Collect data
Pull channel metrics, lead actuals and planning inputs from their most reliable available sources.
Source ingestion
Operational reporting system
Project record 04
Updated 18 Jul 2026
Marketing analytics · Data pipelines & decision systems
Advertising Intelligence Dashboards
A family of reporting systems that connect ad-platform data, lead actuals, plan targets, refresh schedules, anomaly warnings and next-action guidance into dashboards that are practical to review every day.
This record documents the reporting architecture, data flow and operating logic. Employer-private data, client-sensitive dashboards, account identifiers and unpublished campaign details are intentionally excluded.
01 · Overview
Raw metrics are not the real product. Better decisions are.
Ad-platform interfaces expose many metrics, but they do not automatically tell a performance marketer what is happening, what is off-track or what should change next.
The dashboard system reframes reporting as an operating layer. It pulls channel and lead data into a simpler structure, compares actuals against plans, watches for anomalies and presents the numbers with decision-oriented context.
The dashboards are designed for repeated use: fast daily review, quick date filters, visible warnings and a structure that can expand from one campaign or channel to multiple reporting environments without rebuilding the logic from scratch.
02 · Review workflow
The dashboard follows the way a marketer actually reviews performance.
The system is built to answer four questions in order: what happened, how it compares, why it changed and what to do next.
Normalize structure
Map fields, define metric logic and ensure plan, actual and target values can be compared safely.
Metric modelCalculate and compare
Compute trends, deltas, efficiency ratios, pacing and plan-versus-actual performance.
Business logicWarn and explain
Surface anomalies, possible causes, lagging freshness and areas that likely need attention.
InterpretationPrompt action
Convert the review into a next step such as scaling, investigating, fixing tracking or revising the plan.
Decision support03 · Reporting architecture
Use simple components, but keep the flow explicit.
The strongest version of the system separates data extraction, storage, logic, presentation and human interpretation.
Ad platforms
Google Ads, Meta Ads and occasional manual channel inputs provide raw spend, traffic and conversion data.
Channel truthGoogle Sheets
Sheets hold historical daily rows, lead actuals, manual plan values and one-off adjustments when automation is incomplete.
Working datasetApps Script & API sync
Scripts and connectors pull fresh data into sheets on a daily or higher-frequency schedule when appropriate.
Automated refreshHTML report layer
Dashboard pages calculate summaries, trends, plan pacing, warning states and recommendation blocks.
Decision logicHuman operator
The report informs the marketer, but the final interpretation remains tied to business context and channel knowledge.
Action owner
Ad platforms→Scripts / API→Google Sheets→HTML dashboard→Warnings→Next action
04 · Key decisions
A useful dashboard is an opinionated system, not a metric dump.
DR-01Accepted
Track actual leads separately from ad-platform conversions
Platform conversions can be missing, duplicated or delayed, especially when the business funnel extends beyond the ad click.
- Why
- Actual lead logs create a second truth that is closer to the business outcome.
- Trade-off
- Manual or CRM-fed lead maintenance becomes part of the process.
DR-02Accepted
Compare every key metric against a plan when a plan exists
Absolute numbers alone make it difficult to judge whether the system is performing acceptably.
- Why
- Plan-versus-actual framing makes pacing and underperformance immediately visible.
- Trade-off
- The dashboard must clearly distinguish when a plan is absent or incomplete.
DR-03Refined
Add fast date windows instead of forcing manual filtering
A dashboard should support short review loops such as 3-day, 7-day and 30-day checks.
- Why
- Shortcuts encourage regular use and faster anomaly detection.
- Trade-off
- Date logic must remain consistent across summary cards and table views.
DR-04Accepted
Use warning blocks, not only charts
Visuals are helpful, but underperformance still needs clear language that tells the operator where to look.
- Why
- Warnings make the report practical for daily scanning, not just presentation.
- Trade-off
- Warning thresholds and explanation logic require regular tuning.
DR-05Accepted
Allow manual overrides for unsupported one-off spend sources
Not every platform or spend source warrants a full pipeline immediately.
- Why
- Small manual adjustments can keep total spend or lead views accurate while the system matures.
- Trade-off
- The report must show which numbers are automated and which are manually maintained.
DR-06Accepted
Separate data freshness from performance interpretation
A dashboard can look wrong simply because the source data has not updated yet.
- Why
- Freshness indicators prevent the operator from making decisions on stale or delayed data.
- Lesson
- “Real-time” claims should match the actual limitations of the source pipeline.
05 · Warning system
The dashboard becomes valuable when it knows what deserves attention.
Actuals behind plan
Trigger a visible warning when spend, leads or qualified outcomes are materially below the expected pace.
Cost rising faster than outcomes
Flag CPC, CPL or CPA deterioration when the change is sustained rather than a single-day spike.
Missing or implausible conversions
Warn when spend is arriving but conversions or leads are unexpectedly zero or inconsistent with prior behavior.
Source data lag
Show when the latest row is older than expected so the operator knows to verify the pipeline before diagnosing performance.
One channel distorting totals
Call out when a new manual or low-scale source changes overall spend or lead totals and needs explanation.
Next-step recommendation
Conclude the warning section with a practical instruction rather than stopping at the diagnosis.
06 · Public / private boundary
Share the reporting system, not the accounts behind it.
The public record should preserve the architecture and operating lessons without exposing business-sensitive campaign details.
Safe to document
- Pipeline structure across ad sources, sheets and reports.
- Metric categories such as spend, leads, plan and warnings.
- Date-filter, pacing and anomaly-detection design patterns.
- Data-freshness logic and dashboard usability decisions.
- General lessons about automation, reporting and decision quality.
Never publish
- Private account IDs, API keys or connector credentials.
- Employer-private, client-private or unpublished spend data.
- Campaign names or targeting details that are commercially sensitive.
- Lead-level information, CRM exports or personal data.
- Unreleased business plans, internal targets or internal recommendations tied to confidential context.
07 · Lessons learned
The report is only as good as the measurement model behind it.
Decision-ready metrics beat full-platform exports
The most useful report often contains fewer metrics, but arranged around the actual decisions the operator needs to make.
Plan and actual should coexist naturally
Reports become sharper when targets live beside performance instead of in a different file or mental model.
Data freshness is part of trust
An elegant dashboard still fails if the user cannot tell whether the numbers are current enough to act on.
Warnings need explanation, not just thresholds
Highlighting a bad number is not enough; the report should guide the next investigation step.
Manual inputs are acceptable when they are explicit
Small hardcoded or sheet-maintained adjustments are useful as long as the system makes them visible and temporary.
Automation should reduce review time, not create another system to maintain blindly
The right level of automation is the one that improves visibility and still remains understandable.
08 · Next priorities
Move from reporting coverage to stronger diagnostic intelligence.
The next evolution should reduce interpretation time further and make anomalies more precisely attributable.
Formalize metric definitions
Document which fields are raw, derived, planned, actual and manually adjusted across every dashboard.
Add anomaly confidence logic
Distinguish between single-day noise and meaningful multi-day signals before generating a warning.
Improve cross-channel attribution context
Show when performance changes may be driven by funnel-stage behavior outside a single ad platform.
Expand delivery options
Allow summary exports or notifications while keeping the dashboard itself as the main review destination.
Extract reusable reporting playbooks
Turn the strongest dashboard patterns into SOPs and templates for future systems.
Project record status
Fourth encyclopedia record complete.
The advertising dashboards now have a public record covering their data flow, decision logic, warnings and privacy boundary—without exposing private campaign information.