TL;DR
- The research object is the unit of work. A structured record per account carrying the trigger, the committee, the evidence, and the source for every claim is what makes everything downstream possible.
- Scoring against a written profile is the second highest-return application, and it only works if the profile uses observable attributes that a person could verify in two minutes.
- Draft sequences, never send them. The failure mode is confident specificity that happens to be wrong, and a fabricated detail in an outbound email is a credibility loss you cannot recover.
- Campaign preparation compresses from most of a day to well under an hour. That is the honest, defensible claim, and it is enough to justify the work.
- Every claim in a research object carries a source field. No source, no use in a message. This single rule prevents the majority of the risk.
Contents
- What actually takes the time in outbound
- The research object and why it comes first
- Scoring accounts against a written profile
- Drafting sequences from the research object
- The verification layer that keeps it safe
- What this does not fix
- FAQ: Claude Code for Go-To-Market
What actually takes the time in outbound
Ask anyone running a signal-based outbound programme where their hours go and the answer is rarely writing.
Finding out enough about an account to say something specific. Reading the job posting, checking the stack, finding the recent announcement, working out who actually owns the problem. Twenty to forty minutes per account done properly, which is why almost nobody does it properly.
Deciding whether the account is worth the effort. A judgement made against a profile that usually exists in someone's head rather than on paper.
Mapping the committee. Working out which three or four roles need to hear from you and finding them.
Writing. The smallest share, and the part everyone assumes is the bottleneck.
The first and third are context assembly, which is mechanical. The second is judgement applied against criteria, which is semi-mechanical if the criteria are written down. The fourth needs a person in the loop for anything that gets sent.
| Activity | Time share | Automatable | Risk if wrong |
|---|---|---|---|
| Account research | Highest | Yes | Low, caught in review |
| Committee mapping | High | Yes | Low, caught in review |
| Qualification decision | Moderate | Partly | Moderate, wasted effort |
| Writing the message | Lowest | Draft only | High, reaches prospects |
This is the reason the honest claim about campaign preparation compressing from most of a day to well under an hour is both true and unspectacular. It is not that writing got faster. It is that the two thirds of the work that was never writing got faster.
The research object and why it comes first
Build this before anything else. Everything downstream either works from it or does not work.
A research object is a structured record per account with a fixed schema. Fixed matters: a consistent shape is what lets you score it, template from it, and audit it.
A workable schema:
Trigger. The observable event that put this account on the list, with a date. Hiring into the relevant function, a stack change, a negative review in your category, a launch, an executive appointment, a posted project brief.
Trigger source. The exact URL. Not a description of where it came from.
Trigger date and decay status. When it happened and whether it is still inside the useful window. Launch signals last 24 to 72 hours, project briefs 7 to 21 days, hiring and stack changes 30 to 90 days, executive appointments 60 to 180 days, funding three to nine months as a qualifier only.
Firmographics. Size, sector, geography, and whatever else your profile requires, each with a source.
Committee. Three to five roles with names, titles, and the reason each one matters to this purchase. Not one contact.
Evidence. Two or three specific, verifiable observations about this company that could appear in a message. Each with a source.
Disqualifiers. Anything found that argues against pursuing this account. This field is the most valuable one and the one most often omitted, because a system that never says no is a system that scores everything as a fit.
The rule that makes the whole thing safe: every claim carries a source field, and any claim without a source cannot appear in a message. Enforce it in the schema rather than in a guideline.
Scoring accounts against a written profile
Scoring only works if the profile is written in observable attributes. If a criterion cannot be verified from public information in about two minutes, it cannot be scored, and including it produces a number that looks precise and means nothing.
Write the profile as a checklist with weights. Firmographic band, trigger type present, committee reachable, disqualifiers absent. Keep it short enough that a person could apply it manually, because that is the only way you can check whether the automated version is behaving.
Score in bands, not to two decimal places. Strong fit, possible fit, no fit. False precision invites people to trust the number more than it deserves, and the underlying data does not support finer resolution.
Require the reason alongside the score. Every score should come with the two or three factors that drove it. A score without a reason cannot be audited and will drift without anyone noticing.
Calibrate against known outcomes. Run the scoring over accounts you already won and accounts you already lost. If your closed-won cohort does not score highly, the profile is wrong, and no amount of tooling fixes a wrong profile.
Re-audit quarterly. Compare a sample of scored accounts against a human judgement on the same accounts. Divergence indicates that either the criteria drifted or your market did.
| Scoring practice | Why |
|---|---|
| Observable criteria only | Unverifiable criteria produce fake precision |
| Bands, not decimals | Data does not support fine resolution |
| Reason recorded with score | Enables audit, prevents drift |
| Calibrate on closed-won | Tests the profile, not the tool |
| Quarterly human sample | Catches drift in criteria or market |
Drafting sequences from the research object
With a populated research object, drafting becomes assembly against a structure rather than invention.
Route by trigger type, not by segment. An account that entered on a hiring signal needs a different opening than one that entered on a negative review, and applying a single sequence across every trigger discards the advantage the research just bought. This is the most common way well-built research programmes produce mediocre outbound.
Draft the structure, not the finished message. Four steps, varying angle, with the trigger reference in step one and the offer no earlier than step three. Keep each message short: on email, short enough to read on a phone without scrolling; on LinkedIn, under 150 characters lifts replies by roughly 22%.
Generate variants for the committee roles. The same trigger means something different to the person who owns the budget and the person who will use the product. This is variant work with a clear specification, which is exactly the kind of task that compresses well.
Never let a draft send itself. A person reads every message that reaches a prospect. Not because the prose is poor, but because a fabricated specific reads exactly like a real one and the recipient will assume you checked.
Benchmarks to hold the output against: cold email reply rates average 3.4 to 5.8% with top performers at 10 to 18%, LinkedIn acceptance runs 40 to 55% with a personalised note, LinkedIn reply rates 18 to 27%. If drafted sequences underperform these materially, the problem is almost always the list rather than the copy.
The verification layer that keeps it safe
Four checks, all of which should run automatically before anything is queued for human review. This layer is not optional and it is what separates a working deployment from an eventual incident.
Source presence. Every factual claim in every drafted message traces to a source field in the research object. A claim with no source is stripped before review, not flagged for review, because flagged items get approved under time pressure.
Decay check. The trigger is still inside its useful window. Referencing a launch three weeks late is worse than sending nothing, because it advertises that you are running a delayed automation.
Prohibited content scan. Superlatives you do not want to make, competitor claims you cannot substantiate, and anything on your written exclusion list.
Deliverability and compliance check. Sending domain authenticated with SPF, DKIM, and DMARC, one-click unsubscribe present per RFC 8058 and honoured within two days, volume within the mailbox cap, bounce rate under 3%, and complaint rate below 0.10% and nowhere near 0.30%.
| Check | Runs on | Action on failure |
|---|---|---|
| Source presence | Every claim | Strip the claim |
| Decay window | Every record | Remove from sequence |
| Prohibited content | Every draft | Block and flag |
| Deliverability | Every send batch | Halt the batch |
One process rule worth adopting alongside these: read operations run freely, write operations require an approved plan. Research, scoring, and drafting are reads. Sending is a write.
What this does not fix
Worth stating plainly, because the gap between expectation and outcome is where most disappointment originates.
A weak list stays weak. Research depth applied to the wrong accounts produces well-informed messages to people who will never buy. Targeting is upstream of everything here and no amount of preparation compensates for it.
A weak offer stays weak. If the thing you sell does not obviously solve a problem the recipient has, a better-researched email surfaces that faster rather than fixing it. High reply rate with low positive reply rate is the signature of this problem.
Positioning cannot be generated. Models produce the centre of the distribution of what already exists, and positioning is the deliberate act of not being there. This one is worth defending firmly.
Volume is not the constraint you think it is. Producing three times the drafts does not produce three times the pipeline, and it usually produces a review bottleneck plus a temptation to skip review. Measure time from list to approved sequence, not drafts generated.
Relationships are unaffected. Every conversation, every negotiation, and every recovery from a bad month is human work, and prospects notice immediately when correspondence stops sounding like a person.
Signal-Based Outbound Playbook: Download Free
Get the practical framework for applying this article to your GTM system.
Download the Outbound Playbook →
FAQ: Claude Code for Go-To-Market
What is a research object and why does it come first?
It is a structured record per account with a fixed schema carrying the trigger and its date, the exact source URL, the decay status, firmographics, three to five committee roles with the reason each matters, two or three verifiable observations that could appear in a message, and any disqualifiers found. The fixed shape is what makes scoring, templating, and auditing possible downstream. The rule that makes it safe is that every claim carries a source field and any claim without one cannot appear in a message.
How much time does this actually save?
Campaign preparation compresses from most of a day to well under an hour, which is a real gain and also the honest ceiling. The saving does not come from faster writing, since writing is the smallest share of outbound hours. It comes from account research and committee mapping, which together consume the majority of the time and are pure context assembly. Measure time from list to approved sequence rather than drafts produced, since draft volume usually indicates a review bottleneck forming rather than progress.
Can it score accounts against our ideal customer profile?
Yes, but only if the profile is written in observable attributes that a person could verify from public information in about two minutes. Score in bands rather than to decimal places, since the underlying data does not support finer resolution and false precision invites misplaced trust. Require the two or three driving factors alongside every score so it can be audited, and calibrate by running the scoring over accounts you already won and lost. If your closed-won cohort does not score highly, the profile is wrong.
Should drafted sequences send automatically?
No. A person reads every message that reaches a prospect, not because the prose is poor but because the failure mode is confident specificity: a fabricated detail reads exactly like a real one and the recipient will assume you checked. Adopt the convention that read operations run freely while write operations require an approved plan, which makes research, scoring, and drafting free while keeping sending gated. Applying an approved batch takes minutes; the preparation was the expensive part.
What verification checks should run before review?
Four, all automatic. Source presence, stripping any claim that lacks a source rather than flagging it, since flagged items get approved under time pressure. Decay checking, removing records whose trigger has passed its useful window, because referencing a stale event advertises delayed automation. A prohibited content scan for superlatives and unsubstantiated competitor claims. And a deliverability check confirming SPF, DKIM, and DMARC authentication, one-click unsubscribe per RFC 8058, bounce under 3%, and complaint rate below 0.10%.
What problems does this not solve?
A weak list stays weak, since research depth applied to the wrong accounts produces well-informed messages to people who will never buy. A weak offer stays weak, and shows up as a high reply rate with a low positive reply rate. Positioning cannot be generated, because models produce the centre of the distribution of what already exists and positioning is the deliberate act of not being there. Relationships are entirely unaffected, and prospects notice immediately when correspondence stops sounding like a person.
Yananai A. Chiwuta
Author
Yananai Chiwuta is a Go-To-Market Architect and founder of Forma Nôrden. He builds Go-To-Market growth systems for B2B companies selling into enterprise and mid-market accounts.
Celine Sky
Reviewer
Celine ensures technical accuracy and strategic alignment across all Forma Nôrden playbooks and resources.
