DATA. DECIDE.
Data and decisions that move the needle. More traffic, revenue, satisfaction. Dacide builds Company Brains, runs swarm simulations and sets up board decisions that leave a record.
/ track record
2019. Before anyone outside tech had heard of an LLM or GPT, we ran our first growth hacking programme on machine learning. Within a few months it added 5 million CHF in revenue for our client. We were paid an 8% share of the profit.
Today is . The needle still moves the same way. The tools are an order of magnitude better.
This is the board's view of one decision.
In 1999, serious people said they would never buy anything online. Today, serious people say agents will never run real work. Most boards still decide on memory, a slide deck and the loudest voice in the room. When a shareholder, a court or an auditor asks how, the minutes rarely answer. Opinions are cheap. Evidence is a choice.
Three fields, tested in audits.
Your company's memory, with footnotes.
A Company Brain compiles what your people wrote into pages that cite their source and show who checked them.
Open →A market reaction you can run again.
Many simulated stakeholders run through your plan and show how far their reactions scatter. Labelled as simulation every time.
Open →A decision record the next chair can read.
The question, the evidence, the probability and the dissent are written down before the vote.
Open →Real books, read without names.
We strip every name from property and yacht accounts, then grade each finding against the books.
Open →Six numbers from our pages, each with its source.
drag or scrollto 155'000
Don't read about it. Run it.
Simulation · scripted agents · not a forecastA wind tunnel, not a crystal ball.
Set four levers and see which segments hold, strain or break.
No archetype matched. Pick the closest one:
This device shows a static view of the same model and readout.
1 seed (k = 7)
Simulation
What this model does not know
- Agents are scripted archetypes, not your market.
- Coefficients are illustrative, not fitted to data.
- A real engagement replaces both with your evidence and a full swarm run. How a full run works
Show the formula
L = (lever - 50) / 50 four levers, each 0 to 100 z = b + wm·Lm + ws·Ls + wp·Lp + wr·Lr + ε ε ~ Normal(0, 0.35), seed k: mulberry32(k), Box-Muller σ = 1 / (1 + e^-z) holds σ < 0.40 · strains 0.40 to 0.70 · breaks σ ≥ 0.70 drag D = round(100 · Σ share · σ)
The flow shows seed k = 7. Run 20 seeds uses k = 1 to 20 and reads P10, P50 and P90 per segment; status and drag use P50. Coefficients b and w are set by hand.
Computed on this device.
| Segment | Share | Stress σ | Status | Scripted line |
|---|
What we will not claim.
Teams we have worked with.
Three ways to start.
scope and price in writing before anything beginsRehearse one decision
Your evidence, a full swarm run, a board-ready record.
Audit one set of books
Property or yacht accounts, anonymised before any AI reads them.
Build one Company Brain
One knowledge domain compiled, cited and versioned.
We do not ask for your email first.