Where we are
You are an AE at Cadence, carrying eight named accounts into a new quarter. b1 gave you the frame: the deal loop (Target, Reach, Discover, Advance, Record), the trust line (the read, the promise, the price stay human), and the five levers of an email that earns a reply. Tonight sits one stage upstream of all of it. Nobody wants this session. Everyone wants b3, where you get to write. But the b1 simulator already showed you the ceiling: Jonas Weber capped at 1.8%, Anna Kovac at 1.2%, Marcus Hale at 0.0%, at every rung of the ladder, forever. Those are Target failures, and Reach cannot repair a Target failure.
An ICP that excludes people 6 min live
Most ICP documents are a description of everyone the company would happily take money from. That is not an ICP, that is a wish. A useful ICP is defined entirely by who you say no to, and you can test it in one move: read it out and see whether it excludes any account currently sitting in your pipeline. If it excludes nobody, it is doing nothing. Four exclusion tests do the work, and they run in this order because each one is cheaper than the one after it.
LiveThe four tests, and what each one actually asks4 min▶
Each test is a question with a yes or no answer. If you find yourself writing a paragraph to justify a row, the row has already failed.
| Test | The question | Where the answer comes from |
|---|---|---|
| 1 · Price floor | Can this company buy at your entry price without a special deal? | Headcount, sites, revenue band. One field, one glance |
| 2 · Owner of the pain | Is there a named person whose week is worse because of this problem? | Their org page, LinkedIn titles, job postings |
| 3 · Trigger in 12 months | Did something dated and public happen that makes now different? | Part 2 · the six-minute hunt |
| 4 · Path to a decision | Could someone say yes this year, or is this a two-budget-cycle build? | Funding, hiring, procurement signals, prior deals |
Notice what is not on the list: industry, geography, tech stack, and "shows intent". Those are filters, and filters make a list bigger by making you feel informed. Exclusion tests make a list smaller, which is the only direction that helps you.
The cheapest list to build is now the most expensive to send. Platform-average B2B cold-email reply rates sit around 3.4%, down from roughly 5.1% two years earlier, while top-quartile senders still land 15-25%. Volume went up and the average went down. The teams in the top quartile are not writing four times better than everyone else. They are sending to a quarter as many people.
Self-studyThe ICP floor is a business decision, not your opinion3 min read▶
Reps hate test 1 because it feels like turning away money. It is the opposite. An account under your price floor costs you the same six minutes of research, the same email, the same follow-up cadence, and produces either a no or a deal that loses money in support. Worse, it costs you sending reputation on a domain you need for the accounts that can actually buy.
- Anna Kovac, Ops Lead at Bright Harbour, 12 staff. She is a genuinely good fit for the product and a genuinely bad fit for the price. Her row in the b1 simulator is capped at 1.2% with every lever on, and a perfect email to her is still a no.
- Write the floor down as a number. Not "SMB is hard", a number: minimum sites, minimum ops headcount, minimum spend. If you cannot get that number from your own leadership, that is the escalation, not the workaround.
- Keep the under-floor rows somewhere. Kill them from this quarter's list, not from the world. Bright Harbour at 12 staff is a different company at 40.
Trigger hunting in six minutes 7 min live
A trigger is a dated, public thing that happened at their company. Not an inference, not intent data, not a hunch. Timeline-based hooks outperform problem-based hooks by roughly 2.3x in reply rate, which is the entire reason this is the one piece of per-prospect homework worth paying for. Pain and proof you write once per segment and reuse for a quarter. The trigger you earn one account at a time, on a clock, from seven sources that actually carry dates.
LiveLet AI find. Never let it write here.5 min▶
This is the cleanest split in the whole track, and b3 depends on you getting it right tonight. Finding dated facts across seven sources is exactly what a model is good at: it is fast, tireless, and the output is checkable because every claim carries a link. Writing the first line of a cold email is exactly what it is bad at, because the value of that sentence is that a person read something.
Company: Northwind Logistics, 3PL, Singapore, roughly 900 staff.
Find every dated, public event at this company in the last 12 months:
facility leases & openings, exec hires, job postings, funding, earnings
remarks, product launches, regulatory or permit filings.
Return a table only. Columns:
date (YYYY-MM-DD) | what happened, one sentence | source URL |
which of these three pains it plausibly creates
(1) decisions lost between meetings during a ramp
(2) new sites out of sync with HQ
(3) new managers with no shared operating record
Rules: no event without a date and a working link. If you cannot find
the source, leave the row out. Say "nothing found" if that is true.
Do not write an email. Do not suggest subject lines.
Research Northwind Logistics and write me a personalised outreach email to their VP of Operations. Make it relevant to what they are going through right now.
The bad prompt collapses finding and writing into one step, so you never see the evidence. The model produces a confident opening line about a "rapid expansion phase" that no source supports, you send it, and Mira Chen reads a sentence about her company that is not quite true. That is a trust-line breach in the first ten words, and it is unrecoverable.
Self-studyWhich sources carry dates, and which just feel like research3 min read▶
Ranked by how often six minutes there produces something you can open an email with:
- Leases, permits and facility openings - the strongest of all, because they are dated, public, expensive, and imply a deadline. Northwind's Jurong DC lease five weeks ago is this category.
- Job postings - three open warehouse-ops reqs is a sentence about what they are trying to build and what is currently short-staffed. Free, dated, and nobody else in your patch reads them.
- Earnings-call remarks - rare but decisive, because the pain is in the buyer's own words. Sarah Lim, COO of Kestrel Retail, said on a Q2 earnings call that cross-site decisions were taking too long. You will never write that sentence better than she did.
- Exec hires - a new person with something to prove in their first 90 days is the most reachable buyer in B2B.
- Product launches and company news - dated, but often marketing about marketing. Useful, thinner.
- LinkedIn posts - good for voice and for the pain vocabulary, weaker as a trigger, because a post is not a commitment.
- Intent data and "visited your pricing page" - not on the diagram on purpose. It is a signal about you, not a dated event at them, and it cannot open an email without sounding like surveillance.
The account brief, and the rows you kill cards · read in any order
Research that does not fit on one screen will not be read before a call, including by you. Everything below is about compression: what earns a place in the brief, how fast the data underneath it rots, who else you need before you need them, and the two rows on your own list that are already capped.
LiveThe one-screen account brief4 min▶
One screen. If it scrolls, it is a research document, not a brief. Six fields, in this order:
| Field | What goes in | Northwind example |
|---|---|---|
| The trigger | One dated event, with the date and the link | Jurong DC lease signed five weeks ago |
| The pain guess | Which segment pain the trigger plausibly creates | Decisions lost between meetings during a ramp |
| The seat | The person who owns that pain, and why you think so | Mira Chen, VP Ops, owns the ramp |
| Second and third | Two more names and what they will each care about | Priya Raman (CFO), Jonas Weber (IT manager) |
| The proof | One comparable customer, one number, one period | Pacific Freight, Tuas ramp |
| The ask | The one question you want answered in a reply | Is this already a problem in the ramp, or not yet? |
What never goes in: the company boilerplate paragraph, the funding history nobody asked about, a competitor teardown, a list of everything the AI found, and any sentence you would not say out loud to Mira. If a fact is in the brief, you must be willing to be quoted on it.
Your proof point for this segment is already written. Pacific Freight, where Daniel Ortiz runs supply chain, took action items closed within a week from 41% to 88% in one quarter on the Tuas ramp. One customer, one number, one period. That single line does the work of a case-study PDF, and unlike the PDF it survives being read on a phone in a warehouse corridor.
LiveEnrichment data and how fast it rots3 min▶
Enrichment vendors sell you a field and let you assume a date. Every field in a prospecting tool has a half-life, and the ones that decay fastest are the ones your email leans on hardest.
- Job title - decays fastest of all. People move, get promoted, or absorb a second remit. An email addressed to a title someone left eight months ago is a tell all by itself.
- Headcount and site count - the numbers your price-floor test depends on, and usually the stalest thing in the record. Check it against their own careers page before you kill or keep a row.
- Email address - pattern-guessed more often than verified. A bounce costs you domain reputation, which is a shared asset across your whole team.
- Tech stack and "intent" - inferred, not observed. Treat as a hint, never as a fact you state back to the buyer.
The working rule: AI-enriched fields decide which rows you look at. A human-verified fact decides what you write. Anything that appears in the actual email must have been read by you, on a page, with a date on it.
LiveMultithreading before you need it3 min▶
Single-threaded deals die when one person changes jobs, changes priorities, or goes quiet. The time to find your second and third contact is now, while it costs you two minutes and no political capital, not in month three when asking for an introduction signals that the deal is in trouble.
- Mira Chen, VP Ops - the seat that owns the pain. She cares whether the ramp lands on time.
- Priya Raman, CFO - cares what it costs and what it replaces. She will never be your champion, and she can end the deal in one sentence. Know her before b7.
- Jonas Weber, IT manager - cares about integration, access and recording policy. He is not the buyer, and he can absolutely be the blocker.
Put all three in the brief now with one line each on what they care about. You are not emailing two and three tonight. You are making sure that when Mira forwards your email, you already know who receives it and why they might object.
LiveThe wrong seat and the under-ICP row, named4 min▶
In b1 you found three capped rows in the simulator without being told which they were. Here they are by name, and here is the honest handling of each. These are not simulator artefacts. They are the three most common ways a real week of sending goes nowhere.
- Jonas Weber, IT manager at Northwind (capped at 1.8%) - the wrong seat. Right company, right trigger, wrong human. He does not feel the pain and he does not hold the budget. Best case he forwards your email, worst case he answers it with a security question and you have opened your account through the objection door. He belongs in the brief as contact three, not as the person you cold email first.
- Anna Kovac, Ops Lead at Bright Harbour, 12 staff (capped at 1.2%) - under the ICP floor. Good fit, wrong size. A 5/5 email to her is still a no, and it spends sending reputation you need for accounts that can buy.
- Marcus Hale, VP Ops at Continental Cold Chain (capped at 0.0%) - prior opt-out. This one is not a targeting judgment call, it is a rail. Permanent suppression, forever, regardless of how good the trigger is. b3 and leader session a5 cover the regimes properly.
The discipline is to kill these rows in writing, with the reason. Not to leave them in with a low score. A row you keep because deleting it feels like giving up will get sent to on a Friday when the number is short.
Self-studyKeep your own trigger library3 min read▶
Six minutes an account, eight accounts a week, is roughly forty hours of trigger hunting a year. Almost nobody keeps any of it, so every quarter starts from zero. A trigger library is one file with four columns: date, account, the event, the source link, plus a fifth column you fill in later: did it get a reply.
- It teaches you your own patterns. After a quarter you will know that lease filings out-reply exec hires in your patch, or the reverse, and that is worth more than any published benchmark because it is your market.
- It makes nurture real. The accounts you routed to nurture tonight get rechecked next quarter against a list you can actually search.
- It survives you. When you hand the patch over, the library is the handover. b8 makes the same argument about the CRM.
Benchmarks are directional. Vendor samples skew to the vendor's own users, so treat published numbers as the shape of the curve and let your own library tell you where you sit on it.
Build the eight-account list, live ★ 25 min · everyone builds
Open your own list beside this page. Six minutes per account, timed out loud, and every account leaves this session in one of exactly two states: a dated trigger with a source link, or routed to nurture. There is no third state, and "looks promising" is not a state.
Write your four exclusion tests as four yes-or-no questions. Read them out. If none of them excludes an account currently in your pipeline, rewrite them harder.
Run tests 1, 2 and 4 across your whole raw list first. They are fast and they are free. Count what is left before you look at a single trigger.
Pick your top eight by gut. Set a timer for six minutes on account one and run the finding prompt from Part 2 across the seven sources.
Found something dated with a link? Write the date and the URL into the brief. Found nothing when the timer goes? Route to nurture, note today's date, move on.
Fill the six-field brief for each surviving account: trigger, pain guess, seat, second and third contact, proof, ask. One screen each.
Now kill at least two rows out loud, with the reason written next to each: wrong seat, under the floor, no trigger, no path to a decision, or suppression.
Count what you have left. If it is fewer than eight, that is the honest number and it is a better week than 240 was. Bring it to b3 and write to it.
Try it yourself - this week ◐ 60 min total
- Build your own eight. Real accounts from your real patch, run through all four exclusion tests, with the test each rejected row failed written beside it.
- Find one dated trigger per surviving account, six minutes each, timed. Date and source link or it does not count. Anything that runs dry goes to nurture with today's date on it.
- Kill at least two rows and write one sentence each on why. If you cannot bring yourself to kill two, your exclusion tests are still descriptions rather than tests.
- Fill the six-field brief for each survivor, one screen each, including the second and third contact. b4 preps a call from exactly this brief.
- Start the trigger library file: date, account, event, source link, and a reply column you will fill in after b3.
- Answer the course question again: AI just did the reading for you across seven sources. Where did the saved time actually go?
Official sources covered
This session teaches the targeting half of the published cold-outreach evidence: what actually lifts reply rate before a word is written, and the hook types that carry it. Benchmarks are directional, since vendor samples skew to their own users. This page covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · You started with 240 companies from the tool and finished with 8. What makes the 8 the better list?
An ICP is defined by who you say no to. The value is not the eight rows, it is the 232 rejections you can justify - price floor, no owner of the pain, no dated trigger, no path to a decision. A list you never say no to is a directory, and it is the reason platform-average reply rates fell while everyone reported more activity.
2 · Six minutes are up on an account and you have found no dated, public trigger. What is the correct move?
Timeline hooks beat problem hooks by roughly 2.3x, so swapping in a problem hook is choosing the weaker email and pretending you did the work. The twenty extra minutes have no stopping point and produce the afternoons that vanish. Finding nothing is information about the account, not a failure of effort.
3 · Jonas Weber is the IT manager at Northwind. Anna Kovac runs a 12-person ops team. Why do both rows stay capped no matter how well you write?
Target is upstream of Reach. Jonas does not own the pain or the budget, and Bright Harbour cannot buy at your entry price, so a 5/5 email to either is still a no while costing you the same six minutes and the same domain reputation. The third capped row, Marcus Hale, is different again: a prior opt-out is a rail, not a judgment call.