the provider. Used where a pricing page alone is insufficient.
4. WhichPayroll field research: systematic observation and recording of provider-published information at a specific date. Used where no formal pricing page exists but pricing is discoverable.
5. WhichPayroll estimate or model: a calculated value based on disclosed inputs and documented assumptions. Always labelled explicitly as an estimate. Never presented as a provider disclosure.

How Do We Classify Claims?

Every claim in a WhichPayroll Research publication must use one of these seven wording classes. Deviations are treated as publication errors.

How Do We Set Confidence Levels?

Every data record carries one of three confidence levels:

• High: sourced directly from official legislation or a provider’s published pricing page, verified within the last 30 days.
• Medium: sourced from provider documentation or field research, or high-confidence data observed more than 30 days ago.
• Low: sourced from secondary references, or where primary source could not be confirmed. Displayed with a caution label. Not used for legal certainty language.

How Do We Keep Data Fresh?

Stale records are labelled “Verification required” and suppressed from live index table rankings until refreshed. Stale country compliance data is not used for legal certainty language under any circumstances.

How Do We Handle Missing Data?

Where data is not publicly available, WhichPayroll records the field as “Not disclosed” rather than estimating or omitting it. “Not disclosed” is itself a finding. It tells you that the provider does not publish this information publicly. WhichPayroll does not publish estimates to fill missing provider pricing fields.

How Do We Calculate a Provider Score?

Every provider score we publish is a weighted composite of four dimensions on a 0 to 10 scale. Three of the four are read from datasets on this site. The fourth, integration depth, is our own assessment, and we label it as such wherever the score appears. The four weights sum to 100%, there is no fifth term, and no part of the score is bought, negotiated, or adjusted for providers who pay us commission.

The point of publishing the arithmetic is that you can check it. Take a provider's four inputs, run them through the steps below, and you should land on the number in the ring at the top of that provider's review. If you cannot, we have made a mistake and we want to hear about it.

Multiply each dimension by its weight, add the four results, multiply by 10, and round to one decimal place. That is the whole calculation.

Which country count the coverage term uses

Most providers publish more than one country number, and they are rarely the same number. Papaya Global markets 180+ countries for payroll and says 160+ for compliant hiring. Oyster HR puts 180+ on its homepage for contractor contracts and 120+ on its employer-of-record page. Deel leads with 150+ and states 130+ where it owns the entities. Feeding the bigger number into the score instead of the smaller one moves a provider by several tenths of a point, which is enough to reorder a shortlist, so the choice cannot be left to whoever last edited the table.

One rule, applied to all seventeen providers on the same day:

• We use the employment count. The score basis is the number of countries a provider says it can legally employ someone in, because that is what an employer of record is bought for. Counts for payroll only, contractors only, visas, benefits or country-guide libraries do not qualify, however prominently they are printed.
• It has to come from the provider. We read the figure off the provider’s own site. Analyst counts, directory listings and our own tallies never override what a provider publishes about itself.
• A rounded headline is taken at its word. “100+” scores 100. We do not swap a marketing headline for a larger number counted off the provider’s own territory list, because doing that for one provider while the rest sit on headline figures would quietly flatter that one. Plane is the live example: it publishes 100+, its own hire list names 115 territories, and it scores on 100.
• If a provider publishes two employment counts, the lower one wins. We score what a provider will stand behind on every page, not its best number. Papaya Global is the case that forced this: its employer-of-record page shows “180+ countries supported”, its pricing page sells that plan as “compliant hiring in 160+ countries”, and 180+ is also what its menu prints for payroll on every page of the site. We take 160, the count attached to the product you actually buy.
• Silence scores zero, and silence means the whole site. If a provider publishes no employment count anywhere on its own site, coverage scores zero and the table reads “Not disclosed”. We do not estimate it. This is a disclosure index, and a number nobody published is not evidence; the alternative is printing a figure no source supports. Gusto is in that position as at 6 August 2026. Rippling was recorded there on 3 August and should not have been: that check read only its employer-of-record product page, which still names no number, while Rippling’s own blog has carried “live in 80 countries” since June 2025. A missing figure on the product page is never on its own enough to score a provider zero, so we now read the blog and the newsroom before recording anyone as non-disclosing. Rippling scores on 80.

Every coverage input carries the address it was read from and the date it was checked. No provider’s score moves on its own: the whole table is re-verified together, every published score is then recomputed from it, and each one is checked to confirm it still reproduces from the figure printed beside it.

A worked example you can check

Safeguard Global is the clearest one to follow, because its four inputs pull in opposite directions. Safeguard states 187 countries, the widest network on the index, so coverage is 187 divided by 187, which is 1.0, weighted to 0.300. Its pricing transparency is Low, so 0.25 weighted to 0.063. It publishes no trust or security page and no processor data processing agreement, so 0 points of 5 gives 0.00, weighted to 0.000. We band its integration depth as Low, so 0.3 weighted to 0.060. Those four add to 0.423, and multiplied by 10 that is 4.2, which is the score on the Safeguard Global review.

Look at what that example exposes. Safeguard scores a perfect 10 on coverage, the only provider on the index to manage it, and a zero on security disclosure, which no other provider manages either. Coverage carries 30% of the weight, so a perfect score there cannot rescue a composite when most of the remaining 70% is empty. If country count matters more to your business than what a provider will tell you before you sign, the number in the ring is the wrong summary for your decision and the coverage row on the review page is the right one. That is true of any weighted score. It is why every review states each dimension separately instead of only the total, and why we would rather you argued with the inputs than trusted the headline.

What the score deliberately leaves out

Three things buyers care about are not in this number, and pretending otherwise would be worse than saying so plainly.

• Cost friction. Deposits, FX margins, per-cycle processing fees, and non-refundable seat minimums are not scored. They are frequently the difference between two providers at the same headline price, so we cover them in the review body instead.
• Offboarding and exit. Notice periods, termination handling, and what happens to your data when you leave are not scored.
• Depth behind the country count. Coverage counts countries, not owned entities. A provider serving 170 countries through partners scores exactly the same as one serving 170 through its own entities. Which of those you are buying is a question the review answers and the score cannot.

Support quality is absent too, for the plain reason that we have no way to measure it consistently across 17 providers without treating vendor-supplied references as data. When those gaps close, the weights will change and we will date the change on this page rather than quietly reissuing the numbers.

When a review page shows a different number

A few providers are bought for a job this index does not describe. Gusto is the clearest case. It is a domestic US payroll product that partners with Remote for employer-of-record work and publishes no country count of its own, so it scores zero on a coverage term built to measure global employment reach. That is an accurate reading of what Gusto discloses and a poor summary of what it is actually good at. Where a review publishes a score weighted for a different job, three rules apply: the score is captioned so you can see it is not the WhichPayroll disclosure index, the page states the WhichPayroll disclosure index figure as well so both numbers are in front of you, and the page shows the weights it used, exactly as this page does.

We also do not send search engines a rating that differs from the score on the page. Where the two would disagree, the page carries no machine-readable rating at all.

How the Pricing Transparency Index is scored

The Pricing Transparency Index feeds the 25% transparency term above, and it is scored on its own before it gets there. It assigns a rating of High, Medium, or Low based on a provider's composite disclosure across eight fields: public price availability, starting price amount, pricing model clarity, setup fee disclosure, FX markup disclosure, contractor pricing disclosure, payroll pricing disclosure, and minimum contract term disclosure.

• High: provider publishes a clear starting price and discloses at least three additional fee categories.
• Medium: provider publishes some pricing information but leaves significant fee categories undisclosed.
• Low: provider does not publish a starting price, or discloses fewer than two fee categories.

Ratings are applied by WhichPayroll researchers and reviewed quarterly. Providers may request a re-review by submitting updated public source documentation, and the same route is open for every input above. Send us the published page that contradicts our figure, and we will recompute the score and date the correction.

How Do We Handle Corrections?

WhichPayroll maintains a correction process for providers and third parties who identify factual errors. Corrections are accepted when accompanied by a verifiable public source. WhichPayroll does not remove evidence-backed observations solely on a provider’s request, only on verified correction. Superseded values are retained in the dataset change log with the date of correction.

If you spot a factual error, use the contact form with the dataset name, the specific field, the current value, the correct value, and the public source URL. WhichPayroll targets a 5 working day response for correction requests.

How Do We Maintain Commercial Neutrality?

WhichPayroll Research is funded by referral revenue from provider comparison pages, not by research subscriptions or provider sponsorship. Research findings are produced independently of commercial relationships. No provider pays for inclusion in or exclusion from WhichPayroll Research outputs.

Providers with commercial relationships are subject to the same evidence standards as all others. When you rely on WhichPayroll Research for procurement decisions, you can treat all provider scores as independent.