Online Optimisers · The Mycelium
OO · Confidential
Data
Route
Loop
The Learning Network

Every signal teaches.
Every batch converts better.

The Mycelium is the intelligence layer running under the execution layer. It does not just run campaigns. It routes, acts, and compounds on every result it produces.

DATA IN ROUTE agent match AGENT ACTS RESULT feeds back THE MYCELIUM
Signal flowing
How It Learns

Four steps. One continuous loop. No reset.

Each step feeds the next. Each batch informs the one after it. There is no manual intervention between steps.

1

Ingest

Every data signal from the AMP feeds in: intent score, location, engagement history, channel preference, beacon match. Nothing is discarded until it has been scored.

AMP records
2

Route

The network matches each signal to the highest-converting channel and message frame. It does not treat all 2,500 records the same. It knows which ones to prioritise and how to reach them.

Agent match
3

Act

An agent executes the routed instruction: generates the copy, triggers the send, logs the outcome per record. Every action is tagged with a record ID before it fires.

Agent fires
4

Compound

Every close, open, and response is fed back into the routing model. The signal of what worked sharpens the filter for the next batch. The system improves without being retrained manually.

Feeds back
What Improves Each Week

Three surfaces. All compounding. All measurable.

Copy Quality

Phrasing, subject lines, and CTA structures that convert in this market, for this audience, at this intent level. The copy layer does not run a static template. It learns which words move these records.

What shifts

Subject line construction · CTA framing · Opening hook · Length calibration

Record Targeting

Which signals are worth acting on. Which records to skip. The filter gets sharper each batch so spend concentrates on records that are statistically closer to closing.

What shifts

Beacon threshold · Intent signal weight · Channel affinity score · Skip logic

Conversion Attribution

Which channel closed which install. Which message frame preceded the close. The feedback loop tightens spend around the patterns that produce outcomes, not just engagement.

What shifts

Channel credit model · Close-to-touch sequence · Install source tagging · Spend reallocation

The Compounding Principle
"Every batch makes the next batch smarter."

This is not a metaphor. Each record outcome is a data point. Each data point adjusts the routing model. The network does not perform the same on batch 10 as it did on batch 1 because it cannot. The loop will not allow it.

Learning source
Every closed install
Feedback lag
One batch cycle
Manual override required
None
See the agent roster → ← Back to overview
Online Optimisers

The Mycelium · Confidential · Morell Portfolio