From broken weekly ad emails to BigQuery/dbt-validated feed and CDN checks in Campaign Monitor

Overview

A grocery retailer’s weekly ad emails frequently shipped with broken images and mismatched offers because product updates landed late and the image CDN wasn’t checked before send. Merchandising changed items close to launch, URLs shifted, and the email builder had no safety net. We built a product feed manager with validation checks and an image CDN health monitor, wired the curated feed into Campaign Monitor templates, and added fallback offers when assets failed. Deployments became predictable, broken assets were caught before customers saw them, and fewer sends required last?minute fixes.

Client Profile

• Industry: Grocery retail (omnichannel)
• Company size (range): Regional chain with central marketing and local merchandising
• Stage: Scaling weekly promotional email operations
• Department owner: Marketing & Customer Engagement (CRM / Lifecycle / Email)
• Other stakeholders: Merchandising, E?commerce, Creative/Studio, Data & Analytics, IT/Integrations, Legal & Compliance, Customer Support

The Challenge

The weekly ad pulled from multiple sources: the product information system, pricing and promo calendars, and a creative asset library. Late product swaps and path changes were common. When a hero item shifted, its image path or crop changed but the email template kept the old URL. CDN caches weren’t always invalidated, and the send went out with empty slots. Customers clicked into missing images or outdated offers, and the support team fielded complaints.

The build process was brittle. Feeds arrived in different formats, fields were optional, and there was no automated check that images existed or met size requirements. The team relied on manual spot checks and a staging send. If something broke in the last hour, the choice was to delay or ship with known defects. Over time, this created a pattern of emergency edits and lost confidence in the weekly cadence.

Why It Was Happening

Inputs and timing were fragmented. Product data, price and promo flags, and creative assets moved on different schedules. The email tool consumed a raw feed with minimal validation, so missing images, stale URLs, and malformed fields passed through. CDN behavior added another variable: image changes did not always propagate to edge caches before the email landed in inboxes.

Governance arrived late. There was no preflight gate that said “do not send” when a required asset failed, no fallback logic in templates, and no audit trail explaining why an item was swapped. QA depended on manual checks in staging, and even then, inbox rendering differed once the CDN and clients were in the mix.

The Solution

We introduced a governed feed and preflight pipeline that validated product data and assets before the email could ship. A product feed manager assembled a curated list of items for each send window, applied schema and content checks, and verified image health against the CDN. The final, validated feed flowed into Campaign Monitor templates that supported fallback offers when any asset did not meet standards. A monitor tracked origin and CDN behavior and blocked sends when essential assets were not ready. Nothing was replatformed: Campaign Monitor remained the email engine, product data sources stayed intact, and the new layer standardized timing, validation, and fallback behaviors around them.

• Curated product feeds modeled and validated in a warehouse (for example, BigQuery) with tests for schema, required fields, promo windows, and image URLs (dbt)
• Image health checks that issued HEAD/GET requests to the CDN and origin, validated response codes and dimensions, and recorded status
• CDN monitoring and origin failover readiness to reduce stale or missing assets (CloudFront origin failover)
• Email templates in Campaign Monitor that consumed the curated feed and included fallback offer logic and alt?text defaults
• Preflight gate that required a green signal from validation and image health monitors before scheduling
• Automated cache invalidation for changed assets and a recheck loop prior to final approval
• Reason codes and audit logs for item swaps, asset overrides, and send gating decisions
• Monitoring dashboards for feed freshness, validation failures, image health, and send readiness
• Role?based permissions so Merchandising and Creative could update offers and assets while Ops controlled send gates
• Human?in?the?loop exception review for high?visibility hero slots and late changes

Implementation

• Discovery: Mapped product and promo sources, asset storage and CDN behavior, and Campaign Monitor template structure. Collected examples of broken images, stale caches, and mismatched offers. Identified required fields for weekly ad items and the approval path across Merchandising and Creative.
• Design: Defined the curated feed schema and validation rules, image health checks, and send gating criteria. Authored fallback logic and alt?text defaults in templates. Planned cache invalidation triggers and a recheck window. Documented reason codes and audit fields for overrides and blocks.
• Build: Implemented feed models and dbt tests in the warehouse; built image health checkers and logging; configured CDN monitors and origin failover; connected Campaign Monitor templates to the curated feed; added preflight gates and approval steps; and stood up dashboards and alerts.
• Testing and QA: Replayed recent weekly ads through the pipeline, verified that known bad assets were flagged, validated fallback rendering in major clients, and exercised cache invalidation and rechecks. Ran controlled inbox tests to confirm CDN propagation behavior matched the preflight health status.
• Rollout: Operated the pipeline in advisory mode alongside the legacy process, logging validation and health outcomes without blocking sends. After teams validated signal quality and fallbacks, enabled gating and made curated feeds the default input to templates. Kept a manual override lane for time?sensitive swaps with post?send documentation.
• Training and hand?off: Delivered quick guides for Merchandising on feed fields and promo windows, for Creative on asset specs and alt?text patterns, and for CRM Ops on reading dashboards and handling exceptions. Established change control for schema, validation rules, and fallback logic.
• Human?in?the?loop review: Routed failed hero slots and last?minute asset updates to a reviewer queue. Approved decisions captured rationale and updated allowlists or rules to reduce repeat exceptions.

Results

Weekly deployments stabilized. The curated feed and image health monitor caught missing or stale assets before scheduling, templates rendered fallbacks when needed, and cache invalidation ran automatically for late changes. Support saw fewer complaints about broken promos, and the team spent less time triaging day?of issues.

Operationally, the workflow shifted from manual checks to governed signals. Merchandising and Creative worked within clear specs, CRM Ops relied on dashboards and gates instead of inbox spot checks, and leadership gained confidence in the cadence. When exceptions occurred, reason codes and logs explained what changed and why, which made retros productive rather than speculative.

What Changed for the Team

• Before: Raw feeds flowed straight into templates. After: A curated feed with validation and image checks fed Campaign Monitor.
• Before: Broken images surfaced after send. After: Preflight gates blocked sends until assets passed health checks or fallbacks were set.
• Before: Cache behavior was unpredictable. After: Automatic invalidation and a recheck loop aligned CDN state with the email build.
• Before: Fallbacks were improvised. After: Templates included governed fallback offers and alt?text defaults.
• Before: Fixes were urgent and undocumented. After: Reason codes and logs captured swaps, overrides, and gating decisions.

Key Takeaways

• Validate product feeds and assets upstream; schema and image checks prevent downstream surprises.
• Gate sends on readiness signals; a simple preflight pass/fail reduces emergency edits.
• Design templates for failure; built?in fallbacks and alt?text keep emails resilient when assets change late.
• Coordinate with the CDN; cache invalidation and health monitors align edge behavior with deployment timing.
• Log decisions; reason codes and audit trails turn incidents into improvements.

FAQ

What tools did this integrate with?
The curated product feed and validations ran in a warehouse such as BigQuery with tests in dbt. Email builds and dynamic content used Campaign Monitor. Image delivery and monitoring leveraged CDN capabilities, including origin failover in Amazon CloudFront. Lightweight functions handled asset checks and cache invalidation (for example, Cloud Functions), and dashboards surfaced status and failures.

How did you handle quality control and governance?
We encoded required fields and promo windows in the feed schema, added dbt tests for completeness and format, and built image health checks against the CDN and origin. A preflight gate blocked scheduling when critical checks failed. Templates enforced fallback content and alt?text, and all overrides carried reason codes in an audit log. Change control governed schema, validation rules, and fallback behavior.

How did you roll this out without disruption?
The pipeline ran in advisory mode first, producing readiness signals without blocking. Teams compared results with legacy staging sends and tuned rules. Once confidence was high, we enabled gating and shifted templates to the curated feed. A manual override remained for late swaps, with post?send documentation to improve rules.

How were image failures detected and handled?
Automated checks issued HEAD/GET requests to the CDN and, if needed, the origin, validating response codes and dimensions. Failures triggered cache invalidation and a recheck. If the asset still failed near send time, the template rendered a fallback offer with safe copy and alt?text, and the dashboard recorded a reason code for follow?up.

How did you keep pricing and inventory accurate?
The curated feed pulled current price and promo flags from the source systems and enforced promo windows during validation. Items outside the window or missing required fields were excluded or swapped with reason codes. The preflight ran close to scheduling, so last?minute changes were captured, and templates referenced the final validated feed to avoid drift.