Performance & Reconciliation
POCNormalized cross-platform reporting and platform-to-online-sales reconciliation. One reporting period, common definitions, one source of truth.
Normalized cross-platform reporting
Common definitions and one reporting period, normalized from platform exports · POC.
| Channel | Spend | Impressions | CTR | CPC | Purchases (reported) | CPP | ROAS |
|---|---|---|---|---|---|---|---|
| Meta | RM 8,100 | 620,000 | 1.4% | RM 0.93 | 260 | RM 31.15 | 4.20x |
| TikTok | RM 5,200 | 940,000 | 1.8% | RM 0.31 | 190 | RM 27.37 | 4.04x |
| Shopee | RM 4,800 | 310,000 | 2.4% | RM 0.65 | 235 | RM 20.43 | 4.38x |
| Total | RM 18,100 | 1,870,000 | — | — | 685 | — | 4.20x |
Reported revenue rolls up to RM 76,000 — matching the Dashboard Reporting Truth tile.
Platform vs online-sales reconciliation
Matched against periodic online-sales export · POC. Near-real-time reconciliation is Full Production.
| Channel | Reported purchases | Matched online orders | Reported revenue | Matched revenue | Unmatched | Revenue reconciliation |
|---|---|---|---|---|---|---|
| Meta | 260 | 232 | RM 34,000 | RM 29,800 | RM 4,200 | 88% |
| TikTok | 190 | 172 | RM 21,000 | RM 18,900 | RM 2,100 | 90% |
| Shopee | 235 | 213 | RM 21,000 | RM 19,700 | RM 1,300 | 94% |
| Total | 685 | 617 | RM 76,000 | RM 68,400 | RM 7,600 · 68 orders | 90% |
Agency vs agent-assisted comparison
Illustrative result format using synthetic demo data. Controlled comparison on a limited campaign set · POC. Continuous benchmarking is Full Production.
| Arm | Setup | Spend | Purchases | CPP | Platform revenue | Platform ROAS | Matched revenue | Matched-sales ROAS |
|---|---|---|---|---|---|---|---|---|
| Agency campaign | Existing agency | RM 5,000 | 120 | RM 41.67 | RM 21,000 | 4.2x | RM 19,000 | 3.8x |
| Agent-assisted | Agent workflow | RM 5,000 | 138 | RM 36.23 | RM 24,000 | 4.8x | RM 21,500 | 4.3x |
Illustrative result format using synthetic demo data. Matched-sales ROAS is computed on screen as matched revenue ÷ spend, so the reconciliation gap between platform and matched figures is visible and auditable. Results remain directional because both arms may rely on platform optimization.
Autonomy mode
Earn autonomy as trust builds. Only Review is active at POC.
Agent recommends; every material change waits for human approval.
Agent applies changes within pre-approved bounds (budget, creative, audience).
Agent optimizes end-to-end with escalation only on exceptions.
Recommendations
Human-in-the-loop. Approve or dismiss — the agent never auto-applies.
Agent-assisted arm delivered CPP RM36.23 vs agency RM41.67 at equal spend, with a stronger matched-sales ROAS.
Expected: Sustain the CPP advantage into the next reporting period.
RM7,600 of platform-reported revenue did not match the online-sales export — likely UTM / click-ID coverage on Meta and TikTok.
Expected: Higher reconciliation confidence and cleaner ROAS reporting.
The agent prepares and coordinates the campaign package while the agency provides final creative polish where needed.
Expected: Clear separation of scope, no duplicated effort.
Normalized reporting and reconciliation now run end-to-end from exports — the agency's manual weekly deck is redundant.
Expected: Faster reporting cycle and one source of truth.
Continuous optimization
Automated benchmarking, fatigue detection, alerting, and budget/creative changes.
Continuous variant scheduling, holdouts, and lift readouts.
Per-channel frequency, CTR decay, and refresh cues.
Alerts when reconciliation, ROAS, or CPP drifts outside band.
Applies safe optimizations within Guardrail bounds.