$47M of validated decision patterns from 1.5 years of marketing archive

For CIOs & architecture leaders. Companion piece for marketing leaders: The $75M playbook hiding in 1.5 years of campaign data →

Marketing organizations already produce the intelligence they need. Every campaign decision, every outcome measured, every brand manager's judgment about what worked. The information exists in the org. The architecture to retain it doesn't.

Enterprise marketing software was built as a system of record. It captures what happened, aggregates it, reports on it. It wasn't built to capture the decisions teams made, the conditions they made them under, or how to apply that the next time someone faces the same setup. So when the same conditions repeat, the team starts from a clean planning sheet.

This is the insight-to-action gap as a data-architecture problem. The intelligence the organization has already produced is in the building. It just isn't queryable.

The architecture problem

The client runs a multi-hundred-brand portfolio and thousands of campaigns a year. Years of campaign decisions sit logged across CRMs, performance dashboards, agency reports, and brand manager handovers. None of it is queryable as a decision history. There is no shared schema across sources. There is no link between the conditions a decision was made under and the outcome that followed. There is no mechanism to surface what the organization has already learned the next time someone faces the same setup.

A brand manager pivots an audience mid-campaign on Brand A. Sales climb 58% year-over-year, a top-5 outcome in the brand's history. The decision lives in a readout. Six months later, that brand manager rotates. The next campaign on Brand A starts from a clean planning sheet. A year after that, Brand B faces the same conditions and the team makes a different call. No one in the room remembers the +58% pattern.

What we built

A.Team built a learning system on Assemble, our AI platform for unifying enterprise marketing data and making it actionable. For this client, the system is the Continuous Learning capability of Marketing & Media Performance. It runs in six steps.

Ingest and normalize. Connectors to sales data, campaign performance, audience definitions, and market context, normalized into a single structured layer against the brand's own business ontology. The ontology is the part that makes the rest of the system possible. Without it, AI agents read the data as undifferentiated text. With it, they read it the way a brand manager would.

Decision log. Past and active campaigns are structured as records of context, action, and outcome, linked to the entities they touch (brands, audiences, products, markets). Every campaign becomes a queryable row: what was the brand's momentum, what action did the team take, what happened, attributed to which entities.

A slice of Brand A's decision log: every move, its measured outcome, and the pattern the system anchored to each.

Pattern codification. When the same context-action-outcome sequence repeats with consistent direction and magnitude, it gets promoted to a play with a confidence score, an expected value, and explicit trigger conditions. The promotion is automatic above defined thresholds; a human curator reviews edge cases before plays go into circulation.

In-workflow surfacing. Plays surface inside the brand manager's existing tools. No new dashboard, no separate workflow. "Do this" plays carry confidence and expected value. "Avoid this" plays carry the cost of getting it wrong. The surfacing happens at the moment the decision is being made, not after.

Cross-brand transfer. Patterns that mature on one brand get reviewed for portability. AI plus a human curator decides whether trigger conditions are brand-specific or brand-agnostic. If brand-agnostic, scope expands across the portfolio. A pattern's reach grows as its evidence base grows.

Value over time. As more decisions enter the log, learnings compound and propagate. New plays codify. Existing plays sharpen. The system gets more accurate the longer it runs and more useful as it spans more of the portfolio.

What 1.5 years of unified data revealed

Six patterns surfaced with enough evidence to codify. Three "do this" patterns: $47M in cumulative lift across the dataset. Three "avoid this" patterns: $28M in damage averted. A $75M signal sitting in the archive, invisible until the architecture made it queryable.

The same record, distilled into a playbook every downstream agent can read: three moves to lean into, three to pull back from, each weighted by dollar impact and confidence.

The strongest "do this" pattern: when the brand was in a healthy YoY stretch with campaigns running on custom audiences, shifting targeting on one running campaign produced the largest lifts. Same creative, same channels, sharper audience. Seven independent episodes in the data, 92% confidence, $11.7M in cumulative lift attributed to the play.

The mirror pattern was the same action in the wrong state. Audience pivots applied during declining momentum, without a creative refresh, destroyed value every time they ran. -62% YoY in one stretch. -60% two months later. -57% two months after that. Same lever, different system state, opposite outcome.

The remaining four covered channel mix in healthy stretches (+$25.5M), audience structure when momentum existed (+$10M), campaign volume during declines (-$9.2M), and pivot cadence without creative refresh (-$7.3M).

Results

Validated pattern value across the 1.5-year dataset: $47M in cumulative lift from three "do this" patterns and $28M in damage averted from three "avoid this" patterns. The strongest single pattern accounts for $11.7M at 92% confidence across seven episodes.

Current system performance: 67% predictive accuracy on the originating brand. Two out of three times, the system correctly anticipates whether a planned campaign decision will hit or miss before it runs.

What that playbook is worth over three years: a projected 2.7× lift over baseline as learnings accumulate, mature, and fully integrate.

Projected value trajectory by maturity state, three-year cumulative incremental sales:

→ Current baseline (one brand, current accuracy): $5.8M

→ Matures during first 12 to 18 months (more patterns codified, accuracy improves): $10.3M

→ Fully mature (validated across portfolio, embedded in planning workflows): $15.6M, a 2.7x lift over baseline

The trajectory is a function of three operational variables: decision volume entering the log, pattern maturation rate, and cross-brand transfer success rate. The first cross-brand transfer landed at $9M projected lift on Brand B, drawn from Brand A's track record on the same pattern.

Operational changes at scale:

Decision latency dropped. A 92% confidence pattern with a known historical value doesn't require a six-week analysis cycle to justify. It requires a check that the trigger conditions hold.

Error visibility improved. Avoid patterns become high-confidence blocks in the planning workflow, coded with cost of error. The play surfaces at decision time, not at post-mortem.

Most enterprise software depreciates without added investment. Learning systems accumulate organically.

The diagnostic question for marketing AI architecture

When marketing makes a decision today, where in your system does it get captured against the outcomes it produces? And when the same conditions repeat next quarter on a different brand, what would have to be true of the architecture for the team to know what's already been learned?

If the answer to the first question is "in a readout that nobody re-reads" and the answer to the second is "we'd have to rebuild the analysis from scratch," the constraint is in the data architecture. The team's intelligence has already produced the patterns; the architecture is what fails to retain them. A system of record can't become a system of action without the connective tissue between them. That connective tissue is the decision log.

See how Marketing & Media Performance works → or Request your Strategic Intelligence Gap Assessment

This is the fifth essay in A.Team's series on how enterprise marketing and planning organizations are closing the insight-to-action gap. It describes a client engagement delivered through A.Team's AI Solutions practice. The client name and brand identifiers are withheld under NDA. Dollar figures reflect the engagement as documented in the client's own performance systems and Assemble's modeled forward case. Realized cumulative pattern value (+$47M / -$28M) reflects historical performance attributed to the codified plays across the 1.5-year dataset. Three-year cumulative projections ($5.8M / $10.3M / $15.6M) reflect Assemble's compounding-intelligence model under three maturity scenarios.