Manage outreach as a portfolio of campaign cells, not a send queue
Split outbound into campaign cells by company size, industry and persona, count each eligible pool honestly, then decide where to build, fix or stop sending.
Jared Hawkins-Young · Founder, Lost Star TechnologyOctober 8, 20265 min readPart 8 of 14Key takeaways
- A campaign cell is one company size band, one industry and one persona, with its own eligible pool, message play and performance record.
- The eligible pool is the contacts in a cell who are verified, in scope, not suppressed and not already in a sequence. It is never the whole CRM.
- Comparing each cell's positive-reply rate with the portfolio median shows where to build the pool, where to fix the play and where to stop sending.
- Tight cells reply better: Hunter's 2026 report found sequences with 21 to 50 recipients drew a 6.2% reply rate against 2.4% for sequences of 500 or more.
Run outbound as a portfolio of campaign cells: each cell is one company size band, one industry and one persona, with its own eligible pool, message play and record of results. Then manage the cells the way an investor manages positions: add to the ones that earn replies, fix the ones that lag, and close the ones that do not pay back. A send queue cannot tell you any of that. It can only tell you how much went out.
Part 7 argued that the advantage moved from how much a team can send to how well it decides. This part is the first piece of the operating model that follows: the unit you decide about.
What is a campaign cell?
A cell is the smallest group of contacts you would expect to answer the same message for the same reason. For a B2B software company the useful splits are company size, industry and persona. Take a concrete case. You sell software that sales teams buy, so your personas are the head of sales, the founder and the RevOps lead. Size is headcount, because revenue is rarely public for private companies and headcount is. Industry is the market the account sells into, because a fintech's sales motion and an HR tech company's are different problems.
Northwind, a logistics software company of about 120 people, sits in one cell for its head of sales, Dana Reyes, and in two more for its founder and its RevOps lead. Each of those cells gets a different reason to write, a different message and its own result.
- Head of salescell
- Foundercell
- RevOps leadcell
- Head of salescell
- Foundercell
- RevOps leadcell
- Head of salescell
- Foundercell
- RevOps leadcell
A campaign cell: one size band, one industry, one persona, with its own eligible pool, signals, message components, cadence and performance record.
Where the eligible pool is deep, which cells reply, where relevance is weak, and which cells deserve more of your attention or none at all.
The tree also decides who at a company hears from you first. Hunter's State of Email Outreach 2026, built on 31 million emails its users sent in 2025, found that contacting one or two people per company drew a 5.1% reply rate against 3.5% for three or more. Start with the persona cell that owns the problem. Move to the next persona only when the first goes quiet.
How big is the eligible pool, really?
Most outbound plans start from the size of the database. That number is wrong for planning. The eligible pool is the contacts in a cell who pass four filters, counted per cell, before a single send is planned:
- Verified. The mailbox accepts mail and was checked recently. A bounce costs reputation for every cell that sends from the same domain.
- In scope. The account really is in the band, the industry and the persona, confirmed rather than read off a stale enrichment record.
- Not suppressed. No opt-out, no do-not-contact flag, no prior complaint. The FTC's CAN-SPAM guide requires opt-outs to be honored within 10 business days and makes no exception for business-to-business email.
- Available. Not a customer, not in an open deal, not in another active sequence, not contacted inside the cooling-off window you set.
The pool shrinks fast once those filters run. A cell with 300 eligible contacts can carry a careful play for a quarter. It cannot carry a weekly volume target, and pretending it can is how a good cell gets burned.
What does the portfolio view show?
Once every send carries its cell ID, results roll up by cell. The view a head of sales needs on Monday morning is a table: each row a size band and industry, its positive-reply rate, the gap to the portfolio median, the eligible pool, and the move. The numbers below are illustrative, not measured results.
| Size band | Industry | Positive reply | vs median | Eligible pool | Move |
|---|---|---|---|---|---|
| 51 to 200 | Logistics software | 2.4% | +0.7 pts | 1,180 · 14% | Build pool |
| 51 to 200 | Fintech | 1.3% | −0.4 pts | 3,940 · 46% | Improve the play |
| 201 to 500 | Dev tools | 2.7% | +1.0 pts | 2,100 · 25% | Build pool |
| 501 to 1,000 | HR tech | 0.9% | −0.8 pts | 1,300 · 15% | Stop for now |
Read the example the way you would read your own. Logistics software and dev tools beat the median, yet together they hold only 39% of the eligible pool. The next move is not more volume into fintech because it is the biggest cell. It is to verify and grow the two pools that already earn replies, and to protect the message and cadence working there. Fintech gets a message test. HR tech stops until something changes.
The cell is also what makes every later step in this series work. Timing is learned per cell in part 11, and when replies drop, the matched control in the part 12 runbook is the same cell in an earlier week.
When should you stop sending to a cell?
Stopping is the move teams skip, because a send queue rewards activity. Stop when a cell sits below the median after its data and delivery checks come back clean and a matched message test has also failed. At that point the problem is fit: that industry or band does not feel the pain you solve right now. Sending more only teaches mailbox providers that your domain sends mail nobody wants. Google's sender guidelines ask senders to keep the spam rate shown in Postmaster Tools below 0.3%, and that rate belongs to the domain, so one bad cell spends reputation every other cell needs.
Small cells are not a constraint to work around. The same Hunter report found sequences with 21 to 50 recipients drew a 6.2% reply rate against 2.4% for sequences of 500 or more. A tight cell forces a specific reason to write, and the reason is what earns the reply.
Who this is not for. If you sell to a few dozen named enterprise accounts, cells are overhead: an account plan per company does this job better. The portfolio model earns its keep when the market is thousands of accounts and one person has to decide where the next week of outreach goes.
Next in the series
Part 9, A signal earns the right to become a message, goes inside one cell: how public signals are scored for fit, freshness, source quality and relevance before anyone writes an email, and how the research travels to the rep when the buyer replies.
Common questions
- How many campaign cells should a small sales team run at once?
- As many as you can give a real eligible pool and a distinct reason to write. For a head of sales with a few reps, that usually means a small number of active cells with a few hundred verified contacts each, rather than dozens of thin ones. Start with the cells your won deals point to and add one at a time.
- What counts as a positive reply?
- A reply that shows interest or asks a real question: tell me more, send details, talk next month, or a pointer to the right person. Out-of-office notes, opt-outs and wrong-person replies are logged but not counted. Measure positive replies per contacted lead in the cell, so cells of different sizes compare fairly.
- Should campaign cells use revenue bands or headcount?
- Headcount, unless you hold reliable revenue data. Revenue for private software companies is rarely published, while headcount is visible on company pages and professional profiles and changes often enough to keep the bands current.
Sources
Written by Jared Hawkins-Young from the system he builds and runs. Diagrams are drawn from the same specs as the software. Published October 8, 2026. Illustrative numbers are labelled as such; every other figure carries its source.