The panel
The panel is a network of service providers, such as tax advisors, insurance brokers or IT providers, whose daily work produces private data about their clients before it becomes public. Read together and calibrated over years, it shows where a market is moving months before any statistic does.
Worked example
One tax advisor is not a panel, and neither are 50. Austria has about 3,000 tax advisors; a platform used by 1,500 of them reaches half the country's small-business bookkeeping in one hop. That is a panel: cash cycles, margin trends and hiring by sector, visible before they reach any filing.
The same threshold exists elsewhere. Germany has about 46,000 insurance brokers. The US has about 50,000 property-management firms running more than 20 million homes. One IT provider is noise; 200 across a national market show which sectors are hiring, which are hit by ransomware and which are cutting back.
What it holds
| Layer | What it shows | Who it serves |
|---|---|---|
| Engagement | Who uses what, how often, for how many clients | Due diligence: proof the network is real |
| Practice signals | Staffing changes, client counts, owner age | Finding firms whose owners want to sell, only where they opted in |
| Meaning | The aggregated financial reality of the clients behind the firms | Banks, insurers, funds and, later, states |
How it differs from
- A data set. A data set is a snapshot. A panel is valued by length and freshness together: one quarter of data is noise; five years plus this morning's feed is an asset.
- A bank's view. A bank sees its own customers' payments without knowing what they mean. The bookkeeper knows whether €50,000 coming in is revenue, a loan or the owner's own money.
- AI agents inside firms. Agents see how a company works. They do not see what it earns, owes and pays.
Questions
Who pays for a panel? Insurers pricing small-business risk, banks pricing credit, private equity firms sourcing deals and, eventually, central banks and agencies that today read statistics months late.
Doesn't the data go stale? Single data points do. The panel does not, because its value is the history that makes today's number readable. ADP's monthly jobs report moves markets because ADP has calibrated it against official revisions for decades.
Can a competitor buy one later? No. A panel that started recording in 2026 can never be caught by one that starts in 2028.
Whose data is it? Only anonymized, aggregated signal. Consent sits with the end customer, and anything an owner tells an advisor outside an explicit opt-in never reaches a buyer.
Origin
Defined by Viktoria Izdebska in The Prediction Economy, chapters III to V (2026).