The comparison matrix
Read down the column that matches your question, not the one that matches your budget.
| Platform | Research firm | Consultancy | In-house team | |
|---|---|---|---|---|
| Data basis | Modelled from public sources | Primary research, corroborated | Mixed; often licensed or subcontracted | Usually licensed, plus internal data |
| Breadth | Very high | Scoped deliberately | Moderate | Moderate to high |
| Depth on a named competitor | Low to moderate | High | Variable | Moderate |
| Verification | Generally none | Corroboration standard | Varies by firm | Depends on analyst time |
| Speed to first answer | Minutes | 3–8 weeks | Weeks to months | Hours to days |
| Cost basis | Annual licence per seat | Fixed fee per project | Day rate, typically the highest | Fully-loaded salary plus tooling |
| Marginal cost of one more question | Near zero | Full project fee | Full engagement fee | Near zero |
| Independence from internal politics | High | High | High | Lower by construction |
| Institutional memory | Low | Low unless retained | Low | High |
| Accountability for a number | None claimed | Named analyst | Partner-level | Internal |
| Best at | Monitoring and first-pass sizing | Decisions that must be defensible | Board framing and recommendation | Continuous internal questions |
Talent intelligence platforms
Software licences covering many markets at once.
What they do well
Breadth no research budget can match. Trend lines over time. A first-pass market size in minutes rather than weeks. Continuous availability, so a question at 4pm does not become a project. Consistent methodology across markets, which makes relative comparison genuinely useful.
Where they structurally fail
The data is modelled from public profiles and postings, which means it inherits whatever is stale, absent or self-reported in those sources. Coverage is uneven: excellent for large technology employers in English-speaking markets, thin for the German Mittelstand, Gulf free zones and specialist industrial populations.
It cannot verify a specific headcount, a specific reporting line or what an employer actually paid to close a hire.
What they cost
Annual licence, usually priced per seat with module add-ons. The marginal cost of the next question is near zero, which is the model’s real advantage.
The cost that catches people out is the analyst needed to interpret it. A licence without an analyst tends to disappoint.
Our honest position: if your need is continuous monitoring across many markets, buy the platform. We are a worse purchase for that question and we would rather say so than sell you a project that does not fit.
Research firms
Commissioned primary research, scoped per question. This is what we do.
What they do well
Verification. A named analyst researches the market directly and corroborates each material finding against an independent second source, which is the only way to establish what is true inside a named employer today.
Depth on competitors, reconstructed compensation, coverage of markets where public data is thin, and accountability for a number when someone challenges it.
Where they structurally fail
Breadth and speed. A project takes three to eight weeks and covers what it was scoped to cover. If you want forty markets monitored continuously, this is the wrong model and will be expensive.
There is also no institutional memory unless the relationship is retained — each project starts from your brief rather than from accumulated context.
What they cost
Fixed fee per project: $6,000 to $15,000 for a brief, $18,000 to $45,000 for a study, $60,000 to $120,000 a year retained. The marginal cost of an extra question is a new project fee, which is the model’s real disadvantage.
Full bands by currency are on the cost page.
Strategy consultancies
Workforce and organisation engagements delivered at board level.
What they do well
Framing, board credibility and a recommendation someone senior will stand behind. Integration of workforce questions into a wider strategic or transformation agenda, which neither a platform nor a research firm is positioned to do.
Where to look carefully
Who collects the primary labour market data. It is frequently licensed from a platform or subcontracted to a research firm, which means the verification question lands somewhere you cannot see.
That is not a criticism of the model — it is a reason to ask the question explicitly before you buy.
What they cost
Day rates or fixed engagement fees, typically the highest of the four models by a considerable margin. You are paying for framing, seniority and accountability at board level, with the research layer often bought in at a fraction of the fee.
In-house teams
A dedicated internal talent intelligence function.
What they do well
Institutional memory, instant availability and the ability to join external market data to internal workforce data — which no external provider can do as well, because we cannot see inside your organisation.
At high question volume the marginal cost of an additional question is near zero, which is decisive.
Where they structurally struggle
Independence. An internal team reporting into the function whose plan it is testing is in a difficult position when the evidence is unwelcome, and everyone involved knows it.
Also specialist coverage: one or two analysts cannot hold deep knowledge of eight markets, several languages and a dozen sectors at once.
What they cost
Fully-loaded salary per analyst plus tooling. Experienced talent intelligence analysts command senior salaries in every market we publish for, so a team of two or three carries a six-figure annual cost before a licence is bought.
The business case turns entirely on question volume, and should be run with fully-loaded cost on both sides rather than licence fee against project fee.
How to choose
Five questions that usually settle it.
- Is the question continuous or episodic? Continuous favours a platform or an in-house team. Episodic favours commissioned research.
- Does the answer have to be defensible? If it goes into a board paper, an investment committee or a remuneration committee, you need verification and a named analyst behind the number.
- How many substantial questions a year? Below roughly a dozen, commissioning is cheaper and deeper. Above that, build internally and commission for depth.
- How well covered is your market by public data? Large employers in English-speaking markets are well covered by modelled data. The German Mittelstand, Gulf free zones, Belgian language-split populations and specialist industrial markets are not.
- Does it need to be visibly independent? If the evidence might contradict a plan someone senior has already announced, external research survives the conversation better than an internal memo.
Six questions to ask any provider
Including us. The answers tell you quickly which model you are dealing with.
- Where did this specific number come from, and when was it collected? If this cannot be answered per figure, the data is modelled.
- Is it corroborated, single-sourced or estimated? A provider without a confidence standard has not thought about this, or does not want you to.
- What is your lawful basis in each market in scope? It should be answerable in a sentence per jurisdiction, in writing, before work begins.
- Do I own the dataset permanently? Ownership and access are different things, and the difference appears when you stop paying.
- Who is the named analyst, and can I speak to them? Not an account manager. The person who collected the data.
- What will you tell me you cannot establish? A provider who claims no limitations has either not done the work or is not telling you about it.
Most mature functions run more than one
The pattern we see most often in organisations that do this well is not a choice between models but a deliberate combination: a platform licence for breadth and continuous monitoring, one or two internal analysts to interpret it and join it to internal data, and commissioned research for the decisions that have to withstand challenge — a location case, a reward review, due diligence on a target, or depth on the four competitors that actually matter.
That combination costs more than any single model and less than getting one big decision wrong. It also removes the awkward dynamic where the only available evidence comes from the party with an interest in the conclusion.
If you already licence a platform, the useful conversation is not whether to replace it. It is which three questions a year genuinely need verification, and whether the answers would change what you do.
Comparison FAQs
What is a talent intelligence platform?
A talent intelligence platform is a software product that aggregates labour market data — typically public professional profiles, job postings, company information and official statistics — and presents it through dashboards, market sizing tools and exports, usually on an annual per-seat licence.
Its strength is breadth and continuity: it covers many markets at once and is available the moment a question arises. Its limitation is that the underlying data is modelled from sources it does not verify, so it describes shape well and establishes fact poorly.
Is a platform or a research firm better?
They answer different questions, and the honest answer is that it depends on what you are deciding. If you need continuous monitoring across many markets at low marginal cost, a platform is the better purchase and we will say so.
If you need a verified answer that will be interrogated by a board, an investment committee or a remuneration committee — particularly on a named competitor or on what people are actually paid — commissioned research is the better purchase. Many mature functions run both.
When is software clearly the right buy?
When your questions are continuous rather than episodic, when breadth matters more than depth, when you need a first-pass answer in minutes rather than weeks, when you are monitoring trends rather than establishing facts, and when you have an analyst internally who can interpret what the dashboard shows.
A platform without an analyst usually disappoints, because the product produces data and the decision needs interpretation.
When is commissioned research clearly the right buy?
When a specific decision has to be defensible. When the question concerns a named competitor’s structure. When you need what employers actually pay rather than what postings advertise. When the market is one where public professional data is thin — much of the German Mittelstand, parts of the Gulf, specialist industrial populations. And when you need someone to be accountable for the number.
What does an in-house talent intelligence team cost to run?
More than the licence line suggests. The dominant cost is people: experienced talent intelligence analysts command senior salaries in every market we publish for, and a team of two or three carries a fully-loaded cost well into six figures before tooling.
At high question volume that is excellent value, because the marginal cost of the twentieth question is near zero. At low volume it is poor value, because the fixed cost is carried whether or not questions arrive.
Can a research firm replace a platform?
Not economically, for breadth. If you want a standing view of forty markets refreshed continuously, commissioning that as research would cost far more than a licence and would be slower.
What a research firm replaces is the illusion that the platform's numbers are verified. Use the platform for breadth and the research for the decisions where being wrong is expensive.
How do I evaluate a talent intelligence provider?
Six questions. Where does this specific number come from, and when was it collected? Is it corroborated, single-sourced or modelled? What is your lawful basis in each market in scope? Do I own the dataset permanently? Who is the named analyst and can I speak to them? And what will you tell me you cannot establish?
A provider who answers all six without hesitation is doing the work. One who cannot answer the first two per figure is reselling a dataset.
Do consultancies do talent intelligence?
Strategy consultancies deliver workforce and organisation work at board level, and some of it is excellent. The practical question is who does the underlying labour market research, because it is frequently subcontracted or built from licensed data rather than collected first-hand.
That is not a criticism, but it changes what you are paying for: you are buying framing and a recommendation at a premium rate, with the primary research layer bought in. Ask who collects the data and on what basis.
Should we build an in-house team or outsource?
Volume decides it. Below roughly a dozen substantial questions a year, commissioning is cheaper and gives you deeper answers per question. Above that, an in-house team with a platform licence is more economical, and the institutional memory is genuinely valuable.
The most common mature pattern is a small in-house team handling continuous questions, commissioning external research for depth on named competitors, for markets outside their expertise, and for anything that needs to be visibly independent of internal politics.
Is modelled data unreliable?
Not unreliable — just different. Modelled data is excellent for shape, direction and relative comparison, and it covers ground no research budget could reach. It is weak at point facts: a specific headcount, a specific pay figure, a specific reporting line.
The failure mode is not the data, it is using it as if it were verified. A modelled estimate presented in a board paper as a fact is a risk the platform never claimed to carry for you.
Can you work alongside the platform we already licence?
Yes, and a good share of our work is exactly that. Clients use their platform for breadth and commission us for the four competitors that matter, the market where the platform’s coverage is thin, or the compensation reconstruction that no modelled dataset can produce.
We will also tell you when your existing licence already answers the question, which happens often enough to be worth saying out loud.
What about free sources?
Official statistics, company registers and regulatory registers are free, authoritative and genuinely useful for framing a population — we use them in every engagement and cite them on every market page.
What they cannot give you is the current picture inside a named employer, because they lag and they aggregate. Start free, then pay for the layer that free sources structurally cannot provide.