Knowledge Centre
Compound PatternsEnterprise Intelligence··4 min read

What are Compound Patterns?

The most dangerous condition in an enterprise is the one in which everyone is right. Legal is right that the contract is signable. The people team is right that the hiring plan is broadly on track. Technology is right that the migration is proceeding. Every function reports green because every function, judged on its own terms, is green. The problem lives in the combination, and the combination belongs to nobody.

Compound Patterns are conditions that only become visible when signals from several functions are read together: facts that are individually acceptable and jointly a problem. A contract commitment, a hiring gap and a technology date can each pass their own review and still, in combination, describe a promise the firm cannot keep. The pattern is real, it is specific, and until someone reads across the functions, it is invisible.

Why no single function can see them

Enterprises are organised vertically, and this kind of risk is horizontal. Each function builds tools, thresholds and review cadences scoped to its own domain. A contract clause is evaluated against legal exposure, not against the hiring pipeline. A role slipping a few weeks is judged against the workforce plan, not against a commencement date buried in clause seven of a services agreement. Nobody is negligent. The pattern simply falls outside every mandate at once.

The traditional patch is the cross-functional meeting, and it fails for an honest reason: it samples the combinations occasionally, through human memory, at whatever altitude the attendees happen to fly. The combinations that matter are specific, quantitative and perishable. They involve a particular clause, a particular role and a particular date, and they stop being fixable on a timescale no monthly steering call can match. Reading the functions together continuously is a different kind of work, which is why it belongs to cross-module intelligence rather than to a calendar invitation.

What makes a detected pattern trustworthy

Reading signals together is also easy to do badly. Correlation is cheap, and a pattern you cannot interrogate is not intelligence, it is anxiety with a timestamp. In ONX, cross-domain detectors watch for these conditions, and every pattern they raise is held to four properties:

  • Explainable. The pattern names exactly which facts, from which functions, combine, and why that combination matters.
  • Challengeable. Any fact the pattern rests on can be disputed, and the pattern must survive the dispute or be withdrawn.
  • Traceable. Every input fact carries its provenance and its evidence quality state: measured, modelled, inferred, stated or unmeasured. A pattern built on stated facts announces itself as one.
  • Falsifiable. The pattern declares what evidence would prove it wrong. If that evidence arrives, the pattern is resolved, not quietly forgotten.

Detected patterns also have a lifecycle: raised, challenged or confirmed, then resolved or expired. A pattern that cannot die accumulates as noise, and noise is how monitoring systems train their owners to stop looking.

One concrete example

Clearly illustrative, with no customer implied. A firm signs a services agreement with a commencement clause fixing a start date. Legal reviews it: standard language, acceptable risk. The workforce plan shows one specialist role still mid-hiring, forecast to fill a few weeks after that clause takes effect. On its own, tolerable: roles slip all the time. The technology plan has an integration completing a fortnight after the workforce date. Also fine, on its own. Read together, the three facts say something none of them says alone: the earliest date the firm can actually stand the service up falls after the date the contract binds it to. Three green rows in three systems, one commitment the firm cannot keep. A compound detector reading contract facts, workforce facts and technology dates together raises the pattern while the notice period still allows a renegotiation, which is precisely the point.

From detection to decision

Detection is not the product. The decision is. A pattern that terminates in an alert email joins every other alert email, and the firm gains a slightly better-documented incident. Every confirmed compound pattern instead routes into a Decision Room, where the candidate options are priced and ranked by the earliest date each one clears every binding constraint, and a human decides while more than one real option remains. Overrides are allowed and recorded: who, when, against what evidence. The pattern’s whole job is to move the decision inside its window.

This is what enterprise intelligence means in practice, and it is the working heart of Enterprise Decision Intelligence: not a smarter dashboard for each function, but an organisation that can see the spaces between its functions, where the expensive surprises have been hiding all along.

Common questions

What are Compound Patterns?

Compound Patterns are conditions that only become visible when signals from several business functions are read together: facts that are individually acceptable and jointly a problem. A contract commitment, a hiring gap and a technology date can each pass their own review and still, in combination, describe a promise the firm cannot keep. Because each function monitors only its own domain, the combination belongs to nobody, which is why compound patterns go undetected until they become incidents.

Why can’t a single function see a compound pattern?

Because enterprises are organised vertically and this kind of risk is horizontal. Each function’s tools, thresholds and reviews are scoped to its own domain: a contract clause is evaluated against legal risk, not against the hiring pipeline; a delayed role is judged against the workforce plan, not against a commencement date in a contract. Nobody is negligent. The pattern simply falls outside every mandate, and cross-functional meetings sample the combinations only occasionally, through human memory.

What makes a detected pattern trustworthy?

Four properties. It must be explainable: it names exactly which facts combine and why the combination matters. Challengeable: any fact it rests on can be disputed, and the pattern must survive the dispute or be withdrawn. Traceable: every input fact carries its provenance and its evidence quality state. Falsifiable: the pattern states what evidence would prove it wrong, and it is resolved if that evidence arrives. A detected pattern also has a lifecycle, so it is confirmed, resolved or expired rather than accumulating as permanent noise.

What happens after a compound pattern is detected?

It routes into a Decision Room rather than an inbox. The point of detection is a decision, not an alert: candidate options are priced and ranked by the earliest date each clears every binding constraint, and a human makes the call while options still exist. If the human overrides the recommendation, the override is recorded with who, when and against what evidence. Detection that ends in an email joins every other email. Detection that ends in a decision changes the outcome.

Part of the pillarEnterprise Decision Intelligence, the complete philosophy in one essay

Related reading

See a decision run live

Watch evidence land, options reorder against the binding constraint, and the outcome get scored.