Every offer competes on value. Every decision explains itself.

• 01

  Every offer on one scale

  Each candidate is reduced to one expected-value score, so a broadband save and a handset upsell from different portfolios compete fairly for the same customer.

  The portfolio is set by expected value, not by whichever team shouts loudest.

• 02

  Live behaviour moves the answer

  Events such as roaming, a low balance or the end of a contract lift propensity within a fixed bound, and each signal fades at its own pace.

  Offers reach customers while they are in the market, and no single event can override commercial intent.

• 03

  Compliance no one can switch off

  Consent, Do Not Contact and permitted contact hours are hard gates. Fatigue rules relax only through an approved override with a reason and an expiry.

  Lower complaint and opt-out risk, and a contact policy compliance can review on its own.

• 04

  No winner left on the table

  When the best channel is full, the customer moves down the channel ladder instead of being dropped. Capacity losses are reported apart from policy blocks.

  More of the eligible base reached on the capacity you already pay for, and a clear case when you need more.

• 05

  Every decision explained

  Every candidate, selected or not, is kept with each factor, the stage that removed it, the values it evaluated and the release it ran under.

  Answer “why did this customer get that offer?” for auditors, regulators and campaign owners.

• 06

  Agents do the groundwork

  Describe the goal and the KPIs. Agents draft the campaign, compile the rules, generate the tests and explain the results.

  Planners spend their time on strategy instead of SQL, spreadsheets and run logs.

• Campaign design

  • Audiences built from business fields with AND, OR and NOT, compiled to SQL and sized with a dry run.
  • Describe a segment in plain language and get a valid audience rule to review.
  • SMS, email and push treatments, written by hand or drafted with AI.
  • Recurring schedules, delivery windows, offer validity and reminders that stop when the customer responds.
  • Decide per line, per person or per account, and combine a customer's lines into one message.

• Decisioning and arbitration

  • Expected-value scoring across every live campaign, with safe fallbacks for missing or invalid factors.
  • Portfolio, Issue, Group and Offer taxonomy with inherited, dated priorities.
  • The full ranked list kept for every customer, with deterministic tie-breaks.
  • Channel allocation within campaign caps, treatment volumes and channel capacity.
  • Universal and campaign control groups and A/B variants, assigned deterministically.

• Real time and events

  • Next-best-offer API for apps, websites, IVR and the call centre.
  • Response capture for impressions, clicks, acceptances, rejections and opt-outs, with duplicates recognised.
  • A governed event catalogue, with a decay rate per event and unknown events logged, not applied.
  • Event-triggered campaigns that decide for one customer the moment something happens, inside the channel's delivery window.
  • Fallback offers keep channels answering if the decision service is unavailable.

• Contact policy and consent

  • Consent and Do Not Contact as hard gates, where a missing flag means no.
  • Permitted contact hours in the recipient's local time. Out-of-hours contacts are held, not dropped.
  • Frequency caps and minimum gaps scoped to channel groups, from global down to a single channel.
  • Suppression after an outcome, at Offer, Group or Issue level.
  • Overrides relax fatigue rules only, and need approval, a reason and an expiry.

• Batch execution and delivery

  • Portfolio runs where every campaign due in a slot competes in the same run.
  • Run locks, retries, pause and controlled restart, with status recorded for every stage.
  • Validation gates before any file leaves the platform.
  • Export layouts per channel, delivered to cloud storage or over SFTP.
  • Delivery receipts fed back into history, so reporting and contact policy see what was actually delivered.

• Analysis and simulation

  • Run waterfall from eligibility to export, with a reason code at every drop.
  • Rivalry matrix showing which campaigns win and lose head to head.
  • Audience overlap to spot cannibalisation before launch.
  • Customer drill-down that explains a result field by field.
  • What-if simulation of values, levers and caps, with displacement against the live baseline and no access to production.

1. 01 · Brief

   Briefing agent

   Turns a business goal and target KPIs into a review-ready brief or a complete draft campaign, based on what worked before.

2. 02 · Build

   Development agent

   Converts the approved brief into audience rules, prioritisation settings, treatments and volume controls, as a versioned package.

3. 03 · Test Coming soon

   Testing agent

   Generates positive, negative and boundary cases with synthetic data, runs them and records the evidence.

4. 04 · Operate Coming soon

   Operations agent

   Checks approvals and evidence before go-live, starts authorised runs, watches volumes, and retries, pauses or escalates.

5. 05 · Explain

   Explanation agent

   Explains in plain language why a run lost volume or a customer missed an offer, down to the exact field.

Agents draft, people decide. Agent output passes the same validation and four-eyes approval as any human change, and every prompt and approved version is kept.

• Head of CVM

  “Steer the whole portfolio by value, not by whose campaign is loudest.”

  • Raise an Issue's priority once and every campaign beneath it moves up.
  • See who wins, who loses and what it displaces in simulation, before you approve.

• Campaign planners

  “Describe the campaign. Review the draft. Ship it.”

  • Build audiences from business fields, with no SQL to write.
  • Agents prepopulate the campaign from a brief and what worked before.

• Risk and compliance

  “Consent and quiet hours that no override can reach.”

  • Every contact decision records the rule and the values it evaluated.
  • Append-only history, with personal data kept out of the logs.

• Architects and IT

  “Real-time answers that never wait on the warehouse.”

  • Batch, real time and simulation share one decision engine and one approved release.
  • Single sign-on, role-based access and separate staging and production environments.