The Proprietary Intelligence – 2
Layer 2: BrandTwins as Representation
A BrandTwin is an N=1 counterpart for each Best customer. Not a profile. Not a segment label. Not a collection of attributes in a database row. A BrandTwin is a continuously updated model of a single customer’s relationship with the brand.
The BrandTwin knows what the customer values — not demographically but behaviourally, based on what they choose, what they ignore, what they return, what they recommend. It knows what the customer is drifting away from — the early signals of disengagement that precede the visible symptoms. It knows what pulls the customer back — which messages resonate, which offers convert, which experiences rebuild the habit of engagement. It knows what the brand has promised — explicitly through communications and implicitly through prior treatment — and what the customer therefore expects next.
Crucially, each BrandTwin carries a micro P&L. It tracks revenue generated by this specific customer. It tracks cost invested — in incentives, in service, in interventions, in attention. It calculates contribution margin impact. It estimates retention probability. It identifies the next-best investment opportunity — the action most likely to generate incremental value given the current state and trajectory.
This is where Meridian begins to feel less like marketing and more like capital allocation. The BrandTwin’s job is not to maximise clicks or opens or even conversions in isolation. Its job is to maximise relationship value over time, within constraints. It must balance short-term conversion pressure against long-term trust. It must recognise when an aggressive offer would generate a transaction but erode the relationship. It must know when not to act — when silence is more valuable than another message.
The BrandTwin becomes, in effect, the customer’s advocate inside the system. It represents the customer’s interests because representing those interests is how value is maximised. A customer who feels understood stays longer, buys more, forgives occasional failures, and recommends to others. A customer who feels exploited leaves at the first opportunity.
This is not altruism. It is aligned economics. The BrandTwin protects the customer because protecting the customer protects the revenue stream.
Layer 3: Agents as Execution
Meridian’s intelligence is meaningless without execution. Context Graphs can hold perfect understanding. BrandTwins can model customers with exquisite precision. None of it matters if actions do not follow.
This is where agents operate — not as chatbots or assistants but as autonomous operators running continuous micro-experiments across the customer base:
- Adjusting content, timing, and cadence for each customer based on predicted responsiveness.
- Choosing channels based on where this specific customer is most likely to engage, not where the average customer engages.
- Selecting offers based on margin and elasticity — ensuring that discounts go to customers who need them to convert, not to customers who would have converted anyway.
- Triggering service interventions when trust risk rises — a proactive call, a surprise upgrade, a personal note — before the customer reaches the threshold of departure.
- Coordinating on-site discovery experiences with off-site messaging — so that the email a customer receives connects to the products they see when they visit.
- Detecting drift earlier than human reporting cycles can surface it — recognising the pattern of disengagement while intervention is still possible.
Instead of campaign bursts, Meridian runs a living system. Hundreds of decisions per day, per customer base, guided by a consistent model of what matters. This is not “set and forget” automation. It is genuine agency — the capacity to perceive, decide, and act based on goals rather than rules.
The difference matters. Rule-based automation executes the same action whenever a trigger fires, regardless of context. Agent-based intelligence evaluates context, considers alternatives, estimates outcomes, and selects the action most likely to achieve the goal. The first is mechanical. The second is adaptive.
Layer 4: TwinFactory as Scale
A single BrandTwin is interesting. A million BrandTwins are infrastructure.
TwinFactory is the machinery that creates, maintains, and updates BrandTwins at scale. It ingests signals continuously from every touchpoint. It updates Context Graphs and confidence scores as new information arrives. It replays decision traces to transfer learning from one customer’s experience to similar customers’ predictions. It keeps the system consistent as products change, inventory shifts, pricing evolves, and customer behaviour adapts.
This is what transforms an AI concept into an operational system. The challenge of N=1 personalisation has always been scale. How do you deliver individual intelligence to millions of customers without employing millions of marketers? TwinFactory is the answer: industrialised twin creation that is systematic, repeatable, and autonomous.
The navigational metaphor
A meridian is a reference line used for precise positioning. Sailors used meridians to know exactly where they were on the globe — not approximately, not “somewhere in this region,” but precisely. The meridian did not move. It provided a fixed reference against which position could be calculated.
Meridian positions each customer precisely in their journey. It knows not just that a customer is drifting but why they are drifting, where they are drifting toward, and what will arrest the drift. It uses signals others cannot see — the subtle patterns in behaviour that precede visible symptoms. It makes decisions others cannot justify — interventions that seem counterintuitive until you understand the underlying model. It delivers outcomes others cannot guarantee — because it is built for guarantee, not hope.
And like any powerful proprietary model, Meridian’s promise is not that you will understand it. Meridian’s promise is that you will measure what it delivers.
