Mercer's compensation surveys are the largest of their kind. They cover thousands of companies across most sectors and most geographies, with decades of methodological refinement. For a Fortune 500 company with a dedicated C&B team, they are a useful reference and a reasonable line item.
For a 200-person SME with one HR generalist, two part-time recruiters, and a June 2026 deadline, the same product is a different question. The cost is significant; the deliverable is a sector-level distribution; the work to translate the distribution into operational salary bands is the same whether the survey is in hand or not.
The point of this article is not that Mercer surveys are bad. They are exactly what they are designed to be. The point is that what they are designed to be is not what most SMEs need.
How the data is actually collected
Survey data does not fall from the sky. It is collected through a process that determines its accuracy and its latency.
Each year, Mercer (and equivalents — WTW, Korn Ferry, Radford for tech) sends participation forms to HR teams at subscribing companies. The forms ask the HR respondent to fill in compensation data for each role at each level within their company. The HR respondent then estimates — using the company's HRIS, payroll exports, or recent offers as references — what the compensation looks like for "Senior Software Engineer," "Marketing Manager," and the other roles in the survey taxonomy.
The estimates are aggregated, anonymised, and published as sector-level distributions. The participating company gets the survey results 4–6 months after the cycle closes. Non-participating subscribers get the published results without contributing their own data.
Four steps where accuracy is shaped:
- Mapping company roles to survey roles. The HR respondent decides which of the company's roles correspond to which of the survey's standardised role descriptions. A company's "Senior Engineer III" might map to the survey's "Senior Software Engineer" — or might not.
- Compensation basis decisions. The respondent decides what counts as "base" vs "total comp." If the company has unusual variable pay structures (equity at startups, profit-sharing at family businesses), normalisation involves judgment.
- Population definition. Which employees count for which role. Tenured Senior Engineers vs new-hire Seniors vs Engineers acting as Seniors in a flat structure — the boundaries are interpretive.
- The lag. The data the respondent enters is "current as of survey reference date" — typically January or April of the reporting year. Publication is 4–6 months later. By the time a CHRO consults it, the underlying data is 9–12 months old.
What €50k actually buys
The total annual cost of a typical Mercer engagement for a 200-person SME, factoring access fees, role mapping, and the consultancy work to translate survey output into operational bands:
Whether this is "expensive" or "reasonable" depends on what it produces. For a 200-person SME, €50k is roughly the all-in cost of one HR generalist for six months. The question is whether the survey delivers six months of HR generalist value.
What it delivers, and what it doesn't
The Mercer engagement delivers three things that are genuinely valuable:
- Sector-level distribution per role × level. A statistically robust median and range across the survey participants. Useful as one external reference point in the salary band methodology.
- A polished deliverable. A bound document or interactive tool with role-by-role bands, comparable companies, market trends. Useful for board reporting and pay equity conversations with works councils.
- The credibility of the brand. "Our bands are calibrated against the Mercer survey" carries weight in some conversations. The signalling value is real.
The Mercer engagement does not deliver:
- Recent data. The 9–12 month lag is structural. The survey reports on last year's compensation; this year's market has moved.
- Company-specific operational reference. The output is a sector-wide distribution. Translating it into "what should we pay this specific candidate, for this specific role, given our specific cohort" is the consultancy step — and it is the same work that would be needed without the survey, just informed by a sector reference.
- The methodology behind the data. Subscribers see the aggregated results but cannot inspect the per-company inputs. The accuracy depends on the participating HR respondents — and the variability across respondents is not visible to the subscriber.
This article is not arguing that Mercer survey data is wrong. The aggregated medians and distributions are statistically defensible for the questions they answer. The argument is that the questions they answer are not always the questions the SME has. Mercer answers "what does the sector look like for this role." Most SMEs need to answer "what should we offer this candidate, today, given our existing team." The first is the input to the second; the second is the deliverable.
When the Mercer model is the right fit
Three signals indicate the survey is genuinely the best option for the company:
- Company size >500 FTE. The compensation team and the role count are large enough that the per-employee cost of the survey is small and the deliverable can be operationalised internally.
- Multi-geography operations. Comparing bands across countries benefits from the survey's standardised cross-geography methodology.
- Board / works council requirements that name a survey. Some governance frameworks explicitly require benchmarking against named surveys; the requirement makes the choice.
For SMEs without those signals — single geography, <500 FTE, board not asking for a named survey — the survey is one possible reference but not a necessary one. A self-built band structure calibrated against publicly observable signals (job adverts, recent offers, candidate-disclosed ranges) can satisfy the directive's requirement without the survey's cost.
Mercer surveys are not the wrong product. They are the right product for the companies they were designed for. The SME bracket is not one of those companies.
Where the diagnostic starts
The choice between buying a survey and building internally is the wrong frame. The right frame is what reference data the company needs to support its salary-band methodology. The directive accepts any methodology that is transparent, documented, gender-neutral, and applied consistently. Where the reference data comes from matters less than whether the methodology is intact.
The ReadinessCheck™ surfaces whether the methodology is in place — by axis of the directive's principal obligations. The decision about reference data sits downstream of that.
The reference matters. The methodology matters more.
ReadinessCheck™ takes about 20 minutes and requires no salary data. It produces an observational position view of where the company sits on pay-setting methodology, category construction, and the inputs the directive's reporting obligation depends on — independent of whether the company uses a survey, signals, or a hybrid.
Start the ReadinessCheck →