NuLoop
Sign in

AI marketing learning engine · Retail-first

Every campaign starts
sharper than the last.

NuLoop is the marketing learning engine for D2C and retail brands running paid acquisition, brand, and lifecycle work on Google, Meta, ecommerce, and CRM. Every campaign, customer signal, and review becomes memory the next campaign starts from.

1% better every day · Built by Nurix

Start a 90-day pilot
Loop status · running
  • 01Catalog readiness checked
  • 02Guarantee language retired
  • 03Repeat-buyer creative split out

The cost of not compounding

Every week you ship the same learnings you shipped last week.

Six pains we hear from retail teams running real spend on Google, Meta, and ecommerce. NuLoop exists to close each one.

  • 01

    Campaigns restart from scratch.

    Prior learnings live in people, decks, agency Slack, and dashboards.

  • 02

    Stale offers reach launch.

    Out-of-stock SKUs, expired coupons, margin-killing discounts go live.

  • 03

    Customer objections die in tickets.

    Returns, reviews, support, and call signals rarely change next week's creative.

  • 04

    Reviewers repeat themselves every week.

    Brand, legal, merch, and performance comments don't compound into rules.

  • 05

    Dashboards report. Nothing learns.

    You see what happened. You never see the decision chain behind it.

  • 06

    Brand and performance work in parallel.

    Two brains, two tones, two stacks. Nothing tells one what the other proved.

1% better every day · in one loop

One festive sale, narrated.
What the loop saw, decided, learned.

A real shape of a working loop: catalog reality before creative, brand memory before copy, operator approval before spend, customer signal after launch, and a learning the next loop starts from.

Promotion loop · Bundle A · sample

Day 0 → Day 8

Completed

  1. Day 0 · 09:14

    Goal entered

    Promotion Loop · festive sale on Bundle A · 5-day window

  2. Day 0 · 09:21

    Product Agent flags 2 SKUs at low stock

    Bundle downgraded to 3-SKU variant before creative work begins

  3. Day 1 · 11:08

    Creative Agent drafts 12 variants

    Brand Agent retires guarantee language — rejected twice prior loop

  4. Day 2 · 15:42

    Suggest mode · operator approves

    Meta launch · 4 ad sets · margin floor enforced · cap held

  5. Day 5 · 18:30

    CRM cohort signal returns

    Repeat buyers click warranty over discount · stronger CTR

  6. Day 7 · 09:00

    Weekly Learning writes memory

    Repeat-buyer creative split from prospect creative · medium confidence

  7. Day 8 · 09:00

    Next loop opens with learning pre-applied

    Prospect and repeat segments diverge in plan from the first stage

Sample loop · numbers illustrative

Learning memory

Repeat buyers respond to warranty over discount.

Evidence. CRM cohort · stronger CTR vs prospect creative · 184 events across 5 days · stronger post-purchase open rate.

Confidence
Medium · re-check at 90 days
Owner
Performance + Brand · co-approved
Reversal rule
Two consecutive lower-CTR loops
Next applied
Bundle B festive · planned for Day 9

Every memory entry in NuLoop carries evidence, confidence, owner, expiry, and a reversal rule.

What actually compounds

Every loop starts where the last one ended.

  1. Loop 01retired hook
  2. Loop 02stale offer blocked before spend
  3. Loop 03brand rule promoted
  4. Loop 04winning proof point reused
  5. Loop 05segment split (repeat vs prospect)
  6. Loop 06objection → landing copy
  7. Loop 07creative variant retired

Compounding is retained learning per loop.
Not a daily multiplier on revenue.

NuLoop IQ · shared context layer

One layer everything reads from.
One layer everything writes to.

Brand voice, product truth, customer signal, approvals, channel rules, and outcome memory live in one versioned layer. Every agent reads from it. Every loop writes back to it.

Every brief, generated output, recommendation, review, and learning records the NuLoop IQ version it used.

Context snapshot

NuLoop IQ v24 · Summer retail push

82% ready

Brand voice

Ready

Tone, vocabulary, banned language

Visual style

Ready

Logos, color, photography rules

Product catalog

Live

SKUs, PDP, margin, status

Pricing & offers

Review

Promo, coupon, expiry, margin floor

Inventory

Live

Stock, low-stock alerts, returns window

Audience

Ready

Segments, personas, lifecycle stage

Proof library

Ready

Claims, reviews, ratings, lab tests

Channel rules

Ready

Google, Meta, landing, email, SMS

Review policy

Ready

Approvers, SLA, risk routing

Customer signals

Live

CRM/CDP, support, conversations

Budget & autonomy

Ready

Caps, margin, suggest-mode rules

Outcome memory

Live

Orders, returns, repeat, retention

Recommendation

Refresh the bundle campaign before scaling spend

Needs review

The learning engine sees enough signal to recommend a creative and offer refresh, but not enough proof to launch automatically.

Confidence
74%
Expected lift
+8–12%
Autonomy
Suggest
  • Offer is valid for 11 more days
  • Review team rejected guarantee language twice
  • Repeat buyers respond better to bundle proof
  • Meta creative fatigue crossed the threshold

Agent library

Specialist agents.
Jobs, authority, review gates.

Twelve visible specialists, each with a clear job and a published write authority. The grid is your operating org chart, not a list of features.

NuLoop earns authority. It does not assume it.

  • Conductor

    Suggest

    Routes goals into pipelines and enforces gates.

  • Strategy

    Draft

    Goal → plan, level, assumptions, risks.

  • Brand

    Suggest

    Applies voice, visual style, channel nuance.

  • Product

    Suggest

    Catalog, SKU, inventory, price, PDP readiness.

  • Audience

    Draft

    Segment hypotheses from CRM and outcomes.

  • Creative

    Draft

    Channel-ready copy, concepts, asset briefs.

  • Compliance

    Review · block

    Claims, policy, substantiation, risk.

  • Activation

    Bounded

    Google / Meta writes under guardrails.

  • Lifecycle

    Draft

    CRM and owned-channel drafts.

  • Optimization

    Bounded

    Budget, creative, and audience moves.

  • Reporting

    Suggest

    Performance and decision explanations.

  • Learning

    Review · block

    Reusable patterns, memory proposals.

Autonomy ladder · 0 → 4

Earned per workspace, per pipeline, per action type.

  1. Level 0

    Draft-only

    Plans, copy, concepts, reports. Basic context.

  2. Level 1

    Suggest

    Platform changes prepared for human execution.

  3. Level 2

    Human-approved

    Launch after explicit approval.

  4. Level 3

    Bounded auto

    Low-risk moves inside budget and margin limits.

  5. Level 4

    Learning auto-apply

    Approved memory patterns apply to future runs.

Customer signal loop

Customers are already telling you what to run next.

Reviews, returns, support, and conversations carry the next campaign's hook, proof, segment, and offer. NuLoop reads them so the next loop responds to them.

  • Customer signal

    Support ticket cluster

    12 tickets this week mention delivery confusion. Customers can't tell shipping promise from dispatch time.

    Action · Landing copy — hero line

    Fast shipping

    Arrives in 3 days. Tracked from dispatch.

  • Customer signal

    Returns spike · size S

    Size S returns up 38% on Bundle B. Review reasons cluster on fit, not fabric or color.

    Action · Ad set + landing proof line

    Soft, breathable, made for everyday

    True-to-size. Fits as ordered, 92% of reviews.

  • Customer signal

    Review · 4.7★

    Top-helpful review on Bundle C used the phrase “weightless and serious” — repeated by 3 other buyers.

    Action · Next creative variant

    Premium build. Designed to last.

    Weightless and serious — in the reviewer's own words.

Signal source · customer touchpoints — never a call-center product.

Repeatable loops

Eight repeatable loops. Reusable across campaigns.

Every loop is a template, not bespoke automation. Inputs, gates, outputs, activation surface, and memory writes are versioned and shared across the workspace.

  • 01

    Product Launch

    Take a new SKU, bundle, or collection from goal to launch-ready plan plus owned-channel drafts.

    Paid · or suggest mode

  • 02

    Promotion

    Plan and activate a discount, sale, festive campaign, or bundle offer under margin and approval gates.

    Paid · or suggest mode

  • 03

    Paid Acquisition

    Plan, create, check, launch or suggest, optimize, and report for Google and Meta.

    Paid · or suggest mode

  • 04

    Brand Campaign

    Brand-safe narrative, messaging, creative concepts, and owned-channel drafts in review.

    Owned-channel drafts

  • 05

    Creative Refresh

    Detect fatigue, generate variants, risk-route, approve, and activate or queue for review.

    Paid · or suggest mode

  • 06

    Lifecycle

    Winback, repeat purchase, replenishment, cross-sell, abandoned cart, and post-purchase drafts.

    Owned-channel drafts

  • 07

    Weekly Learning

    Summarize what changed, what worked, what failed, and what to try next.

    Report + memory proposals

  • 08

    Data Readiness

    Check sources, freshness, connectors, permissions, and autonomy eligibility.

    Report + memory proposals

What it is. What it isn't.

NuLoop owns the learning loop above your stack.

NuLoop is

  • The marketing learning engine above your stack.
  • Shared context, agent execution, review, activation, outcome memory.
  • Memory that the next campaign starts from.
  • Trust evidence on every move.

NuLoop is not

  • A CRM
  • A CDP
  • A creative model
  • A DAM
  • A marketing cloud
  • An ad optimizer
  • A call-center product

NuLoop integrates with these systems. It owns the learning loop above them.

Where the signals come from

  • Shopify
  • Google Ads
  • Meta Ads
  • Klaviyo
  • Hightouch
  • Salesforce
  • GA4
  • Customer touchpoints

Trust center

Every move is reviewable.

Trust is a surface, not a footnote. Audit, provenance, model choice, cost, and connector health are visible on every run.

  • Audit

    Every agent decision logged.

    Run-level audit, human overrides, rollback events. Nothing happens off the record.

  • Provenance

    Every generated output traced.

    Model, prompt version, NuLoop IQ version, source assets, license, and likeness risk.

  • Models · cost · connectors

    You see what ran, where, and what it cost.

    OpenAI and Gemini · per-run cost telemetry · connector health agent surfaces freshness and degradation.

90-day pilot

Bring one campaign.
We'll show you the loop that keeps improving.

Bring brand rules, product/offer truth, customer signals, and one campaign goal. NuLoop will show shared context, repeatable pipeline, agent decisions, review gates, suggest-mode packet, and learning summary.

Start a 90-day pilot

Built by Nurix.