The same shape of decision, in every industry.
AnantCausal is not built for a sector. It is built for a shape of decision: discrete, defensible, and not yet measured — where a model can be elicited from what the organization already knows, and where someone will later be held to the number.
Where a decision is continuous, or an experiment already settled it, or nobody has to defend it, this is the wrong tool. Four verticals follow, each with a worked example run against a shipped scenario — including the places the answer is no.
What makes a decision fit
Four questions. If the answer to any of them is no, this will not help and the page should say so rather than finding a way.
- 01
Is there a decision, or only a question?
Somebody has to be choosing between options with a horizon attached. "Why did this happen?" is a question; "should we move the effort here?" is a decision, and only the second one has an answer you can act on.
- 02
Can the structure be elicited?
The variables and their relationships have to be knowable from documents, experts or data — not only from a trial nobody ran. If the structure is genuinely unknown, structure learning from data is the honest answer, and that is a different tool.
- 03
Is the outcome discrete?
Variables here are discrete, and a continuous outcome reaches the model through binning, which is lossy and labeled as such. That is a real limitation, not a presentational one.
- 04
Will someone be held to the number?
This is the one that matters. If nobody will ever ask where it came from, buy something cheaper — and if they will, the record is the product, not the number.
Every vertical below is one of those four shapes. Where a decision is none of them, the industries list is the wrong place on this site to be looking.
The four verticals
Each has a shipped scenario in the engine with a known structure, so the numbers are checkable rather than asserted. Each page shows the effect, the band, the control advice, and one question the scenario cannot answer.
Healthcare payer
Where should documentation effort go?
Where to spend clinical-documentation effort: which conditions, which populations, which review process. The options are two levers on the same claim, and the question is which one moves denial.
Read the worked example
Wealth management
Which clients should the desk call?
Which clients to call, and with what offer. The capacity is small and the value per client is large, so the ranking matters more than the average — and the ranking is a causal quantity per subgroup, not a propensity score.
Read the worked example
Manufacturing operations
Is operator training worth the investment?
Whether to train operators on the line, and on which shift. The action is the training standard; the outcome is a defect; the unit is a batch.
Read the worked example
Clinical trials
What drives treatment response?
Whether to invest in an adherence program, and whether dose escalation is worth it alongside. The action is adherence; the outcome is response; the unit is a participant.
Read the worked example
Why the numbers are checkable
Each scenario is a small generative model — variables, levels, and the mechanism that produces each from its parents — so the table it emits has a known structure. The screen can then be scored against that structure rather than eyeballed, which is how precision and recall appear on these pages at all. You can download the data for any scenario and reproduce every figure above, including the misses.
Four verticals, four decisions, and the same method. What changes between them is the structure; what does not is that the structure is written down, the refusals are reported, and the numbers are checkable.