We ran this once,
with forty analysts.
In the 2000s Valuatum built a freelance equity-analyst network in Finland. Revenue sharing, per-company rankings, accuracy scoring against realised prices, and peer review as a paid role. Here is what happened, and why we are doing it again.
What it was
Valuatum provided the professional tools; freelance analysts used them to build equity analysis and updated it over the internet into a database our customers used. The work was part-time and remote — evenings, at home. Active investors, business controllers analysing their own competitors and customers, and finance and accounting students all did it.
The distribution was real. Full analyses went to paying institutional customers, and restricted versions reached a wide audience through online financial media — at the time, the front pages of Taloussanomat and Arvopaperi.
How the money worked
Analysts were paid by revenue sharing: half of the revenue generated by sales of their analysis. The full split was 50% to analysts, 10% to coaching analysts, and 40% to the system.
Revenue was first allocated to companies, then to analysts. Customers who bought analysis on specific companies made those companies more valuable to cover — one company could be worth ten times another. Revenue from customers who bought everything was split evenly across all covered companies, on the reasoning that a small company's analysis is relatively more valuable precisely because fewer people follow it.
Within a company, the majority of the revenue went to the best analyst. Rank one earned twice rank two, rank two twice rank three, and so on. Analysts with very poor points earned nothing.
How analyses were ranked
Every covered company was ranked separately, so one analyst held a different position for each company they followed. Points came from three categories:
- Administrator points — human review of the analysis, given by Valuatum or by senior analysts.
- Customer points — the readers who paid for it.
- Automatic points — computed from the accuracy of estimates and recommendations against what actually happened.
The weights were designed to shift over time. In the beginning almost everything was administrator points, because there was no data yet. As data and paying customers accumulated, automatic and customer points were meant to take over and the human weight to fall toward zero.
The admin scale had teeth. Above 1.1 meant an excellent analysis with something extraordinary in it: industry context, competitor positioning, informed corrections to the figures the company itself had published. Above 1.0 meant very good, informative estimate comments. Between 0.5 and 0.99 meant the essentials were there but thin. Below 0.5, the analysis was labelled unreliable, dropped out of the ranking lists, and earned nothing.
Automatic points: was the call right?
The most durable idea in the whole system. Monthly, every analysis was compared against the realised share-price move and scored against thresholds that depended on the recommendation — a Buy is right if the share rose, a Hold is right if it went nowhere, a Sell is right if it fell. The raw score became a percentile against every other analysis that month, and months were combined with the recent ones weighted heavier, so old mistakes faded and a bad month was recoverable.
"This analysis ranks better than 76% of all analyses" is a sentence that needs no explaining. That is why the percentile, not the raw number, was the thing shown.
Coaching analysts: peer review as a paid role
Any analyst holding at least one analysis rated 1.1 or better could apply to become a coaching analyst and review other people's work. Coaching analysts shared 10% of the cash flow.
The loop mattered as much as the money. An analyst signalled that their work was ready by making a feedback request; coaching analysts were notified, one of them booked the request so two people would not duplicate the work, and the reviewer gave both points and concrete improvement suggestions. The analyst fixed it and asked again, and the reviewer re-rated. Rating alone ranks people. Rating plus a fix path is a ladder.
What actually came of it
Around forty analysts worked in the network in Finland. Within nine months, six of them were hired as analysts by brokers or corporate finance units.
That is the number that matters, and it explains what the participants were really buying: not the fees, but a public, dated, scored track record they could put in front of an employer. An objective accuracy score against everyone else in the pool is a stronger claim than anything a CV can assert.
What changed, and what did not
Back then the analyst built the entire model in Excel before publishing a single number, which is exactly why they earned half of the revenue. The engine does that part now, in minutes, for any listed company — and the split stayed at half anyway. Not out of nostalgia: the modelling was never the scarce part, and paying for the scarce part is what the number was always for.
What did not change is everything the money was measuring. There is still exactly one scarce input, and it is judgement: which assumption is wrong, which risk the market is ignoring, which peer group is the honest one. A machine that can produce a competent analysis of any company on demand makes that judgement more valuable, not less, because it is the only part left that distinguishes one analysis from another.
So the mechanics carry over almost unchanged: ranking per company rather than one global leaderboard, scoring recommendations against realised prices, a quality floor that demotes instead of deleting, peer review as a paid role, and reader ratings that only count when they come with a written comment.