The Loop

Why it exists: clienteling is the foundation

Everything on this page serves one relationship: an associate and a customer. The platform gives every associate a Customer 360 that knows the whole story, a unified inbox where the conversation lives — email, SMS and WhatsApp in one thread — capture at the counter that starts a relationship in seconds, and Next Best Actions that say exactly who to reach, why, and when.

The loop is what makes that relationship smarter every time it turns: activity becomes insight, insight becomes an action in an associate's hands, the outcome is measured, and the next suggestion is better than the last.

Most retail platforms have some of these pieces. What makes this one different is that every stage is one connected system, out of the box — the output of each stage is the verified input of the next, every decision can explain itself, and the loop measures its own effect instead of asserting it. And it runs in real time: no overnight batches — actionable insights refresh in seconds — so the platform influences people in the moment, while the customer is still in the store, still in the conversation, still in the basket. Nothing here is a roadmap slide: every capability on this page runs today, end to end, on a demonstration tenant you can see live.

flowchart LR
  A(["Store & customer
activity
sales · captures · messages"]):::world
  B(["Sense
every signal,
one fabric"]):::cap
  C(["Understand
Insight Warehouse —
always fresh"]):::cap
  D(["Think
scores, segments,
governed AI"]):::cap
  E(["Act
NBAs & journeys,
guarded"]):::mid
  F(["Learn
uplift vs
held-back control"]):::brass
  A --> B --> C --> D --> E --> F
  F -. "every turn gets smarter" .-> A
  linkStyle 5 stroke:#A87E3F,stroke-width:2.5px
  classDef world fill:#2B3036,stroke:#4E5651,color:#F2EFE6,stroke-width:1px
  classDef cap fill:#1E4D40,stroke:#163A30,color:#F2EFE6,stroke-width:1px
  classDef mid fill:#2E6B5A,stroke:#1E4D40,color:#F2EFE6,stroke-width:1px
  classDef soft fill:#E7EEE9,stroke:#B9CCC2,color:#1E3A30,stroke-width:1px
  classDef brass fill:#A87E3F,stroke:#8A6630,color:#FFF8EA,stroke-width:1px
  classDef guard fill:#3A3F45,stroke:#A87E3F,color:#F2EFE6,stroke-width:2px
      

The six stages, in detail

Nothing that happens is lost

Every sale, capture, message, appointment, service case, product import and segment change publishes a named business fact onto the platform's internal messaging fabric — not raw database noise, but events with meaning: a sale completed, a customer was captured, someone entered a segment.

• Self-healing delivery. Events are stored durably and retried until handled. A delivery interrupted by maintenance is found and re-sent automatically — and can never be delivered twice.
• One event, many reactions. The same sale simultaneously updates the warehouse, rescores the customer, credits the selling associate, checks whether a journey's goal was met, and notifies any connected external system — without any of those parts knowing about each other.
• Privacy in the envelope. Events carry references, never personal data — so the fabric itself can never leak a name, email or phone number.

The Insight Warehouse

Not a BI export bolted on the side — a genuine analytical layer inside the platform. Each dataset is a configuration document: what it derives from, how it aggregates, which events keep it fresh, and the quality checks it must pass. Adding a new dataset is authoring, not an engineering project.

• Business-time correctness. A migrated customer's five-year history counts as five years of loyalty — never as "imported yesterday".
• Always fresh. Datasets react to the events that affect them and rebuild exactly what changed, in the right dependency order.
• Quality gates. A dataset that fails its checks is quarantined loudly — never silently served to a decision.
• Privacy is structural. Right-to-be-forgotten requests sweep the warehouse automatically, because every customer-level dataset declares how it links to the customer.
• The marts are where it gets exciting. Raw business facts flow through the pipeline into curated, decision-ready marts: a customer insight mart (the single richest view of every customer), engagement marts that learn when outreach converts, per-customer send-window marts, and revenue and product facts that power the plain-English analytics. Each stage builds on the last — facts feed marts, marts feed decisions.

The platform's core currency

On top of the warehouse, the platform computes customer insight per customer, from their real history: recency-frequency-monetary standing, predicted lifetime value, churn risk, engagement, lifecycle stage, category and brand affinity, price point. It refreshes on every relevant event as well as on schedule, and it surfaces exactly where the relationship happens — the Customer 360 an associate opens before saying hello.

• Living segments. Authored from criteria, scripted for anything bespoke, or suggested by AI from your own data. Membership reacts in seconds — a purchase can move someone out of "at risk" before the receipt is printed — and every entry and exit is itself a signal: it can raise an action, enrol a journey, notify another system, and hold back its own control group so the segment's effect is measurable from day one.
• Product intelligence too. A recommendations engine builds bought-together, complementary and popularity models from the retailer's own transactions — so suggestions reflect your customers, not a generic catalogue.
• The glass box — and why it matters. Ask "why is this customer in this segment?" and the engine produces a deterministic, criterion-by-criterion answer. A dedicated activity screen shows every decision the brain made: segment moves, actions raised and why, journey enrolments, timing learned, AI usage and spend. The why: a retailer will only act on intelligence it can interrogate. Black-box scores get ignored on the shop floor and challenged in the boardroom; a decision that explains itself gets used — and can be defended to a customer, a brand director, or a regulator.

Augmented intelligence: AI where it earns its keep, governed everywhere

The design philosophy is augmentation. AI drafts, suggests, summarises and explains; associates and the deterministic engine decide. It makes the associate faster and the platform more articulate — it never replaces the relationship, and it never gets the last word on a customer.

AI sits at the heart of the platform — and the platform is AI-ready out of the box: connect one key and seven AI capabilities light up everywhere language helps; remove it and everything still runs on its deterministic baseline. All of it flows through one governed gateway: the AI can only choose from what the platform hands it, spend is metered to the cent against a monthly budget you set, and every surface degrades gracefully when there is no key, no budget, or no signal.

• Ask your business anything, in plain English. "Roughly how much money did we take through the tills this month?" — answered live from the warehouse, from a governed catalogue of measures the AI cannot go beyond. Extending the vocabulary is one configuration document.
• Entity questions. "Who is our best customer and what should we do next with them?" answers with a person — their name linked to their 360, their value, and their open next-best-actions.
• Everywhere language helps: reply suggestions in the brand's voice; conversation summaries so an associate picks up any thread in seconds; sentiment read on every inbound message, so a frustrated VIP surfaces for care before the relationship sours and the next suggestion accounts for their mood; next-best-action wording; basket completion ideas; segment suggestions from your data; even drafting an automation from a sentence — each one individually switchable, metered, and optional.

Ask it — real questions, live answers

• "Roughly how much money did we take through the tills this month?"Net sales, computed live from completed transactions in the warehouse.
• "Who's our best customer and what should we do next with them?"A person — name, lifetime value, and their open next-best-actions, linked to their 360.
• "Who's our worst customer? How do we win them back?"The other end of the ranking, with the recovery actions already suggested.
• "How many repeat customers do we have?"Customers with two or more purchases — loyalty at a glance.
• "What's our average customer lifetime value?"Straight from customer insight — the same number the associates see.
• "Which product makes us the most money each month?"Product-level revenue ranking from the same warehouse.
• "How much discount are we giving away?"Margin awareness in one question.
• "What's our average basket worth?"And when a question is out of reach, the honest answer: "here's what I can answer."

Actionable Insights: the heartbeat

Intelligence that stays on a dashboard is decoration. Here it lands as Actionable Insights — Next Best Action cards on an associate's home screen that read like a colleague's advice, not a task queue. The name is the promise: not a report to interpret, an action to take, with the insight attached:

• Every card carries its why — the reason in plain words, the signal that raised it, the evidence behind it. "They joined recently and haven't bought yet — a personal hello lands best now."
• Every card carries its worth — an expected value learned from what actually converted, honestly hedged when the evidence is still thin.
• Every card carries its moment — the customer's own best time of day to be contacted when their history says so, the cohort's when it doesn't, and nothing when neither can honestly say.
• Completing a card means something. An outreach card opens the composer and completes on a real send — so completion data is truth, not box-ticking. Captures, conversations, appointments and cases all raise or retire cards, and multi-touch attribution credits the associates whose touches led to the sale.
• The whole business raises them, and every play has a shape. A personal hello for a fresh capture. A winback when a VIP turns at-risk. A churn-risk save from the intelligence engine's own signals. A basket recovery when someone leaves £480 in an online basket overnight ("a nudge from their associate converts better than an email blast"). A relationship repair after a service case ("resolve it, then win them back"). Each type carries its own reasoning, its own learned value, and its own best moment.

For anything longer than one decision, Customer Journeys: multi-step sequences that enrol on any signal, wait durably for days if needed (a wait can hold for the customer's best contact window), branch on the customer's intelligence, act through cards or messages, and exit the moment the goal is reached — a customer who buys leaves the winback journey immediately.

Consent and contact guardrails are decided once, at the platform's choke points, and every acting surface inherits them: opt-outs always win, frequency caps stop pressure stacking, quiet hours have the last word — and a refusal is a recorded outcome, never a failure and never a silent drop.

flowchart LR
  E(["Signal fires
capture · purchase · segment entry"]):::world --> H{"Hold-back
draw"}:::brass
  H -->|"control group"| HB(["Recorded, not actioned
the honest comparison"]):::soft
  H -->|"enrolled"| W(["Wait
durable — hours or days,
optionally until their best window"]):::cap
  W --> B{"Branch on their
intelligence"}:::brass
  B -->|"high value"| A1(["Personal outreach card
to their associate"]):::cap
  B -->|"otherwise"| A2(["Guarded message
consent checked at send"]):::guard
  A1 --> G{"Goal met?
they purchased"}:::brass
  A2 --> G
  G -->|"yes — any time"| X(["Exit immediately"]):::mid
  G -->|"not yet"| W2(["Next step…"]):::soft
  classDef world fill:#2B3036,stroke:#4E5651,color:#F2EFE6,stroke-width:1px
  classDef cap fill:#1E4D40,stroke:#163A30,color:#F2EFE6,stroke-width:1px
  classDef mid fill:#2E6B5A,stroke:#1E4D40,color:#F2EFE6,stroke-width:1px
  classDef soft fill:#E7EEE9,stroke:#B9CCC2,color:#1E3A30,stroke-width:1px
  classDef brass fill:#A87E3F,stroke:#8A6630,color:#FFF8EA,stroke-width:1px
  classDef guard fill:#3A3F45,stroke:#A87E3F,color:#F2EFE6,stroke-width:2px
      

The stage most platforms skip

• A control group at every decision point. A small share of would-be nudges is deliberately held back and recorded — the same decision, minus the action. That's what makes measurement honest.
• Uplift you can defend. Contacted vs held-back purchase rates, per nudge type, per segment, per journey — and the report refuses to print a percentage until both groups are big enough for the number to mean something.
• Send-window intelligence — it learns when, not just what. Every completed outreach records the time of day it went out; the engagement marts turn that history into each customer's own best contact window (their cohort's when theirs is still thin); cards carry the advice, journeys can wait for the moment, and quiet hours always have the last word. The advice appears only when the evidence is real — never a guess dressed as a number.
• The next decision is better than the last. What converted feeds the value on the next card; the control group keeps the measurement honest; and every one of these numbers is visible on a screen, not buried in a model.

Sixty seconds after a sale

The clearest way to see the loop is to follow one receipt. A customer buys a coat at the till. Watch what the platform does — all of it automatic, none of it configured for this specific sale:

• T + 0s — the till

  The sale completes. The associate who nurtured this customer is on the transaction. One business fact — a sale completed — enters the signal fabric.

• Seconds — the fan-out

  The same event, many reactions at once: the warehouse refreshes the datasets this sale touches; the customer's spend history rolls up; multi-touch attribution credits the associates whose captures, messages and appointments led here.

• Seconds — the rescore

  The customer's insight recomputes: lifetime value up, recency reset, churn risk down. They were in the "at risk" segment — this purchase moves them out before the receipt is printed, and that segment exit is itself a new signal.

• Seconds — the journeys react

  They were three days into a winback journey. Its goal — a purchase — is met, so they exit immediately: no awkward "we miss you" message next week. The exit is recorded as "goal reached — they purchased".

• Seconds — the housekeeping

  Any open "win them back" card on an associate's home screen is retired — nobody chases a customer who just bought. Connected external systems get their webhook.

• Minutes — the next suggestion

  The action engine re-crunches: with new value and fresh recency, the right next move might now be a thank-you, a care tip for the coat, or an invitation to the next event — timed to the customer's best contact window, with the reason on the card.

• Days — the proof

  This purchase lands in the uplift ledger. If the winback journey nudged them and the held-back control group didn't buy, the journey's measured uplift just went up — visible on a screen, per journey, per nudge.

That is the loop in one minute of real time: sense, understand, think, act, learn — from a single receipt.

Journeys the loop makes better

Customer journeys

• The welcome. A customer is captured at the counter; the welcome journey enrols the moment the signal fires. It waits a day — then until their best contact window — and asks their associate to say a personal hello. First purchase? The journey exits instantly, goal reached. What the loop adds: instant enrolment, personal timing, a real exit condition, and a held-back control group proving the welcome actually converts.
• The winback. A VIP drifts into "at risk" — the living segment moves them there in seconds and the journey enrols. It branches on who they are: high value gets a personal outreach card; otherwise a guarded message in the brand's voice. The moment they buy, they exit — no "we miss you" after they've returned. What the loop adds: segment-driven enrolment, intelligence-driven branching, consent decided at send, and measured uplift (+41.7 points on the demonstration winback, against control).
• The basket recovery. A customer leaves £480 in an online basket overnight. Their own associate gets the card — the basket contents, the reason, the suggested wording — because a person recovering a basket converts better than a template. What the loop adds: the e-commerce signal landing in a store associate's hands, in the moment.
• The moment. A birthday coming up, a first anniversary as a customer, a loyalty milestone — the insight knows, the card proposes the gesture, the timing is theirs. What the loop adds: the platform remembers so the associate never forgets.

The store colleague's day

• Morning. Open My Hub: today's Actionable Insights, ranked by learned value, each with its why and its moment. No cold list, no guesswork — the day plans itself.
• Before saying hello. One tap to the Customer 360: history, scores, open conversations, cases, next actions — and recommendations built from this customer's own buying patterns. Walk over informed.
• At the counter. Capture a new customer in seconds — consented, deduplicated — and the loop takes over from there: scored immediately, welcomed automatically.
• In the inbox. Replies drafted in the brand's voice, any thread summarised in seconds, sentiment flagging who needs care before it becomes a complaint.
• End of day. Attribution has credited every sale their touches influenced — the relationship work counts, visibly. What the loop adds across all of it: a new starter works with the judgement of a veteran, and the veteran gets their admin time back.

All joined up — out of the box

This is the part that is genuinely hard to build and impossible to fake: every capability below ships connected to the others. One sale at the till touches all of them, with no integration project.

flowchart TD
  subgraph IN["  What happens  "]
    S1(["Sale at the till"]):::world
    S2(["Customer captured"]):::world
    S3(["Message · appointment · case"]):::world
    S4(["E-commerce & imports"]):::world
  end
  HUB(["Signal fabric
durable · self-healing · PII-free"]):::cap
  S1 --> HUB
  S2 --> HUB
  S3 --> HUB
  S4 --> HUB
  subgraph UNDERSTAND["  Understand  "]
    WH(["Insight Warehouse
facts & marts"]):::cap
    ATT(["Sales attribution
who influenced this sale"]):::mid
  end
  HUB --> WH
  HUB --> ATT
  subgraph THINK["  Think  "]
    CI(["Customer insight
value · churn · lifecycle · affinity"]):::cap
    SEG(["Living segments"]):::mid
    REC(["Recommendations
from your own transactions"]):::mid
  end
  WH --> CI
  CI --> SEG
  WH --> REC
  subgraph ACT["  Act  "]
    NBA(["Next Best Actions
why · worth · moment"]):::cap
    JRN(["Customer journeys
wait · branch · act · exit on goal"]):::cap
  end
  SEG --> NBA
  SEG --> JRN
  CI --> NBA
  REC --> NBA
  GUARD(["Consent · opt-outs · frequency caps · quiet hours
decided once, inherited everywhere"]):::guard
  NBA --> GUARD
  JRN --> GUARD
  GUARD --> OUT(["Outreach &
associate action"]):::world
  OUT --> MEAS(["Measured outcomes
uplift vs held-back control · best contact windows"]):::brass
  MEAS -. "tunes the next suggestion" .-> NBA
  OUT -. "back into the loop" .-> HUB
  linkStyle 17,18 stroke:#A87E3F,stroke-width:2.5px
  classDef world fill:#2B3036,stroke:#4E5651,color:#F2EFE6,stroke-width:1px
  classDef cap fill:#1E4D40,stroke:#163A30,color:#F2EFE6,stroke-width:1px
  classDef mid fill:#2E6B5A,stroke:#1E4D40,color:#F2EFE6,stroke-width:1px
  classDef soft fill:#E7EEE9,stroke:#B9CCC2,color:#1E3A30,stroke-width:1px
  classDef brass fill:#A87E3F,stroke:#8A6630,color:#FFF8EA,stroke-width:1px
  classDef guard fill:#3A3F45,stroke:#A87E3F,color:#F2EFE6,stroke-width:2px
      

"If we're using AI, how do we secure our sensitive data?"

flowchart LR
  subgraph PLATFORM["  Inside the platform — sensitive data stays here  "]
    CRM(["Customer records
names · contacts · addresses"]):::world
    WH2(["Insight Warehouse
transactions · history"]):::cap
    SIG(["Behavioural signals
lifecycle · spend level · churn · interests"]):::mid
    ANS(["Answers & rankings
computed in-platform"]):::cap
    CRM -.->|"derives"| SIG
    WH2 -.->|"derives"| SIG
  end
  GATE{"Governed AI gateway
per-capability switches ·
budget · audit"}:::brass
  SIG ==>|"signals only"| GATE
  Q(["Plain-English questions
+ measure catalogue"]):::soft ==> GATE
  GATE ==> LLM(["Cloud AI
your key, your agreement"]):::guard
  LLM --> PROP(["Proposals & wording"]):::soft
  PROP --> VAL(["Schema validation ·
consent & guardrails"]):::cap
  VAL --> ANS
  style PLATFORM stroke:#1E4D40,stroke-width:2px
  classDef world fill:#2B3036,stroke:#4E5651,color:#F2EFE6,stroke-width:1px
  classDef cap fill:#1E4D40,stroke:#163A30,color:#F2EFE6,stroke-width:1px
  classDef mid fill:#2E6B5A,stroke:#1E4D40,color:#F2EFE6,stroke-width:1px
  classDef soft fill:#E7EEE9,stroke:#B9CCC2,color:#1E3A30,stroke-width:1px
  classDef brass fill:#A87E3F,stroke:#8A6630,color:#FFF8EA,stroke-width:1px
  classDef guard fill:#3A3F45,stroke:#A87E3F,color:#F2EFE6,stroke-width:2px
      

Configurable and extensible — by design

Everything above is configuration, not code. That is what makes the loop deployable per brand, per market, per tenant — and extensible without waiting for a platform release.

Journeys themselves are proof of the design: they run on the same general workflow engine — 25+ step types behind one visual designer — that powers catalogue imports, integrations and scheduled operations across the platform. Build a journey today and you are using the same muscle that moves a quarter-million-product catalogue tonight.

And every switch — pillars, AI capabilities, budgets, holdout share, guardrails — is per tenant, so one platform serves many brands at different stages of adoption, safely.

What a retailer gets out of it

The loop's outcomes are the ones a board actually cares about — and each one is a direct consequence of a stage above, not a hope:

• More sales — winbacks that land before the customer is gone, abandoned baskets recovered by a person rather than a template, the right product suggested from your own transaction patterns — and the uplift proven against a control group, per play.
• More loyalty — churn risk spotted and acted on while it's still reversible; every customer's history, preferences and conversations in one place so each touch builds on the last instead of starting over.
• Happier customers — contacted at the time of day they actually engage, never over-messaged (frequency caps), never messaged through a refusal (consent always wins), and reached by someone who knows them — sentiment read from every conversation, so a frustrated customer gets care, not a promotion.
• Targeted communication instead of noise — segments that update themselves in seconds, messages timed by send-window intelligence, wording drafted in the brand's voice — every message fewer, better, and more personal than the blast it replaces.
• Complaints that end in loyalty — sentiment flags the frustration so the customer surfaces for care, the service case tracks resolution against its SLA, and the moment it opens the loop raises the relationship-repair action that follows. Handled this way, a complaint becomes the moment a customer learns the brand actually cares — the strongest loyalty signal there is.
• Fuller appointment books, better visits — when the winning move is face-to-face, the card proposes the appointment; the booking lives on the platform; the associate walks in prepared with the customer's 360 and recommendations from their own history; and the visit itself is a touch that attribution credits and the loop learns from.
• A better day on the shop floor — the platform guides every colleague: who to reach, why, what it's worth, when it lands, and what to say — with the composer one tap away. No guesswork, no cold lists, less admin. A new starter works with the judgement of a veteran, because the guidance carries the experience.

Put together, that is the competitive position: a brand that engages more — and more personally — with every customer than its competitors can, while running leaner: one connected system doing the work of a CDP, a campaign tool, a BI stack, a clienteling app and an automation platform stitched together.

And underneath those, the operational wins:

• Associates with reasons, not lists — every card says why, what it's worth, and when it lands best; opted-out customers are protected without anyone having to think about it.
• A Customer 360 that is actually 360 — history, scores, conversations, appointments, cases and next actions in one place, with every derived number explainable.
• A business you can interrogate directly — plain-English questions, live answers, and a straight "here's what I can answer" when a question is out of reach.
• Journeys that respect customers — enrol on real signals, exit on real goals, never message through a refusal.
• Provable impact — uplift measured against a held-back control, per journey and per nudge, on a screen. Not a vendor claim: your own numbers.
• Day-one demonstrability — a fresh tenant with demonstration data shows the entire loop running, end to end, before a single integration is built.

…and the loop is one pillar of a whole platform

Everything above rides on — and feeds — the rest of the platform: full point-of-sale and commerce (baskets, pricing, promotions, discounts, gift cards), order management and in-store fulfilment (pick, pack, dispatch, reserve & collect, the stock ledger), appointments and events, service cases with SLA hooks, unified communications across providers, and multi-touch sales attribution.

It is multi-tenant to the core, integrates first-class with Salesforce, Shopify, Akinon, SAP and commerce clouds — catalogue imports measured in hundreds of thousands of products — and every capability, including this entire loop, is configured, observed and explained from one Console.

And crucially, the loop is not a module or an add-on: it is foundational — first-class, native, and glued into every pillar of the solution. The till feeds it, the inbox feeds it, the appointment book feeds it, the case queue feeds it — and every one of those surfaces gets smarter because of it. No connectors, no sync jobs, no second system of record.

Native, yet loosely coupled — and that is deliberate. Every stage talks to the next through events and contracts, never hard wiring, so any stage can be swapped without touching the rest: bring your own propensity model, Snowflake-computed scores, or another intelligence engine or data warehouse behind the same contracts — they land in the same action cards; swap messaging providers; swap AI providers; point a stage at your own systems. Loose coupling is how this architecture is always designed. The loop ships complete out of the box — and opens up wherever you want to bring your own.

The loop is the brain; the platform is the whole body it thinks for.