Skip to main content

Audience Insights & Ad Attribution Architecture

DriveMetaData (DMD) provides an analytical post-evaluation engine designed to answer two mission-critical questions before and after audience activation:

  1. Audience Profiling: What are the statistical characteristics, demographic clusters, behavioral triggers, and reachability limits of this audience?
  2. Channel Performance & Attribution: Which marketing channels, ad networks, and campaigns drove the conversions, and how efficiently was capital deployed?

1. Post-Evaluation Analytics Pipeline​

Unlike static customer lists that only track whether a customer qualifies for a rule, enterprise audience management requires deep diagnostic post-evaluation.

When an audience is saved or evaluated in DMD:

  1. Rule Resolution: The analytical engine evaluates the audience criteria across ClickHouse partitions.
  2. Statistical Profiling: The insight engine aggregates the matching cohort across seven analytical dimensions (demographics, purchase intent, recency risk, traffic source, engagement frequency, commerce metrics, and technographics).
  3. Attribution Aggregation: The ad attribution engine maps customer conversions back to touchpoints recorded in the AttributionAdsPerformanceDaily cube.

2. Multi-Touch Ad Attribution Modeling​

To measure channel performance objectively, DMD attributes marketing revenue and conversion events across paid channels (Google Ads, Meta Ads, TikTok Ads) and organic touchpoints using deterministic and probabilistic attribution models:

Attribution Models Supported​

ModelWeighting StrategyBest Used For
First Touch100% of conversion credit is assigned to the initial touchpoint.Acquisition analysis and top-of-funnel brand discovery.
Last Touch (Direct-Excluded)100% credit is assigned to the final non-direct marketing interaction.Immediate conversion driver evaluation and tactical ad spend optimization.
Linear / Multi-TouchEqual credit is distributed across all recorded touchpoints in the lookback window.Balanced full-funnel customer journey analysis.

Blended Spend & Cross-Channel Normalization​

Ad platforms report performance in disparate currencies, timezones, and metric formats. DMD's query engine performs:

  • Currency Normalization: Automatic currency conversion into the tenant's base reporting currency based on the exchange rate at event timestamp.
  • Workspace Timezone Alignment: Realigns daily ad spend batches and ClickHouse conversion events to the active workspace timezone (getActiveWorkspaceTimezone()), preventing day-boundary attribution skew.

3. Statistical Audience Profiling & Reachability​

Before deploying high-budget campaigns, marketing operators must understand audience viability. An audience with 100,000 users is useless if only 5,000 can be matched on Google or Meta.

Reachability Rate Calculation​

The Reachability Rate quantifies the proportion of qualifying customer profiles that possess valid, consented identifiers required for third-party ad synchronization:

Reachability Rate (%) = (Profiles with Verified Channel Identifiers / Total Qualifying Audience Size) * 100
  • High Reachability (>70%): Ideal for direct paid activation (hashed email, phone number, or Mobile Ad ID present).
  • Moderate Reachability (40%–70%): Suitable for blended omnichannel campaigns (push notifications, on-site personalization, and paid retargeting).
  • Low Reachability (under 40%): Triggers an Activation Guardrail in the UI, warning operators that paid ad network match rates will be severely degraded.

Confidence Scoring Engine​

The insight engine assigns a statistical Confidence Score (High, Medium, Low) based on:

  1. Cohort Sample Size: Audiences with fewer than 50 records display low confidence warnings to prevent over-indexing on statistical noise.
  2. Data Freshness: Indicates whether behavioral touchpoints have synced within the last 24 hours.
  3. Definition Version Stability: Evaluates whether the audience rules have remained stable across execution runs.

4. Analytical Cube Architecture: AttributionAdsPerformanceDaily​

Channel performance reporting queries the pre-aggregated ClickHouse analytical cube AttributionAdsPerformanceDaily. This cube indexes advertising data into optimized dimensions and measures:

Dimension MemberDescriptionExample Values
channelMarketing channel or ad platformGoogle Ads, Meta Ads, TikTok Ads, Email
campaignSpecific ad campaign nameSummer_Sale_2026_In, Retargeting_LTV_Tier1
countryUser geographic countryIndia, United Arab Emirates, Singapore
deviceClient hardware platformmobile, desktop, tablet
timeAggregation date (metrics.timestamp)Daily partition records
Measure MemberCalculationDescription
spendsum(spend)Total marketing investment in base currency.
impressionssum(impressions)Raw ad impression volume.
clickssum(clicks)Total user click actions.
purchasessum(conversions)Number of attributed orders.
revenuesum(conversion_value)Gross attributed revenue generated.

5. Multi-Tenant Security & Privacy Controls​

  • Tenant Data Segregation: Every query executed against the analytical engine strictly enforces Client-ID and Workspace-ID partition filters at the ClickHouse query compilation level.
  • Sensitive Attribute Masking: Performance tables automatically filter out direct PII columns (isDimensionMetadataMember), ensuring aggregated marketing reports never expose sensitive personal data to unauthorized team members.