Map your first loop →
← Back to Resources Map your first loop →
← Resources · Field guide
Field guide · loop anatomy

Anatomy of a loop.

Every Pipelinestack loop — ICP sharpening, campaign-to-cash, ABM land, renewal watch — is built from the same six bricks. Five do work you could automate with point tools. The sixth turns a clever workflow into a system that compounds. Get the full field guide, brick by brick, with worked examples.

  • All six bricks — Sense, Decide, Act, Hand off, Measure, Learn
  • Why the sixth brick is the one that makes a loop compound
  • A worked example — one account traced through all six
  • The five-brick trap — and how to design around it
Get the field guide free · instant access
No spam · unsubscribe anytime · we never sell your data
✓
Access granted.

The full field guide is unlocked below — read on, or jump straight in.

Read the guide ↓
Field guide
pipelinestack
Field guide
Jun 2026
Loop anatomy
Anatomy of
a Loop
6 bricks
one of them makes it learn
6
Bricks per loop
1
Feedback path that closes it
24/7
Runs unattended
every cycle
Gets a little sharper
✓ Field guide unlocked.  Anatomy of a Loop · the six bricks Map your first loop on your stack →
Pipelinestack · Field guide · June 2026

Anatomy of a loop.

The six bricks inside every Pipelinestack loop — what each one does, a worked example through all six, and why the sixth is the one that makes the loop learn.

ByThe Pipelinestack team
Reading time9 min
ForCMOs · CROs · VP RevOps
01A loop is not a pipeline

A pipeline runs once. A loop comes back.

Most "GTM automation" is a pipeline: a lead enters one end, gets enriched, scored, routed and sequenced, and falls out the other end as a win or a loss. It runs left to right. Then it forgets everything and runs the same way for the next lead — including the parts that didn't work.

A loop is the same line of work with one structural difference: the outcome at the end is wired back into the start. What converted, what stalled, which message landed, which account was misjudged — all of it returns and changes how the next cycle behaves. The line becomes a circle. That circle is what we call a loop, and it's the unit Pipelinestack builds, runs and tunes.

Every loop in the catalog looks different on the surface, but underneath they're assembled from the same six bricks, in the same order. Learn the anatomy once and you can read any loop on the wall.

6
bricks in
every loop
1
feedback path
that closes it
24/7
the loop runs
unattended
every cycle
it gets a little
sharper
02The six bricks

Sense → Decide → Act → Hand off → Measure → Learn.

Read left to right, then follow the dashed line back. The first five carry work forward; the sixth carries what was learned backward, into the bricks that started the cycle.

Anatomy of a Pipelinestack loop closed-loop
01Senseingest
→
02Decideprioritize
→
03Actexecute
→
04Hand offescalate
→
05Measureinstrument
→
06Learntune
↺ feeds back into Sense + Decide — the loop closes here
1

Sense

Signal → loop

The loop's eyes. Watches the signal sources that matter — intent, product usage, firmographic and job-change moves, replies, funding, support tone — continuously, not on a nightly batch. No trigger, nothing to act on; a loop that can't sense is just a schedule.

Example · renewal-watch loopProduct logins for a key account drop 40% in a week and their champion quietly updates their LinkedIn title — Sense catches both the day they happen.
Receives from: live systems + the Learn brick, which tells it which signals proved predictive.
Ingest
Cadence
continuous · real-time
2

Decide

Signal → priority

The judgment layer. Turns raw signal into a ranked who, what, now — fit × intent × capacity — and routes it to the right owner instead of a round-robin. This is where most pipelines quietly leak: a static rules table can't weigh a live situation.

Example · speed-to-lead loopAn inbound demo request is scored enterprise-fit and routed to the right AE pod in 90 seconds — not the round-robin that would've parked it with an SMB rep on PTO.
Receives from: Sense; recalibrated by Learn each time a score is proven right or wrong.
Prioritize
Cadence
per signal
3

Act

Priority → motion

The hands. Does the work the decision implies — drafts the one-page account brief, opens a multi-thread sequence, books the meeting, updates the record. Personalised per account, not a generic blast, because it's acting on the context Sense and Decide assembled.

Example · ABM land loopAn agent writes a one-page brief per target account and opens a multi-threaded sequence that references each account's specific hiring signal — no generic “just checking in.”
Receives from: Decide; its playbooks are rewritten by Learn as winning patterns emerge.
Execute
Cadence
per decision
4

Hand off

Motion → human

The loop is autonomous, not unmanned. At the single highest-leverage moment — the live conversation, the negotiation, the save — it hands a human the full context and steps back. People do the irreplaceable part; the loop does everything around it.

Example · outbound loopA prospect replies “send me times.” The loop books the meeting and drops the brief plus the full thread into the AE's Slack 20 minutes before the call.
Receives from: Act; passes the human's outcome straight to Measure.
Escalate
Cadence
at threshold
5

Measure

Outcome → truth

The scoreboard — and a strict one. Captures what actually happened against a held-out control, so the result is causal, not just correlated. This is the number you can defend to a board, and the raw material the next brick learns from. No clean measurement, no honest learning.

Example · campaign-to-cash loopThe loop holds out 5% of the audience; the treated cohort books 2.3× more meetings, proving the lift is causal — not seasonality or a lucky quarter.
Receives from: Hand off + Act; emits a verified outcome to Learn.
Instrument
Cadence
per outcome · vs holdout
6

Learn

Truth → back to the start

The brick that makes it a loop. It writes every verified outcome back into Sense and Decide — reweighting which signals matter, resharpening the scoring, rewriting the playbooks Act runs. Remove it and bricks 1–5 still work; they just repeat the same mistakes forever. Keep it, and the loop is measurably better next cycle than it was this one.

Example · ICP-sharpening loopThe loop learns that “pricing re-view + RevOps hire” predicts wins, weights that signal up, and scores the next account that fires the same pattern higher — on its own, no ticket filed.
Feeds: Sense + Decide + Act — closing the circle.
Tune
Cadence
every cycle · compounding
03The brick that learns

Five bricks ship pipeline. The sixth ships compounding.

Bricks one through five are visible and easy to admire. They're the demo: a signal fires, an agent scores it, a sequence goes out, a meeting lands, a dashboard lights up. Plenty of tools can stitch that together. It looks like a loop, and for one cycle it behaves like one.

The difference shows up on cycle ten, and cycle fifty. The Learn brick takes each verified outcome and edits the loop's own behaviour: the signal that kept predicting wins gets weighted up; the score that kept overrating a segment gets pulled down; the opening line that kept getting replies becomes the default. Nobody files a ticket. The loop you turn on in week nine is not the loop you have in month six — it's sharper, because it has been quietly editing itself the whole time.

The whole point

Without the sixth brick you have a fast workflow that plateaus. With it, you have an asset that's worth more every quarter you run it — and that's almost impossible to bolt on afterward.

That last part matters. Learning can't be added as a final step; it has to be designed into how the other five bricks record what they do. Measure has to capture causal truth, Act has to log why it chose a play, Decide has to expose its scoring. A loop built without the sixth brick in mind usually has to be rebuilt to get it — which is exactly why we build it in from brick one.

04One loop, in motion

The same account, through all six.

Here's a single account moving through an ICP-sharpening loop — anonymised, but the shape is real. Watch the sixth row: it's the only one that changes what happens to the next account.

Brick
What happens
When
01Sense
Three buying-committee execs view pricing twice in 48h, and the account posts two roles for the buyer persona.
Day 0
02Decide
Scored 91/100 on fit × intent × open capacity; routed to the enterprise pod that owns the territory — not round-robin.
+2 min
03Act
Research agent writes a one-page brief; sequence agent opens a multi-threaded touch referencing the specific hiring signal.
+9 min
04Hand off
A positive reply trips the threshold; meeting is booked and the AE gets the brief + full thread in Slack before the call.
Day 2
05Measure
Meeting becomes an opportunity; lift is measured against a held-out control cohort so the win is provably causal.
Day 18
06Learn
"Pricing re-view + persona hiring spike" is confirmed high-converting — Sense weights it up and Decide raises its score for every future account.
Next cycle

By the next account that fires the same pattern, the loop reacts faster and scores it higher — because account number one taught it to. Run that a thousand times and the loop's judgment is no longer a guess your team configured once; it's an edge the system earned.

05Five bricks is a workflow

What you get with five — and never more.

Plenty of teams build the first five bricks, call it done, and wonder why the gains flatten after the first quarter. Three things happen to every loop missing its sixth brick.

01

It runs once, forever

Every cycle starts from the same assumptions as the first. A win and a loss leave the loop in exactly the same state — fast, tireless, and never any wiser.

02

It has no memory

A pattern that printed pipeline last month is invisible next month. The team relearns it by hand in a QBR, ships a manual tweak, and the clock resets.

03

It drifts in the dark

The market moves; the loop doesn't. Conversion decays a point at a time and no one can say why, because nothing is comparing this cycle to the last.

None of these are tool problems — the five bricks are doing their jobs. They're architecture problems. The fix isn't a better sequencer or a smarter score; it's the feedback path that makes the other five accountable to their own results. That's the sixth brick, and it's the reason a Pipelinestack loop is an appreciating asset instead of a faster version of what you already had.

From anatomy to action

Want to see which loop to build first on your stack?

In a working session we map your highest-leakage motion to a six-brick loop, show you exactly where the feedback path closes, and size the lift before you commit to anything. You leave with the diagram either way.

Map your first loop → Back to Resources