The method
No specific fact, no send.
Never contact anyone empty-handed. Show up already knowing something specific, true and recent about their problem — in their own words — and the conversation starts at “how did you know that?” instead of “who are you?”
The rule
A guideline is a suggestion.
This is a constraint.
Every other tool tells you to personalise and then ships a “send all” button anyway. We made it structural instead: a draft cannot exist in our database without a foreign key to a real, dated, sourced observation about that specific prospect. There is no setting that relaxes it, because it isn't a setting.
That one rule is what separates this from spam, and it's the reason the reply rates further down this page are reachable at all. It also protects the thing amateurs destroy first: your sending domain.
schema · drafts
-- the whole product, in one line
cited_signal_id uuidNOT NULL
REFERENCES signals(id)
-- and on every signal
evidence_url : "…/reports"
evidence_text: "script[src*=semrush]"
detected_at : 2026-07-22
The test
What counts as a specific fact
The blocked half of each pair is what everyone else sends. It's why cold email has the reputation it has.
Counts — cite it
“You're running SEMrush's widget on /reports”
detected, dated, URL
Doesn't count — blocked
“You're a marketing agency”
no source, no date, no send
Counts — cite it
“You posted an AIO Specialist role 9 days ago”
ATS link
Doesn't count — blocked
“Companies like yours care about AI search”
no source, no date, no send
Counts — cite it
“You left a 2★ on [competitor] in March”
review URL
Doesn't count — blocked
“You might be frustrated with your current tool”
no source, no date, no send
Counts — cite it
“Your last 4 case studies are all local SEO”
page URLs
Doesn't count — blocked
“I noticed your great work”
no source, no date, no send
The funnel
Ten stages, two of which are new
Every competitor sells FIND and ENRICH. Almost nobody sells “here is what this market is angry about, in their exact words, with sources”.
- 01
Find
Who could plausibly buy this?
Multi-source discovery across maps data, directories and competitor customer lists.
- 02
Enrich
How do I reach them, and is the address real?
Crawl for the contact, then verify it. Bounces above 5% are an enrichment problem, not a copy problem.
- 03
Signal
What is true about this specific company, right now?
The competitor script on their site, the role they posted nine days ago, the 2★ review they left in March. Every signal stores a URL, the exact text, and the date it was found.
- 04
Pain
What is this whole market angry about, verbatim?
Public reviews of your competitors plus the forums your buyers actually post in, clustered into themes and ranked by how often and how angrily they come up.
- 05
Approach
What do I say that only I could say?
One draft per prospect, welding their signal to the market's pain — then held in a queue until a human reads it.
- 06
Demonstrate
What real value proves it before they pay?
The pre-computed artifact. Something useful you can hand over without their permission or participation.
- 07
Trial
How do they try it?
A small, real run on their own product. Fifty sends tells you more than any survey.
- 08
Payment
How do they say yes?
The offer ladder, with the cheapest rung genuinely worth buying on its own.
- 09
Referral
Where does the next one come from?
One intro asked for from every customer. It is the single highest-yield minute in the week.
- 10
Learn
Which of this actually worked?
Attribution by segment × pain theme × template. Losers get retired automatically.
Signal and Pain are the two inserted stages, and they are the whole difference.
The rollout
Six phases, Monday to Monday
What a customer actually does, with the exit criteria that stop you moving on too early.
Onboard
Paste a product URL. You get back positioning, named competitors and three to five proposed segments. You edit them — the machine proposes, you decide, and your edits are sticky across re-runs.
Exit criteria
One segment approved and sharply defined. Not three. One.
Starting with three segments is the most common failure: three dossiers, three message sets, and no statistical signal on any of them.
Build the intelligence
Discovery, enrichment, signal detection and pain mining run in parallel. You do nothing while they do.
Exit criteria
Hundreds of verified emails, a live signal on at least 20% of them, and a dossier with real verbatim quotes.
Under 20% signal coverage means your trigger definitions are wrong, not your list. Generic themes — 'expensive', 'slow support' — mean the competitor set was wrong.
The first batch of 50
Fifty drafts. Not five hundred. You read all fifty yourself, with each draft's evidence beside it.
Exit criteria
Fifty approved and exported. Then wait — reply data takes three to five days to mean anything.
More than about ten rejects out of fifty means stop and fix the dossier. Approving a bad batch to get moving is how domains die.
Read the first signal
Fifty sends is too small to be statistically meaningful and absolutely meaningful in one way: did a real buyer reply with a real sentence?
Exit criteria
A decision — scale the angle, change the angle, or change the market.
Open rate is noise. Apple's Mail Privacy Protection made sure of that, and our dashboard omits it on purpose.
Scale what worked
Now volume is justified, because you know what to say. Two to four hundred a week, held to the same specificity rule. Attribution starts telling you which segment, theme and template convert.
Exit criteria
A paying customer from the first segment.
Do not add a second segment until the first one has produced one.
Compound
Every customer is asked for one intro. Every reply, positive or not, feeds the dossier. Every won deal tells you which signal type predicted it, so you can prune the rest.
Exit criteria
The cold-start problem stops existing.
Community channels open here — once you have named customers to reference, not before.
Phase 3, in detail
How to read fifty sends
One thoughtful reply from a real buyer tells you more than any open rate ever will.
| What you see | What it means | What to do |
|---|---|---|
| 3+ genuine replies | The angle works | Scale to 200/week, same angle |
| 1–2 replies, lukewarm | Right market, wrong angle | Change the pain theme, keep the list |
| 0 replies, no bounces | Wrong market or wrong pain | Back to Phase 0 with a new segment |
| Bounces over 5% | Enrichment quality problem | Fix verification before sending anything else |
Scroll for the full table
The arithmetic
Being specific is cheaper
Same product, same market. The only variable is whether the email says something true about the person reading it.
Generic outreach
12,000
sends · ~30 weeks at 400/wk · a domain probably burned
Signal-cited
1,700
sends · 4–5 weeks · reputation intact
Both columns end at roughly three paying customers. Here is every conversion step that gets you there.
| Stage | Generic | Signal-cited | Why |
|---|---|---|---|
| Delivered | 92% | 95% | Verified lists, warmed domain |
| Reply rate | 1–2% | 6–12% | The whole thesis |
| Positive share of replies | ~20% | ~35% | Relevance |
| Positive → call | ~50% | ~55% | — |
| Call → paid | ~25% | ~30% | Better qualified |
Scroll for the full table
Ranges, not promises. A bad market beats a good process every time.
The commitment
Three and a half hours a week
The promise isn't “automate outreach”. It's compressing a professional GTM motion into three and a half hours without it becoming spam.
- Mon30 minReview the week's new signals; approve the discovery queue
- Tue1 hrRead and approve the batch. The one hour that must not be skipped
- Wed–Thu20 min/dayHandle replies. Real humans answer real humans
- Fri30 minRead attribution. Retire one losing template. Ask one customer for one intro
- Monthly1 hrRefresh the dossier and check whether new themes emerged
The hour that isn't optional
Tuesday's batch approval is the only step we will not automate for you. Drafts arrive fifty at a time with their evidence beside them; you read, edit, approve or reject. There is no “send all” button and no setting that turns the gate off.
See the method on your own market.
Give us a product URL and we'll run the Pain stage on it — verbatim complaints from your competitors' reviews, ranked and sourced. No sends, no card, no ask.
Free · no card · yours to keep