The checklist
- ICP written as observable data fields, not adjectives
- Disqualifiers defined before qualifiers
- At least five signal sources wired, one of them custom
- Every signal has a decay date
- Accounts scored A, B or C by stacked signals, not gut feeling
- Enrichment runs cheapest source first, credits on hard finds only
- Every email validated before it leaves Clay
- List refreshes continuously, not quarterly
- Scores sync to the CRM automatically
- Tier A gets outreach within 48 hours of qualifying
Static lists lose before they start
B2B contact data decays at roughly a quarter to a third per year. People change jobs, companies pivot and priorities move. A list exported once and worked for a quarter is stale before the second sequence ends, and it shows: volume outreach on static lists answers at under 2 percent.
Signal based targeting inverts the question. Instead of asking who fits our market, it asks who fits our market and is in motion right now. That single change moves reply rates to 8 to 15 percent, and stacking signals moves them to 15 to 25. The list stops being a file and becomes a living query.
Write the ICP in data, not adjectives
An ICP that says modern B2B companies that value growth cannot be queried. An ICP that says 50 to 500 employees, B2B software or tech services, selling into the US or DACH, at least one revenue role hired in the last year, no GTM engineer on staff, can be built as a table. Every trait must map to a field some source can actually return.
Write the disqualifiers first: industries you will not serve, deal sizes that do not carry your cost, regions you cannot support. A disqualifier costs one filter. A bad-fit deal costs a quarter.
A working signal taxonomy
Signals split into three families. Fit signals say the account belongs in your market: size, stack, business model. Intent signals say something just changed: a funding round, a hiring spree for revenue roles, a tech migration, a pricing page visit. Timing composites combine both: a company that fits, raised within two quarters and is hiring its first SDR is not a lead, it is an appointment waiting to be proposed.
Give every signal a decay date. A funding round is a strong signal for about a quarter. A job post expires when the role is filled. A pricing page visit cools in days. Scores that never decay become seniority lists, not priority lists.
Build it in Clay
Sources come first: a Sales Navigator search for the fit envelope, registries and databases for firmographics, job boards for hiring signals, your website visitors if you run identification. Each source lands in a Clay table and merges on domain.
Enrichment runs as a waterfall ordered by cost: free and flat rate sources first, paid credits only for records the cheap tiers missed, and a validation gate at the end that no address skips. The same discipline as in the deliverability playbook: bounces are reputation damage you paid to acquire.
Scoring stays simple on purpose. Two or more live intent signals on a fit account is an A. One is a B. Fit without motion is a C and gets nurture, not sequences. Complexity in scoring formulas is where signal programs go to die.
The signals nobody else has
Everything above is available to your competitors, and by 2026 most of them use it. Durable advantage comes from signals that do not exist in any marketplace: a scraper watching the niche job board where your buyers hire, changelogs of complementary products, procurement registries, community posts in your vertical. These run as small automations in your orchestration layer and land in Clay like any other source.
One custom signal that fires twenty times a month with high precision beats another generic intent feed shared with everyone bidding on your keywords.
Keep the list alive
The engine runs continuously: new accounts qualify in, decayed signals age out, scores move and the CRM stays in sync with dedupe on domain so sales never sees two versions of one company. Tier A accounts trigger outreach within 48 hours, because the signal that made them Tier A is a clock, not a label.
Five ways signal programs die
- Buying a static list anyway
The old habit with new tools. If the input is a csv from a broker, the output is spray and pray with better formatting.
- Worshipping one signal
Funding alone means budget, not need. Hiring alone means growth, not your problem. Single signals produce plausible lists that do not reply.
- Enriching everything
Running paid enrichment on C tier accounts burns credits on contacts you will not touch this quarter. Enrich at the moment of need.
- Scoring theater
Weighted models with fifteen inputs nobody can explain. If sales cannot say why an account is an A in one sentence, the score is decoration.
- No decay
A signal list without expiry dates quietly turns back into a static list. The whole point was the clock.
What good looks like
| Metric | Target |
|---|---|
| Email coverage after waterfall | 85 to 90 percent of accounts |
| Validation pass before send | 100 percent, no exceptions |
| Share of outreach going to Tier A | above 60 percent |
| Time from signal to first touch | under 48 hours |
| Reply rate, single signal | 8 to 15 percent |
| Reply rate, stacked signals | 15 to 25 percent |
| List refresh | continuous, with signal decay applied |
These numbers assume the sending foundation from playbook 01 is in place. Great targeting through a burned domain is a great message nobody receives.