Insights

What Should Reach Marketing Automatically

Automation should move a signal because a defined marketing decision can still change, not because another source produced another record.

The fastest way to make a marketing intelligence system useless is to send marketing everything.

Every call mention, support ticket, product update, and leadership comment becomes another notification. The feed grows. The important signal looks like the routine event beside it. Marketing turns the automation off and goes back to asking people what changed.

The right question is selective: which signals can still change a defined marketing decision?

Automation Is an Editorial Decision

A source event does not earn a route because it exists. It earns a route when the company can explain why marketing needs it, what decision it informs, and when delivery becomes too late.

This makes automation an editorial judgment before it becomes a technical rule. The system needs a view of what marketing is deciding: public claims, active campaigns, positioning, proof, content priorities, or audience expectations.

The previous article, Capturing a Signal and Routing It Are Different Jobs, separates source preservation from delivery. This next decision sits between them. The company must choose what qualifies for a route.

If no marketing decision can change, keep the record available but do not create a notification. If the source is weak or the interpretation is unresolved, send it to a review state rather than pretending it is ready for action.

Use Four Delivery Modes

Automatic delivery should not mean one shared channel. Different signals need different forms of attention.

Interrupt when current work has become wrong or a live decision is at risk. Queue when the item has a clear owner and action but can wait for the next work cycle. Accumulate when recurrence or segmentation creates the value. Hold for review when the signal may matter but its evidence, permission, or meaning is unresolved.

Each mode protects a different resource. Interrupts protect decision time. Queues protect focused work. Accumulation protects the team from reacting to anecdotes. Review protects the company from turning uncertain evidence into a confident instruction.

Interrupt on Factual Changes That Make Work Wrong

A product boundary changes while an old claim remains on a landing page. Approved pricing or packaging changes while a campaign still uses the prior offer. A proof asset expires or loses approval while sales and marketing continue linking to it.

These signals should reach the owner of the affected work as soon as the underlying decision is confirmed. Include the changed fact, effective time, source owner, affected asset, and replacement guidance when it exists.

Do not interrupt on every product activity. A ticket moving columns or a draft proposal changing does not automatically alter the public truth. The route should begin from an approved change with a known marketing dependency.

Queue Buyer Friction With a Clear Next Action

Repeated objections, direct buyer questions, proof requests, and documented content corrections often belong in a governed marketing queue.

The item should arrive with the exact source passage, commercial context, recurrence state, evidence required, and likely asset. A question may call for an FAQ update. A proof request may expose a missing case study or technical page. An objection may require a supported explanation instead of another piece of reassurance copy.

The guide to starting AEO in sales calls explains how these call signals map to public answer jobs. Automatic routing should preserve that distinction instead of converting every signal into a blog topic.

Accumulate Signals Whose Value Comes From a Pattern

Some evidence becomes useful only when it repeats or shifts within a meaningful segment. Customer language drift, emerging competitor framing, support friction, churn themes, and changes in who is asking a question often belong in this mode.

Store every qualifying source with its time and context. Group comparable events. Route the pattern when the company-defined threshold is met or when a scheduled review window arrives.

There is no universal count that turns an anecdote into a pattern. The required evidence depends on event volume, commercial consequence, source diversity, and how reversible the marketing decision is. Record the threshold in advance so the rule does not change to fit the story somebody wants to tell.

Hold Ambiguous or Sensitive Signals for Review

A memorable customer statement may be commercially important and still be unsafe to automate into work. The account may be unusual. The comment may include confidential detail. The internal interpretation may conflict with the source. The proposed claim may lack public proof.

Keep those signals in a review queue with the evidence attached. Name what is unresolved: permission, factual support, classification, or ownership. Route the result only after the missing condition is resolved.

Automation can move evidence to an editor or fact owner. It should not turn an unreviewed interpretation into a public claim.

Give Every Signal a Stable Event Envelope

A routing rule becomes fragile when it depends on reading free-form summaries. Standardize the context the rule needs.

The open CloudEvents specification requires an event identifier, source, specification version, and type. It also defines optional subject and occurrence time attributes. The standard is technical infrastructure, not a marketing model, but its event envelope provides a useful principle: routing metadata should travel separately from the event payload so a system can inspect it consistently.

For a marketing signal, keep the source, event time, signal class, subject, evidence reference, affected decision, destination mode, and route state. Add account or segment context only where the company has permission and a reason to use it.

Make the Eligibility Rule Inspectable

Google Cloud Eventarc requires event filters and a destination for a trigger. Its trigger documentation states that filters match CloudEvents attributes and every trigger must filter on event type.

A marketing route needs the same clarity at the business level. Write the signal type, source conditions, evidence threshold, delivery mode, destination, and exception path in one record. Show which rule matched each delivered signal.

This prevents a black-box feed. The recipient can see why an item arrived, and the system owner can correct a noisy or silent rule without guessing which prompt or integration caused it.

Keep Failed Deliveries Out of the Void

A signal may qualify and still fail to reach its destination. The owner left. The queue was removed. A permission changed. A delivery action failed.

Cloud event systems use explicit exception paths. Amazon EventBridge documents dead-letter queues for undelivered events so failed deliveries can be processed later. Marketing intelligence needs its own visible exception queue with the source, rule, intended destination, failure reason, and age.

Do not reroute every failure to a general channel. Assign a fallback owner who can repair the route or deliberately close the signal.

Measure Decision Value, Not Notification Volume

Count how many eligible source events were captured, how many matched a rule, and how many reached the intended destination. Track route latency and unresolved exceptions. Review false positives that consumed attention and false negatives discovered through later calls or manual audits.

A large notification count is not evidence that the system is useful. The useful measure is whether the right evidence reached the owner before a relevant marketing decision closed.

Pair this with the marketing intelligence latency metric. Speed matters only after eligibility and coverage are honest.

Start With One Decision and One Source

Choose one decision that repeatedly suffers from missing evidence. Define the source events that can change it, the gate, delivery mode, destination, and fallback. Run real records through the rule and review both what arrived and what the rule rejected.

Northstar's Listening Engine applies the eligibility and delivery rules. The Glass Dashboard keeps the evidence, matched rule, mode, owner, timing, and exception state visible. This is a complete core workflow. The Pinnacle remains optional.

The goal is not automatic awareness of everything the company knows. It is timely delivery of the evidence marketing can still use.

Frequently asked questions

What information should reach marketing automatically?

Automatically route signals that can change a defined marketing decision, come from an attributable source, match a stable classification rule, have a useful delivery window, and land with a named owner or governed queue.

Should every customer comment be sent to marketing?

No. One-off comments without clear consequence, source context, or a defined decision should remain in the evidence record or a pattern watchlist. Sending everything creates a filtering job for the recipient.

Which signals should interrupt marketing immediately?

Use an interrupt route when a factual change makes current public work wrong, a live campaign carries a material expectation risk, or a decision change invalidates work already in motion. The company should define materiality and the decision window.

Which signals belong in a marketing work queue?

Route supported buyer questions, objections, proof gaps, content corrections, and asset feedback into a queue when they have a clear owner and action but do not require an immediate interruption.

Which signals should accumulate before marketing receives them?

Customer-language drift, recurring support friction, churn themes, and competitor framing often become useful as patterns. Keep their sources, segment them, and route the pattern when a defined evidence threshold is met.

What should never become an automatic public claim?

Do not turn raw anecdotes, confidential details, unsupported interpretations, or unresolved internal contradictions into public claims. Automation can route evidence for review, but publication still requires an approved statement.

How does Northstar Stack decide what reaches marketing?

The Listening Engine applies source-specific eligibility and delivery rules. The Glass Dashboard shows the signal class, evidence, route, owner, timing, exception state, and the marketing decision each item can change.