Most lead-gen retainers sell activity: sends, sequences, "touches", a dashboard with a rising line. The invoice is real, the deliverable is a story. The version I run ships in a fixed order, and every step ends with something you can open: a file on your domain, a URL, a number you can argue with. Here is the order, and where the money usually gets lost.
1. An offer that can be judged
Not a positioning deck — a live page: who it is for, what it replaces, what it costs, what happens on click. If the offer can't be read by a stranger in 40 seconds, no amount of sending volume fixes it, it just multiplies the problem.
2. List and criteria — with the exclusion list
Who gets contacted, and who is explicitly excluded. The exclusion list is where wasted budget usually hides: purchased lists, role addresses that never reach a buyer, verticals where your legal team will say no, existing customers, competitors, aggregator domains.
Two rules I keep for the list itself:
- the company domain comes from the posting or record it was collected from — no guessing, no inferred domains;
- every domain gets a live check before it enters the send queue.
A list of 3,000 rows where 400 bounce is worse than 1,200 rows that deliver, because bounces are what gets a domain burned.
3. Sending that survives
This is the unglamorous part that decides whether anything else is measurable:
- separate sending domain, SPF/DKIM/DMARC all passing;
- warm-up before volume, not during;
- per-mailbox daily limits well under provider thresholds;
- one plain-text variant per test arm, no link soup.
Outbound does not start until that checklist passes. Starting earlier doesn't save time, it spends the domain.
4. A test with a kill rule, computed before the test
The required volume is derived from your baseline and the smallest difference you actually care about — at 80% power, alpha 0.05. A strict two-arm test at a 3% base rate looking for a +20% relative lift needs 13,914 contacts per arm. Written down before the first send, with the date you will read the result and the stop rule if the number doesn't cross.
Without the kill rule, "we're still optimising" is a permanent answer, and no result is ever falsifiable.
5. An inbound layer
Outbound that lands on nothing is a rented pipeline. So: pages indexed under the queries your buyers actually type, structured data that answer engines can read, a free tool worth linking to. This is the compounding half — it keeps producing after the sends stop.
6. Measurement and handover
Events wired to the funnel, index status per URL, the whole thing handed over running. If you can't take it to another operator and keep going, you didn't buy an engine, you rented one.
What it is not
- Not a bought contact database resold at a margin. The exclusion list matters more than the row count.
- Not unlimited anything. Scope is fixed, and so is the price.
- Not a contract you cannot leave — the monthly loop cancels with 14 days notice.
- Not a promise of leads. Promises are worth nothing; the volume maths in the test plan is worth something.
The one that's usually missing
Six artifacts, and the one clients almost never have when they arrive is #4 — the number that makes the test falsifiable. Everything else can be bought: lists, tooling, copy. The kill rule is the part that stops a channel from being "in progress" forever.
Reference prices are public: Sprint $900 fixed, Engine $1,900/month, Full build $2,900 one-time. If your own volume can't reach the floor for a readable test, the honest deliverable is diagnostics and a smaller promise — and you should hear that in the first call, not in month nine.
I build this for B2B teams, one client at a time, as an engagement. Scope and public pricing: https://axelfreeman.com/marketing-engineer.html?utm_source=devto&utm_medium=article&utm_campaign=dfy-outbound
The full shipping order, with the deliverability checklist and the FAQ:
https://axelfreeman.com/done-for-you-lead-generation.html
Top comments (0)