CAVEMANAIQ all four storylines
storyline 03 · the audience

Why pay to reach strangers we already know?

GA
ME
TM
G4
SC
MC
DM
DB
01
01 · one console

Seven dashboards to one console

the console · all seven servers, one login
ask anything…
7 servers docked · each holding its own login
The buyers live in the shop’s own database.
02 · you ask · on the record

The question is the whole interface

“ We are paying to reach strangers when we already know who our buyers are, and our customer list covers a year because that is what the platforms default to. Which customers should we be seeding from, which should we be excluding · and are the lists we build actually doing anything?
you
the console
on the record
7 servers docked · each holding its own login
handed to the agent, word for word · nothing runs unasked
▸ we write the request down before we send it
▸ and the answer when it comes back
The shop was paying to reach people who resembled its traffic, while its actual buyers sat unused in its own database.

What the shop had to settle before it could build a single list

Two questions came before any list existed, and neither is a marketing preference.

what had to be answered first

the four kindsbuyers, cart abandoners, site visitors and lookalikes answer different questions · one of them seeds prospecting, another can only ever narrow it
the window is the producthow long these customers actually take to come back, measured from real repeat orders · the median lands either side of a year depending on the category, and the tail runs out to around three
lawful basis firsta customer list is only ever built from people whose data may lawfully be used that way · checked before any row is prepared, as standing practice
prepared, not sentidentities are hashed before they leave the shop · the platform receives a fingerprint it can match, never an address it could read
beneath the argument · the rest of the record
seven states, not onecreated, correctly sourced, receiving matched events, countable, targetable, attached and delivering · seven different things, and confusing any two of them is how this work goes wrong
03 · dispatch

The agent decides who to ask

THE MCP FLEET
THE PLATFORMS · reached by API, not by login
the agent
reasoning: what do I need, in what order?
DB the database 01
who buys, and how long they take to return
ME Meta 02
did the lists land, and what matched
DM Data Manager 03
the same lists into search · approved destinations only
GA Google Ads 04
which lists the live campaigns use
04
GA
Ads
spend · conversions
Google Ads API
your ad account · Google’s cloud
02
ME
Meta
spend · conversions
Meta Marketing API
your ad account · Meta’s cloud
TM
GTM
stands down
GTM API
the tag container running on your site
G4
GA4
stands down
GA4 Data + Admin APIs
your site’s events · browser + server
SC
GSC
stands down
Search Console API
how Google search sees your site
MC
MC
stands down
Merchant Center API
your product feed · Google’s cloud
03
DM
DM
audiences · uploads
audience · conversion upload
your audience lists · Google’s cloud
YOUR OWN RECORDS · READ THROUGH THE DATABASE MCP, NOT A VENDOR CLOUD
01
DB
the database
your own records
the database · your own records
who buys · how long to return
The order history is asked first, and asked properly, because it is the only place that knows how long a customer stays a customer.
04 · the crossing
data plane · no human · always running

An approved list stops being a record and becomes a control

THE DATA PLANE · runs whether or not anyone is watching
a shopper · not you
the site · code running
every visit writes the record · no one at the wheel
Meta’s servers · ads runtime
buyer lists arrive as custom audiences · match rate read back
Google’s servers · audience matching
the same lists into the search side’s matching · approved destinations only
what changed in the plane
The shop’s own order record is now a targeting control inside two advertising platforms.
owned evidence becomes something the platforms serve against.
This is the beat where the work leaves the control plane.
05 · gather

The evidence converges on the session

ME
Meta
lists landed · match rate · attachments
DM
DM
lists into search-side matching
GA
Ads
which lists live campaigns use
DB
the database
who buys · how long to return
one payload · assembling
Meta
lists landed · match rate · attachments
DM
lists into search-side matching
Ads
which lists live campaigns use
the database
who buys · how long to return
all 4 answers in · ready to combine
The lists land, and the two platforms disagree about the same people · matched far better on one than the other, and better again once more than an email address travelled with them.
06 · the readout

Four kinds of list, four different jobs

what had to be answered first
Two questions came before any list existed, and neither is a marketing preference.
the four kindsbuyers, cart abandoners, site visitors and lookalikes answer different questions · one of them seeds prospecting, another can only ever narrow it
the window is the producthow long these customers actually take to come back, measured from real repeat orders · the median lands either side of a year depending on the category, and the tail runs out to around three
lawful basis firsta customer list is only ever built from people whose data may lawfully be used that way · checked before any row is prepared, as standing practice
prepared, not sentidentities are hashed before they leave the shop · the platform receives a fingerprint it can match, never an address it could read
the audience finding
built from ME DM GA DB
the seedthe buyer lists seeded the lookalikes · prospecting reach now comes from customers, not from site visitors
the exclusionrecent buyers kept out of prospecting · budget stops paying to reach someone who has just bought
the windowa one-year list cannot hold a multi-year replacement cycle · the customers actually due to buy again had already dropped out of it
The shop’s own buyers now decide who two advertising platforms go looking for and who they leave alone · and the only thing that cannot be seen from outside is how many people are inside the audiences built from site behaviour.
A customer list does three different jobs and most accounts only ever use it for one.
07 · the loop closes

You decide

you, again
“Yes · upload the approved lists.”
applied · and read back
The same customers seed prospecting and are kept out of it at once. A later read found which lists are actually attached, and to which live campaigns.
The buyer lists go up as custom audiences on the social side and into the search side’s own matching, to destinations approved in advance.
08 · how it ended

How it ended

how it ended

something was built
No incremental revenue is claimed and none was proven.
A match rate is not reach, an uploaded list is not spend, and an attachment is not delivery.
Several attachments sit on ad sets that are paused · that is use rather than current delivery, and historical attachment is never presented as either.
The membership figure for the behavioural audiences is unreadable through the platform surface, and no single defect in the data plane is offered as the explanation of it.