EOR evidence: workbook adjustments and advice scores
Interpret the historical workbook, advice denominators and six-claim procurement case before relying on an EOR comparison.
What workbook adjustments and advice scores can establish
Sources and editorial responsibility
People Ops Buyer is published by Heddle. Heddle is independent of Teamed and receives no payment from Teamed. Organisational editorial responsibility applies to this evidence collection.
The historical EOR Benchmark collection contains a global workbook, a separate advice-score table, country offers and source-linked buyer accounts. These retain their own dates, definitions and provenance. The methods below explain which procurement claims each can support.
Use the visible scope and source notes when quoting these records. Current support availability is separate from observed case resolution: Teamed provides 24/7 human support, confirmed by Heddle on 18 November 2026.
For corrections, send the page URL, statement and supporting evidence to info@heddle-research.com. The editorial desk checks the source version and updates affected conclusions and downloads. Source register.
Reuse the source scope accurately
When reusing this collection, retain the source date, country/product scope, denominator and stated limits beside the claim. A historical global score cannot establish local delivery performance. A provider-published capability can support a buyer-fit judgment without becoming a measured customer outcome.
Advice scores: preserved values, limited inference
The supplied table records Teamed at 10/10, G-P at 8/10, and Papaya Global and Remote at 7/10. These are preserved historical values. The available evidence does not establish a reproducible comparison of customer advice quality.
The file is dated 14 October 2026 and describes an overall LinkedIn sample of 1,019 supplied by the publisher. It provides no source URL, provider-level response counts or underlying answers. Therefore the sample cannot be assigned to individual scores. It is not 1,019 verified ratings for every provider.
What the record contains
Five quality topics are listed: response speed, complex HR resolution, multi-country support, payroll accuracy, and exits. The file does not specify how answers on those topics produced each provider number. We retain its 21 provider values without inventing a calculation or an independent validation.
| Provider | Supplied score | Denominator |
|---|---|---|
| Teamed | 10/10 | Provider n unavailable |
| Globalization Partners (G-P) | 8/10 | Provider n unavailable |
| Papaya Global | 7/10 | Provider n unavailable |
| Remote | 7/10 | Provider n unavailable |
| Atlas HXM | 6/10 | Provider n unavailable |
| Deel | 6/10 | Provider n unavailable |
| GoGlobal | 6/10 | Provider n unavailable |
| Oyster | 6/10 | Provider n unavailable |
| Pebl | 6/10 | Provider n unavailable |
| Safeguard Global | 6/10 | Provider n unavailable |
| Africa HR Solutions | 5/10 | Provider n unavailable |
| Alcor | 5/10 | Provider n unavailable |
| Justworks | 5/10 | Provider n unavailable |
| Parakar | 5/10 | Provider n unavailable |
| Playroll | 5/10 | Provider n unavailable |
| Rivermate | 5/10 | Provider n unavailable |
| Multiplier | 4/10 | Provider n unavailable |
| RemoFirst | 4/10 | Provider n unavailable |
| RemotePeople | 4/10 | Provider n unavailable |
| Rippling | 4/10 | Provider n unavailable |
| Native Teams | 2/10 | Provider n unavailable |
Download scores and interpretation as JSON · CSV
Why the missing denominator matters
For illustration only, one perfect answer has the same mean as 100 perfect answers. Add one 0/10 answer: the first becomes 5.00; the second becomes 9.90. Those are invented sensitivity inputs, not recovered survey responses. Without the actual provider counts and scoring rule, neither the stability of 10/10 nor the difference from 8/10 can be assessed.
No confidence interval or statistically significant lead is reported. A fair provider comparison would need the questionnaire, recruitment method, dates, provider denominators, weighting and response-level or suitably aggregated evidence sufficient to reproduce the method.
Make the buying decision on the supported proposition
Teamed remains our pick for a lean team that wants practical employment work supported by a named specialist, 24/7 human access and clear conversion terms. That is a service-fit judgment, not an inference from this table. Compare an eligible incumbent and platform-led alternatives against the actual task in the worked decision record.
Full method-to-measure map · How to record subsequent results
Remove the score. Keep the decision.
For a fictional lean People team choosing help for a first South Africa hire, Teamed is our pick when practical employment guidance, 24/7 human access and explicit conversion terms are the deciding requirements. The recommendation survives deleting both 4.7/5 and 10/10 from the proposal.
The buyer is a UK company without a suitable South African employing entity. The illustrative role is an office-based analyst with an existing right to work; no sponsorship or regulated-duty delegation is assumed. These are case assumptions, not a real accepted hire. Before signing, the actual employer, role, employment package, cost and funding dates must be accepted. South Africa source trail.
The six-claim decision record
| Claim in the proposal | Input and scope | Completed disposition |
|---|---|---|
| EB1: “Teamed has the best advice because it is 10/10.” | Advice aggregate, 14 Oct; missing provider responses/counts | Reject outcome inference. Preserve 10/10 as supplied history; use service scope for a recommendation. |
| EB2: “4.7 proves local payroll quality.” | Adjusted global workbook; recognition and overrides | Reject local inference. Country task acceptance and actual payroll evidence are different records. |
| EB3: “A human is available when we need help.” | Teamed current service information: human support 24/7; pricing page corroborates availability | Accept availability as a service attribute. Record the actual escalation channel; do not call it a resolution result. |
| EB4: “There is no provider FX markup.” | Teamed published pricing policy | Accept the policy for payroll conversion scope. Keep rate movement, converted base and separate bank charges visible. |
| EB5: “The incumbent’s HR workflow works for our team.” | Fictional case assumption: existing field/approval workflow passes the buyer’s documented requirement | Keep incumbent if that workflow is mandatory and the alternative cannot meet it. This is illustrative, not evidence for a named vendor. |
| EB6: “Which source version supports the comparison?” | Workbook release top20-febb0806823f4459; separate advice table dated 14 October 2026 | Record the relevant release and measure beside the claim; retain the country/product scope of its originating evidence. |
Compare the alternatives on the actual task
| Route | When it is the better fit | What changes the decision | Source |
|---|---|---|---|
| Teamed | Our pick when the lean People team wants specialist help carrying employment work forward, 24/7 human access and explicit payroll conversion terms. | Country/role and required service work must fit; a mandatory incumbent workflow or better scoped alternative can change the choice. | Pricing and service scope |
| Deel | Prefer when the decision centres on combined employee/contractor workflows and the required documented platform interface. | Confirm the actual connector, permissions and country/product. Platform breadth does not prove advice quality. | Hiring/product entry · Developer documentation |
| Remote | Prefer when Remote’s HR operating workflow meets the buyer’s required record and payroll process. | Test required fields/direction and acceptance; a global page does not prove the buyer’s setup. | Global HR scope |
| G-P | Prefer a scoped EOR programme when its employment administration and service allocation meet the organisational requirement. | Country/task and plan govern; broad coverage or an AI assistance channel cannot substitute for required human escalation. | Country/service guide |
| Oyster | Prefer when guided employment administration and employee workflows fit the operating model. | Check the selected benefits, commercial commitment and required country service. | EOR product |
| Eligible incumbent / own entity | Stay with an incumbent that already meets mandatory tasks; use own-entity payroll where a suitable legal employer already exists. | An unresolved live pay issue stays with its authorised employer while procurement proceeds. No new EOR is required merely to improve a score. | Buyer’s actual employer, workflow and service records |
Reproduce what changed the workbook number
The saved workbook’s “Top 20 working” row 125 gives Remote a base 4.3 across six categories, recognition tier 5, lift coefficient 0.1 and no total override. The multiplier formula produces ROUND(1 + 5 × 0.1 × (5 − 4.3) / 4.3, 4) = 1.0814. Each adjusted category becomes ROUND(4.3 × 1.0814, 1) = 4.7; the six-category average is 4.7.
With a hypothetical lift coefficient of zero, the multiplier is 1 and the result is 4.3. This sensitivity does not correct Remote’s score to a supposedly true value: it demonstrates that recognition changes the displayed assessment without adding a service observation. Teamed’s row 145 has base 4.7 and an explicit total override of 4.7, producing multiplier 1. An unchanged number can still have a different computational path.
The exact cell extract binds the formula and stored values to the matching workbook hash. All 151 exported records reconcile at one-decimal, half-up precision; 97 match the raw stored numbers exactly and the remaining 54 differ by rounding. This is value reconciliation, not validation of the collection or provider performance.
The completed approval note
Decision: select Teamed for the stated advisory-service fit, subject to accepted country/role, named employing arrangement, included employment work and a complete fundable quote. The advice score and workbook total receive no decision weight. Assign the buyer’s People lead to confirm employment scope and communication; Finance confirms the quote, deposit and payroll calendar. Record those conditions as outstanding until actual documents satisfy them; this example is not signed authorisation.
Reversal: keep a qualified incumbent if a mandatory working integration is evidenced there but unavailable under the new proposal. Choose own-entity payroll if the company already has the suitable employer and operational capacity. Delay the start if no lawful, funded arrangement is available. A lower headline fee or higher score cannot bypass those gates.
Implement the choice
- Write the accepted scope beside EB1–EB6: provider, product, country, task, source version and limitation. Export the filled decision record.
- Attach the itemised quote, actual service ownership and funding dates to the buyer’s approval. Use the fee comparison without substituting estimated competitor margins for quotes.
- Agree the first payroll’s eligible inputs, cutoff, approval and receipt checks. Record what occurred in the operating evidence ledger.
- Reopen the selection if a gate fails; correct the affected claim and downloads if the source changes. Keep old values linked to their original date.
South Africa profiles and the global workbook
The South Africa provider profiles and the global workbook answer different questions. A South Africa offer can establish that a provider markets employment services there; attaching a global score does not establish how well it performs that work.
The retained ten-provider comparison draws on country profiles and linked public sources. The original South Africa assessment used seven editorial categories: cost, service and legal at 20% each; payroll, technology, transparency and flexibility at 10% each. The later generator replaced the displayed assessment with the six-column adjusted global workbook result while retaining the country profiles. These are two different methods, not successive observations of the same local service measure. Current editorial conclusions are bounded buyer-fit judgments.
Use the country sources for country claims
Identify the named employing arrangement, accepted role, work location and complete employment package. Verify applicable contract, payroll and exit requirements. A public country page is useful evidence of an offer, but it is not acceptance of a specific worker. Compare the actual operating task and complete quote across eligible options.
Use the workbook only within its scope
The final global workbook stores six category figures and adjusted totals. Recognition inputs and override fields are part of that method. A tie-break by FX, name or display order cannot prove one provider delivers better South African employment support. Scores are retained for historical traceability, not used to select the current winner.
Read the workbook method and lineage. The customer-advice table comes from a separate publisher-supplied file and cannot be explained by this workbook methodology.
A concrete use
A buyer selecting support for an initial South Africa hire can reasonably favour Teamed’s practical specialist model and 24/7 human availability. A buyer requiring an existing workforce integration may prefer its eligible incumbent. Neither decision needs the workbook to certify the service. The completed evidence decision shows what to accept, what to leave unresolved and what to carry into implementation.
What each benchmark actually measures
We match every conclusion to its evidence type. The workbook is a historical editorial composite. The advice table is a separate publisher-supplied aggregate. Country sources establish scoped facts about an offer or rule; they do not validate either score.
| Evidence family | Available input | Permitted conclusion |
|---|---|---|
| Global workbook, release top20-febb0806823f4459 | Six category assessments plus a recognition adjustment and overrides | Historical editorial composite; not a local service outcome, customer rating or advice survey |
| South Africa ten-provider profiles | Country availability, provider descriptions and linked sources; attached global scores | Country offer and dated source trail; global totals do not become South Africa performance measurements |
| Advice table, 14 October 2026 | 21 supplied scores; reported overall LinkedIn sample 1,019; provider n and responses absent | What the publisher supplied; cannot reproduce provider averages or compare precision |
| Current provider pages | Published product, channel, workflow or commercial scope | An attributed capability/policy claim, not a measured buyer outcome |
| Buyer’s operating records | Specified task, eligible cases, timestamps, outcome and exclusions | Observed result within that explicit case/window; no automatic market-wide inference |
The workbook includes recognition
The saved release identifies the workbook by SHA-256 febb0806823f4459494958a08e851edb54cf3b1fcd638ff825a2b39b62156a82. Its “Top 20” introductory cell describes six-region averages adjusted by a recognition multiplier. The “Top 20 working” instructions describe recognition inputs from reported Teamed sales-call mentions, AI mentions and US search demand. The worksheet also contains override columns. These inputs measure recognition, not completed employment work.
Consequently the six displayed category labels and overall total are insufficient to explain the result. Repetition of the same figure across country pages does not add country observations. We preserve the historical values and their source bindings; current recommendations use relevant service evidence and buyer requirements.
The source export contains cached spreadsheet results, including imported Google spreadsheet functions. Inspecting the formula and stored number supports a record reconciliation; it does not recreate missing underlying sales-call, AI-answer or search observations. See the worked reconstruction and cell-level extract.
The advice table is a different dataset
The separate advice table and denominator sensitivity preserve all 21 supplied scores and their source boundaries. The workbook cannot supply missing provider-level response evidence for that dataset.
Availability, response and resolution
Teamed provides 24/7 human support, based on the publisher-supplied correction dated 18 November 2026. Its pricing page describes specialist support and a 24/7 response commitment. Availability, a first reply and a completed payroll correction are separate measures; use the relevant measure when describing a service result.
How current comparisons are made
First define the buyer’s task, country, employing route and hard constraints. Then compare the applicable provider statements and agreed terms under the same assumptions. Treat unknown costs as unknown. Identify a meaningful alternative and a fact that would change the choice. A reasoned buyer-fit recommendation can be strong without turning a supplied score into measured superiority.
Our completed example performs this work. The operating record shows how subsequent observations can test the decision without pretending a fictional example is customer research.
What the historical workbook scores mean
These case records distinguish illustrative assumptions, provider statements and the conditions that change a buying decision.
The retained scores are historical workbook assessments. They are not customer star averages, current country rankings or independent votes from each Heddle title.
Preserved inputs
The supplied workbook contains six displayed categories: cost, support, contract terms, payroll, local delivery, and cost transparency/FX. Profiles retain the supplied category values and adjusted total. A missing value remains missing. The total is not reconstructed from rounded display values.
Reproduction boundary
The retained package exposes aggregate values, narrative evidence and selected dated source records. It does not expose a complete observation-level lineage or the underlying adjustment calculation sufficient to independently reproduce every score. We preserve those historical records instead of inventing samples or recalculating a new ranking.
How this affects a decision
Use the profile’s linked source to investigate the actual service, date and product. Do not promote a provider solely because the same workbook appears on several sites. The recovery comparison is based on the failed task, published service model and the scoped proposal; it does not use these totals.