Signal Architecture B2B: How to Route Triggers Into Revenue Actions

Picture a signal feed with 14 detection sources, a Slack channel collecting hundreds of alerts a month, and a CRM where almost none of those alerts became a logged task. I see versions of this setup constantly. The detection is excellent. The pipeline impact is close to zero.

That gap is the whole story of signal architecture B2B teams need to fix. Detection got cheap. Job changes, funding rounds, pricing page visits, tech stack changes and hiring surges are all one integration away. What did not get cheap is the part after detection: deciding who owns a signal, how fast they must act, what action is correct, and when the signal stops being worth anyone’s time.

We already covered which signals to track daily and how to reach buyers before intent becomes obvious. This post covers the layer neither of them specifies: the routing infrastructure. Here is the five stage framework I use to build it, called the Signal Router.

Why Detection Without Routing Fails

Speed matters, and everyone agrees on that. The commonly cited MIT and InsideSales lead response study claims a lead contacted within 5 minutes is far more likely to qualify than one contacted after 30. I would not stake a strategy on the exact multiple; much of what circulates in 2026 traces back to a decade old dataset and vendors citing each other. Treat it as direction, not a constant.

The direction is still right, and it applies more strongly to outbound signals than to inbound forms. A pricing page visit loses value within a day. A social post that mentions your problem is stale in about two days. A funding round gives you weeks. Vendor decay tables disagree on the exact numbers, but they agree on the shape: value falls fast, then flattens.

Here is what I see when signal programs fail. It is rarely a detection problem. It is one of four routing failures:

• No owner. The alert lands in a shared channel. Everyone assumes someone else has it.
• No deadline. The alert has an owner but no clock, so it competes with every other task and loses.
• No action. The rep sees “Acme hired a VP Sales” and has no idea whether to email, call, connect on LinkedIn or do nothing.
• No decay. Six week old signals sit in the queue next to fresh ones, so the queue looks full and nobody trusts it.

Lean Data makes a similar point about inbound leads: slow follow-up usually traces back to routing logic, enrichment delays and unclear ownership rather than lazy reps. I agree. Fix the plumbing before you coach the people.

The Signal Router: Five Stages From Trigger to Revenue Action

The framework is deliberately boring. Each stage has one job, one output, and one failure mode you can measure.

Stage	Job	Output	Typical failure
1. Capture	Collect raw events from every source	Event with timestamp, source, entity	Duplicate and missing events
2. Resolve	Match event to account, contact and CRM record	Clean entity match	Orphan signals with no account
3. Score	Rank by strength, fit and freshness	Priority tier (fast, medium, slow)	Everything marked “high”
4. Route	Assign owner, backup, SLA and action	Task with deadline	Shared ownership
5. Retire	Expire, suppress or recycle the signal	Closed outcome	Zombie signals

Most teams build stage 1 and stop. The value sits in stages 2 through 5.

Stage 1: Capture Into One Event Format

Every source speaks a different language. A job change webhook, a website visit from your analytics tool and a funding alert from a news feed all describe different things. Before anything else, convert each into one event shape with the same required fields:

• Event type from a fixed list (for example: champion_moved, funding_round, pricing_visit, hiring_surge, stack_change)
• Entity (the company domain and, when known, the person)
• Detected at and occurred at timestamps (these differ more than you expect)
• Source and a confidence label
• Raw payload link for audit

The detected versus occurred timestamp split matters. A funding round announced Tuesday and detected Friday has already burned three days of a window. Track the lag per source and you will quickly find which of your feeds is too slow to be useful.

Keep the event type list short. A taxonomy of 60 event types guarantees that nobody writes a routing rule for most of them. Start with seven.

Stage 2: Resolve to a Real Account

This is the dullest stage and the most damaging when skipped. A signal that cannot be matched to an account is noise. Resolution has three steps:

1. Domain match. Normalize to the root domain, strip subdomains, and handle parent and subsidiary relationships.
2. Contact match. Match by email first, then LinkedIn URL, then name plus company.
3. State lookup. Is this account a customer, an open opportunity, a closed lost deal, a prospect in sequence, or unknown?

The state lookup is where routing gets smart. The same signal means different things by state. A funding round at a customer is an expansion trigger. At an open opportunity it is a deal acceleration note. At a cold prospect it is an outbound trigger. At a closed lost account it is a reopen trigger. One event, four different actions.

If your CRM data is poor, this stage will expose it immediately. Industry sources put annual B2B contact data decay anywhere from about 22 to 70 percent depending on the study, and every percentage point of decay is a signal that routes to the wrong person or nobody. We go deeper on keeping that foundation clean in our RevOps system guide.

Stage 3: Score With Three Inputs, Not One

Most scoring models add up points per signal. That works for lead scoring and fails for routing, because routing needs to answer a different question: how fast, and by whom. I score each signal on three inputs and combine them into a tier.

Input	What it measures	Example scale
Strength	How strongly the event predicts a purchase window	1 to 5
Fit	How well the account matches your ICP	A, B or C
Freshness	Hours since the event occurred	Under 4h, under 24h, under 72h, older

Then the tier logic is simple enough to write on a sticky note:

• Fast: Strength 4 or 5, fit A or B, freshness under 24 hours. Respond within 4 business hours.
• Medium: Strength 3 or higher, fit A, B or C, freshness under 72 hours. Respond within 2 business days.
• Slow: Everything else that passes the minimum fit bar. Weekly batch.
• Drop: Fit C with strength under 3, or anything older than its decay window.

Notice what is missing: a 0 to 100 score. A single number invites debates about whether 67 is different from 71. Three tiers force a decision, and each tier maps to an SLA. If you want a points model for lead scoring, our 15 signal point table is a reasonable starting point. For routing, tiers are easier to run.

Stage 4: Route With a Fixed Ownership Order

Routing has four decisions: owner, backup, deadline and action. Make them rules, not judgment calls.

Ownership order. Evaluate top to bottom and stop at the first match:

1. The account owner, if the account is a customer or has an open opportunity
2. The owner of the named contact, if a champion moved or a known person triggered the event
3. The territory or segment owner
4. Round robin across the active pool, weighted by current open task load

Round robin is the fallback, not the default. Pure round robin sends a signal about a customer’s expansion to a rep who has never spoken to them. That is how a good signal becomes an awkward email.

Backup and escalation. Every owner gets a backup. If the task is untouched at 50 percent of the SLA, ping the owner. At 100 percent, reassign to the backup and notify the manager. An alert nobody is accountable for is a decoration.

The action map. The rep should never wonder what to do. Each event type and account state combination maps to one default action:

Event	Account state	Default action	Channel
Champion moved	Past customer	Personal note referencing prior result	Email plus LinkedIn
Champion moved	Never customer	Intro message with one relevant asset	LinkedIn first
Funding round	Cold prospect	Enroll in trigger sequence	Email
Funding round	Customer	Expansion check in, flag to CS	Task for CSM
Pricing page visit	Open opportunity	Alert AE, send follow up within the hour	Email or call
Pricing page visit	Unknown	Enrich, then route as inbound if fit A or B	Email
Hiring surge	Cold prospect	Trigger sequence with hiring angle	Email
Stack change	Any	Weekly batch, research first	Email

This table is the single most useful artifact in the system. Write yours in a spreadsheet before you touch any automation tool. Our breakdown of fifteen trigger events and the angle for each gives you the content to fill the action column.

Stage 5: Retire Every Signal Explicitly

Every signal should end in one of five states: worked (action taken, outcome logged), suppressed (deliberately skipped, with a reason), expired (passed its decay window untouched), merged (combined into a stronger multi signal event) or recycled (parked for a later review date).

Expired is the state that tells you the truth. If 40 percent of your fast signals expire untouched, you do not have a prioritization problem; you have a capacity or routing problem. Count expirations weekly and show them to the team.

Merging matters too. Two weak signals on the same account within a short window often beat one strong signal alone. A hiring surge plus a new VP Sales plus a visit to your comparison page is one conversation, not three tasks to three different people. Build a merge rule: if two or more signals hit the same account inside 14 days, collapse them into one task at the highest tier, owned by one person, with all the context attached.

Decay Handling: The Part Everyone Skips

Signals lose value, and the curve depends on the type. These are working defaults, not laws; I have seen vendor tables disagree by a factor of two for the same signal, so calibrate against your own data after 60 days.

Signal class	Working window	Decay behavior	Routing consequence
Behavioral (pricing visit, demo page)	1 to 3 days	Drops sharply in 24 hours	Fast tier, real time alert
Social and public posts	2 to 5 days	Stale after a few days	Fast tier, same day
People moves (champion, new exec)	30 to 90 days	Strongest in first weeks	Fast first, then keep warm
Funding	30 to 90 days	Slow burn, evaluation starts later	Medium tier, sequence
Hiring surge	30 to 60 days	Moderate	Medium tier
Stack or compliance change	60 to 180 days	Slow	Weekly batch

Two rules fall out of this table.

Match the cadence to the window. Unify’s point on this is worth repeating: a 30 day sequence on a signal with a 24 hour window spends most of its life after the signal has expired. A fast signal gets a short, direct touch sequence of 2 to 3 steps over about a week. A slow signal can carry a longer nurture.

Re-score on a schedule. Freshness is an input to the score, so the tier must change over time. A fast tier signal that nobody touches for three days should not stay at the top of the queue. Run a re-score job every few hours for fast signals and daily for the rest. Anything past its window moves to expired automatically, and the system should record it as a miss, not delete it.

What to Build First: A 30 Day Rollout

You do not need a data warehouse. A CRM, an automation tool and one spreadsheet are enough to run the first version. If you are connecting several tools, our GTM workflows guide covers the plumbing patterns.

Week	Build	Done when
1	Pick 5 to 7 event types. Write the event schema and the action map in a spreadsheet.	Every event type has an owner rule and a default action
2	Wire capture and resolve. Push events to the CRM as tasks against the matched account.	90 percent of events match an account automatically
3	Add tiers, SLAs, backup owner and escalation pings.	Every task shows owner, tier and deadline
4	Add retire states, re-score job and a weekly review dashboard.	You can report worked, suppressed and expired counts

Keep the first version small on purpose. Five signals that all get worked beat fifteen that mostly expire.

The Weekly Review

Block 30 minutes weekly for a signal review. The agenda is fixed: expirations by tier, SLA hit rate by owner, signals that produced a meeting, and one routing rule to change. If a rule does not change, you are not learning.

Metrics That Show Whether the Architecture Works

Measure the system, not the activity. These are the metrics I would put on the first dashboard.

Metric	Definition	Starting target
Detection lag	Occurred time to detected time, per source	Under 4 hours for fast sources
Match rate	Share of events resolved to an account	90 percent or higher
SLA hit rate	Tasks actioned inside their tier deadline	85 percent for fast, 90 percent for medium
Expiry rate	Tasks that expired untouched, by tier	Under 10 percent for fast
Signal to meeting rate	Worked signals that became a meeting, by event type	Track by type; compare to your cold baseline
Signal to pipeline	Pipeline created per 100 worked signals	Compare to non signal sourced pipeline

The targets are starting points I set for a first month, not industry benchmarks, so adjust them once you have your own baseline. The comparison that decides whether to keep a signal is simple: does it convert to meetings better than your cold baseline? If a signal type does not beat cold after 50 worked instances, demote it or cut it. That is how a 15 signal program shrinks into the 6 that earn their place.

Common Mistakes

1. Routing everything to Slack. Slack is a notification channel, not a system of record. A message cannot be assigned, escalated, expired or counted. Put the task in the CRM and let Slack carry a pointer.
2. One tier for everything. If all signals are urgent, none are. Force a distribution: in a healthy system, fast signals are a small slice of the total.
3. Letting reps pick which signals to work. Cherry picking means the easy ones get worked and the valuable ones expire. Assign by rule.
4. No account state check. Sending a cold style sequence to a customer because a funding alert fired is a fast way to annoy your best accounts.
5. Never deleting signal types. Every type has a maintenance cost. Review quarterly and cut the ones that fail the cold baseline test.
6. Automating the message before fixing the route. An AI written email on a signal nobody owns is still an email nobody sends. Route first, then automate the action.
7. Ignoring suppression. Without a do not contact list and a cooldown window, one account can receive three signal triggered touches from three reps in a week.

Where AI Agents Fit

The stages with the most manual drag are resolve and score: matching messy entities, reading a hiring surge and judging strength and fit. Those are good candidates for an agent, as long as the routing rules themselves stay deterministic and visible. The agent proposes the tier and the context; the rule table decides the owner and the deadline. If you want to see how we approach this, our AI agents and outbound automation pages describe the building blocks.

The Plain Version

A signal is a perishable asset. Detection finds it, but only routing makes it worth anything. Pick five signals, write the action map, assign one owner and one deadline to each, and measure how many expire untouched. That single number, the expiry rate, will tell you more about your revenue operation than any dashboard of intent scores.

If you want a second pair of eyes on your own signal routing, book a free growth audit and we will map where your triggers are leaking. You can also try our free AI growth tools at app.momentumnexus.com.