Signal-based selling has a strange problem: almost everyone has bought the signals, and almost nobody has written the plays. Teams now run intent feeds, job change alerts, funding triggers, hiring data and visitor identification simultaneously, then route the whole lot into one Slack channel and wonder why reply rates never moved. The data layer is solved. The operating layer — who acts, how fast, with what message, and when to stop — is where the pipeline is won or lost.
This guide is the operating layer. It covers how to build a signal inventory, how to score signals on the two axes that matter, how to write a play card that a rep can execute without thinking, how to route without drowning anyone, and how to measure the program honestly enough to retire what does not work.
What signal-based selling actually is
Signal-based selling is the practice of triggering a specific, pre-defined action when a specific, observable event occurs at an account. Three words in that sentence carry the weight.
- Observable. Something happened in the world that you can date. Not a score, not a probability, not a guess.
- Pre-defined. The response was written down before the event, so nobody improvises at 9am on a Tuesday.
- Specific. One signal maps to one play, with one owner and one deadline.
The underlying logic is timing. The Ehrenberg-Bass 95:5 rule holds that only a small share of any market is in-market at a given moment, and Gartner finds buying groups spend only around 17% of their journey with suppliers. You cannot manufacture demand at scale, but you can arrive at the moment a specific account develops it. That is the entire proposition.
Why signal-based selling fails
Five failure modes account for nearly every disappointing signal-based selling program. Each has a cheap fix.
| Failure mode | What it looks like | Fix |
|---|---|---|
| The undifferentiated firehose | Every signal type lands in one queue with equal weight | Score and tier before routing |
| No named owner | Alerts go to a channel, so they belong to everyone and nobody | One owner per signal type |
| Generic response | Every trigger produces the same sequence | One play per signal, never a shared sequence |
| The creepy open | Reps recite the trigger back to the prospect | Signal chooses the topic, never the opening line |
| No expiry | Signals accumulate into an unworkable backlog | Auto-close every signal at its decay date |
Notice that four of the five are process problems, not data problems. Buying a better feed fixes none of them, which is why so many teams cycle through vendors without changing their results.
Step 1: build a signal inventory
Signal-based selling starts with an inventory. List every signal you can currently observe, and be honest about which ones you can actually act on. Most teams discover they have more signals than plays, which is the whole problem in one line.
| Signal | Source | Strength | Decays in |
|---|---|---|---|
| Demo or pricing form submitted | First-party | Very high | Minutes |
| Three or more visitors from one domain | First-party | Very high | Days |
| Champion joins a target account | Job change feed | Very high | Weeks |
| First-of-kind senior hire posted | Hiring data | High | 1–3 months |
| Competitor comparison page visit | First-party | High | Days |
| Funding round closed | News and filings | Medium | 2–6 months |
| Category intent surge | Third-party | Low to medium | 2–3 weeks |
| Competitor tool detected | Technographics | Low, but persistent | Months |
Two patterns are immediately visible. First-party signals are strongest and decay fastest, so they need automation and short SLAs. Third-party signals are weaker and last longer, so they belong in campaigns rather than in a rep’s daily queue. Building the queue backwards — humans on weak signals, automation on strong ones — is a surprisingly common mistake.
Step 2: score on strength and decay
A signal without a weight is just a notification. Score every signal on two axes: how strongly it predicts a purchase, and how quickly it stops being true. Those two numbers determine the SLA, and the SLA is the actual product of a signal program.
Then apply three gates before anything reaches a human:
- Fit gate. The account must pass ICP. A strong signal at an unqualifiable company is still unqualifiable.
- Coverage gate. A verified contact in a relevant persona must exist, or the signal routes to enrichment rather than to a rep.
- Suppression gate. Open opportunities, current customers, competitors and opt-outs are removed automatically.
Weak signals should require a second, independent signal to clear the gate at all. One intent surge is a guess. An intent surge plus a relevant hire is a reason to pick up the phone.
Step 3: write the play library
This is the step almost every signal-based selling program skips, and it is the one that separates a program from a subscription. A play is not a sequence. A play is a one-page contract that answers seven questions.
- Trigger. The precise event and threshold that fires the play.
- Entry criteria. The fit, coverage and suppression conditions that must also be true.
- Owner. A named role, never a team or a channel.
- SLA. The deadline for first touch, derived from the decay rate.
- Channel sequence. Which channels, in which order, over how many days.
- Message angle. The point of view, the asset, and the ask. Not a template — an angle.
- Exit criteria. What ends the play: a reply, a meeting, or the decay date.
A worked play card
| Field | Value |
|---|---|
| Play name | Buying group forming |
| Trigger | 3+ unique visitors from one company domain within 72 hours |
| Entry criteria | ICP fit, no open opportunity, 2+ verified contacts available |
| Owner | AE who owns the territory |
| SLA | First touch within 24 hours |
| Sequence | Day 0 email to likely economic buyer, day 1 LinkedIn to likely user, day 3 call, day 5 value email to the group |
| Message angle | A point of view on the topic of the pages viewed, plus one artifact. Never mention the visits. |
| Exit criteria | Reply, meeting booked, or 10 days elapsed |
Write one of these per signal type. Start with three, not twelve. The individual signal guides in this series — job change alerts, hiring signals, funding round signals and website visitor identification — each contain the plays to lift into your own library.
Step 4: route within someone’s actual capacity
Routing failure kills more signal-based selling programs than data quality ever has. Four rules.
- Cap the queue. A rep can meaningfully work perhaps five to ten signals a day. Rank and truncate rather than delivering everything.
- Route into the workflow. Signals belong in the CRM and the sequencer, not only in a chat channel that scrolls away.
- Set a fallback owner. If the primary owner misses the SLA, the signal reassigns automatically rather than expiring silently.
- Report the miss rate. Publish SLA compliance by rep and by play weekly. It is the fastest way to make the program real.
All of this presumes a data layer that can be trusted. If your CRM is full of duplicates and stale titles, routing simply distributes bad records faster — the argument set out in why AI fails without a data layer and the field-level hygiene framework.
Step 5: the four rules of signal messaging
The message is where good signal-based selling programs are most often wasted.
- Never state the signal. The trigger determines what you write about. It is not the subject of the email.
- Lead with the consequence. Name the problem the event creates for them, not the event itself.
- Bring one artifact. A benchmark, a checklist, a worked example. Relevance without substance is still noise.
- Match the ask to the moment. Early signals earn a resource. Late signals earn a meeting request.
If a message would still make sense with the signal removed, it is a good message. If it collapses without the trigger, it was surveillance with a call to action attached.
Step 6: measure, then retire what fails
Report by play, never in aggregate. Aggregate numbers hide the two plays carrying the program and the six that are wasting everyone’s morning.
| Metric | What it tells you |
|---|---|
| Signal-to-conversation rate by play | Which signals are real and which are noise |
| SLA compliance by play | Whether routing works in practice |
| Median time from signal to first touch | Whether you are actually faster than before |
| Win rate, signal-sourced versus cold | Whether timing is worth the tooling cost |
| Cost per sourced opportunity by signal | Which subscriptions to renew |
Then act on it. Any play below your cold baseline after 90 days and a fair volume of attempts should be retired, not tuned. Adding a ninth play while three are underwater is how these programs collapse under their own weight.
A 90-day signal-based selling rollout
- Days 1–30. Inventory every available signal, score them on strength and decay, and pick exactly three to start. Write three play cards. Instrument first-party signals properly before buying anything new.
- Days 31–60. Wire routing into the CRM and sequencer with capped queues, named owners and automatic expiry. Publish SLA compliance weekly from day one.
- Days 61–90. Review precision by play, kill the weakest, and add a fourth only if the first three clear their SLA consistently.
Signal-based selling FAQ
How many signals should we start with?
Three. One first-party signal you already own, one relationship signal such as job changes, and one account-level signal such as hiring or funding. Teams that launch with ten never build a real play for any of them.
Does signal-based selling replace outbound?
No. It reorders it. You still need a defined ICP, a target list and a reason to be relevant. Signals decide who to work today and which angle to lead with, which is a prioritisation improvement rather than a replacement for the motion.
Who should own the program?
Revenue operations or a GTM engineer owns the plumbing, scoring and reporting. Sales owns execution of each play. Splitting it that way avoids the common outcome where a marketing-owned signal feed produces alerts nobody in sales has agreed to work. See the business case for GTM engineering for how that role is usually justified.
Can this work without expensive tooling?
Yes, and most teams should start there. Manual signal-based selling works well at first: site behaviour, public job postings, funding news and a hand-tracked champion list cost almost nothing and cover the highest-value plays. Buy tooling once the manual version is working and the constraint is genuinely volume.
The play is the product
Signal-based selling is not a data-acquisition problem. Every competitor can buy the same feeds tomorrow. The durable advantage is the play library: a written contract for each signal that names the owner, the deadline, the sequence and the angle, plus the discipline to retire what does not work. Start with three signals, three plays and one published SLA report. That beats ten feeds and no owner every single quarter.
For the individual signal deep-dives, see B2B intent data, technographic data and sales trigger events, or browse more on B2B growth hacks.
Ready to run plays instead of reading alerts? ZenBee unifies buying and hiring signals across 35M+ companies with 700M+ verified contacts and multichannel LinkedIn and email outreach in a single workflow. Request a demo, or start for free.