Skip to main content

Key Concepts

Customer workspace​

A Customer workspace is your private environment inside BPP. All your data — audiences, signals, AI models, connections — lives in your workspace and is only visible to users on your team. There is no data sharing between workspaces.

Bytek ID​

The Bytek ID is BPP's unified customer identifier. Every person in your data receives one Bytek ID. All their identifiers — email address, phone number, cookie, CRM contact ID — are linked to the same Bytek ID.

This means BPP can:

  • Count unique people in an audience (not just rows or sessions).
  • Accurately attribute conversion events to the right person.
  • Send the correct identifier to an ad platform even when different data sources use different ID types.

Identity Resolution​

Identity resolution is the process BPP uses to link multiple identifiers from multiple sources to a single person. When an email address from your CRM and a cookie from your website appear together in the same data, BPP links them to the same Bytek ID.

This also works retroactively: when a previously anonymous visitor logs in or submits a form, BPP links all their past anonymous activity to their new identified profile.

See User Reconciliation for more detail.

Signal​

A Signal is a scheduled delivery of event-based data from BPP to an ad platform. Signals are how you send offline conversions, CRM lifecycle changes, and value-based scores to Google Ads or Meta Ads.

Each signal has:

  • A trigger — the event data and filter conditions that identify which events qualify (e.g., won deals, completed purchases, subscription renewals).
  • One or more destinations — the ad platform connections that receive the data.
  • A status lifecycle: Draft → Ready → Running → Complete / Error.

Audience​

An Audience is a segment of customers matching conditions you define. You can filter on any attribute in your data — user characteristics, behavioural metrics (computed by Fields Builder), or AI model scores.

Audiences are synced to ad platforms via one or more destinations. Each audience has the same status lifecycle as signals (Draft → Ready → Running → Complete / Error).

AI Model​

An AI Model is a machine-learning or analytical model that runs on your data and produces scores or segments for each customer. BPP supports four model types:

  • Action Prediction — probability score that a user will perform a specific action (e.g., purchase, renew, churn).
  • pcLTV (Predicted Customer Lifetime Value) — estimated future revenue from each customer over a chosen time horizon.
  • RFM Clustering — segments customers by Recency, Frequency, and Monetary value.
  • Interest — classifies customers' interests from their browsing and event behaviour.

Once a model runs, its output scores appear as filter fields in the audience builder — letting you target the top 20% by purchase propensity, or all "Champions" in your RFM segmentation.

Connection​

A Connection is an authenticated link between BPP and an external ad platform or data destination. Supported types:

  • Google Ads — Customer Match (audiences) and offline conversion uploads (signals).
  • Meta Ads — Custom Audiences and Conversions API (signals).
  • Microsoft Ads — Customer Match (audiences).
  • Anonymised — cookieless audience activation.
  • API Endpoint — any custom HTTP destination.

Connection credentials are stored securely and encrypted at rest.

Export / Sync​

An export (also called a sync) is the process of pushing an audience or signal to a connected ad platform. BPP records every sync run in Sync History, tracking how many records were processed, how many succeeded, and what errors occurred.

Fields Builder​

The Fields Builder is BPP's tool for computing per-customer behavioural metrics from your event data. It aggregates your event history into summary statistics (total orders, average spend, last activity date, etc.) that you can then use as filter conditions in audiences and as value inputs in signals.

See Fields Builder for more detail.

Data Warehouse (BigQuery)​

BPP integrates with Google BigQuery as its data layer. Your BigQuery dataset contains:

  • User Tables — one row per customer, with user attributes and at least one identifier column.
  • Event Tables — one row per event, with a timestamp and at least one user identifier.

BPP reads from your data without duplicating it — it only writes back model results, enriched tables, and metadata.

Scheduled sync​

Each audience, signal, and AI model runs on a daily schedule. When you move an audience or signal from Draft to Ready (enabled), BPP sets up automated daily sync jobs. You don't need to trigger exports manually.

Last updated on