Heyo, Ilya here 👋

8 teams reached out this week. Different countries, different markets, and they opened with almost the same sentence: we're building our outbound strategy right now and we want to launch in a few weeks.

That is what August is. Nobody is generating pipeline. Everybody is planning the quarter that starts in September.

For the next four weeks, pretend I actually joined your team as your GTM lead. Not as a vendor, as the person who has to make the September number work with whatever you already have.

So this is not an issue about what we do. It is exactly what I would do in your seat, in the order I would do it, with the week each thing belongs in.

One thing before the map. Almost every team starts in the wrong place: with tools, or a list, or copy. That is decorating a house before the foundation has set. The order below exists because each step decides the size and shape of the next one, and one of them has a deadline you physically cannot move.

Five workstreams, four weeks. Notice that row two starts this week and everything else can be caught up. That asymmetry is the whole reason this issue is going out today rather than in three weeks.

Before any of it, the thing I would say in our first team meeting.

Roughly half of your outbound result is decided before anyone writes a word or buys a domain. That half is how well you understand your buyer. Not who they are, but what they are dealing with, and how badly.

The other half is data research, copywriting and deliverability. Those matter enormously and they do exactly one job: they put a well-made email in front of the right person. If it lands in the inbox, looks clean, and still does not feel relevant, the result is low. Every time.

Most teams spend August on the second half, because the second half is concrete. You can buy domains. You can install tools. You can watch a progress bar. The first half feels like homework, so it gets skipped, and then the whole quarter runs on a shallow idea of who is being emailed.

So we are doing the first half first.

Not "we want more pipeline". The actual number, worked backward into the volume that produces it.

Here is the chain nobody writes down: revenue goal, average contract value, deals, meetings, replies, contacts, emails, mailboxes, budget. Every stage leaks. The leakage is where plans quietly die, because the spreadsheet assumed none of it.

That example is a 100,000 quarter at a 19,000 average contract value, a 2.8 percent reply rate and a three-step sequence.

It needs about 45,000 emails to about 11,000 contacts, producing roughly 300 replies, 40 real meetings and five closed deals. And 45,000 emails cannot come out of a handful of mailboxes without burning them, so it needs around 138 mailboxes and a real monthly budget across infrastructure, tooling and people.

We built a free calculator that runs this properly, because we got tired of watching teams find the gap in month two.

→ Run yours: plan.growth.band

→ Five-minute walkthrough of the logic:

Most people meet that screen one of two ways. Either the plan is real and now it has a budget, or the goal was never going to happen with the resources attached to it. Both are useful in August. Only one of them is useful in November.

Do this first. Everything below is sized by its output.

Want a second pair of eyes on your numbers? Run the calculator, then book 30 minutes with me. I will tell you whether your assumptions hold and where the plan breaks first.

Here is the mistake I watched a team walk into this week, and it is the most expensive one on this page.

You pick a trigger. A new head of department, a funding round, a hiring spike. You build a list around it, write a campaign for it, send it to a few hundred people, and wait.

Nothing happens. And you cannot tell why.

Was the trigger wrong? The copy? The list? Did it go to spam? You have three hundred contacts and one flat result, and no way to separate those four explanations. You cannot run it bigger either, because only so many companies have that trigger firing in a given month.

That is what starting from the outside costs you. You guess at what matters, then test the guess at a volume too small to learn anything from.

So do the opposite. Go inside first.

You already own better market research than you could buy, and most teams have never opened it:

→ Every call recording with a lead or a client, at any stage, including the ones that died.
→ Email threads with buyers, especially the long ones where they explained themselves.
→ Whatever you once sent a champion to help them sell you internally. That material is not about your product, it is about how a decision gets made in their situation, which is exactly what a cold email has to speak to.
→ Old deals. Take a client who arrived two years ago and read backwards from the close.

Then put the questions to it that a list will never answer:

→ Why did they change at all? What broke, or what stopped being tolerable?
→ What were they using before, and what specifically was missing?
→ Who actually decided, and who had to be convinced?
→ How long did the decision take, and what stalled it?
→ In their words, not yours, what was the problem called?

This is the one job AI is genuinely excellent at. Feed the transcripts and threads in and ask for patterns rather than summaries. Recorded calls are the most valuable asset most B2B companies own and almost nobody mines them. Two years of them sitting in Gong or Fireflies is two years of market research you have already paid for.

Not a persona document. Three to six sub-segments, each defined by a genuinely different situation.

And here is the test that matters: if your segment can be described by a filter, it is not a segment yet. It is a market.

"Companies of 100 to 1000 people who outgrew their current system" sounds specific. It is not. A 120-person company and a 750-person company are different businesses with different problems.

The smaller one adds five people a quarter and absorbs that manually. The larger one replaces a hundred people a year, runs twenty interns at a time, and every one of them has to be onboarded, trained and compliance-checked without a human explaining it twice.

Those are not one segment with one message. They are two segments, and in that case the dividing line was something as ordinary as whether the company runs interns at all.

You will not find that line in Apollo. You will find it in the fourth call recording you review.

Only now do you build triggers. Once you know the real dividing lines, a trigger is just the detectable proxy for one of them, and you write the message in the language your buyers already used. That order is the whole difference. Trigger first is a guess. Trigger last is a translation.

Now you know how big the infrastructure has to be, so build it, and start the clock today.

Warm-up takes three weeks minimum before the first cold send. Then the sending ramp runs after it, roughly 30, 55, 75 and 100 percent of target volume, about a week per step.

✅ Start now and you send on September 1 and reach full volume near the end of the month.

❌ Start on September 1 like everyone else and you are sendable on the 22nd and at full volume in late October.

Three weeks of hesitation costs most of a quarter.

The setup rules that matter, briefly:

→ Never send cold from your main domain. Separate sending domains, redirected to the main site.
→ Maximum three mailboxes per domain, separate workspaces. Reputation is shared across a domain and bleeds across a workspace, so one damaged mailbox drags its neighbours down with it.
→ Measure the ESP split of your real audience first, then buy capacity in that ratio, plus about 20 percent kept warm and idle as a rotation pool.

Plain text, open tracking off, bounce under 1 percent. That is the whole of it. Infrastructure is a solved problem, which is exactly why it is the wrong thing to spend August thinking about.

And while it runs, do not trust the dashboard.

Book an independent placement test for the week before launch. Here is why that is not paranoia.

Warm-up tools work by having mailboxes email each other and mark the result as important. In cheap pools those other mailboxes belong to other cold senders, so you mark each other not spam and the dashboard reports that exchange. The number is real. It just describes a closed loop containing none of your prospects. Across hundreds of independent audits we have seen dashboards reporting 98 to 99 percent while real placement came back between 30 and 70 percent on the same domains, on the same day.

We test our own fleet every 36 hours across 32,000+ mailboxes and hold 98.9 percent, and we test that often precisely because the number does not stay still by itself.

Read the result per provider, never as one average. Google at 97 and Microsoft at 20 averages to a number that describes nothing and hides the entire problem.

Now you have three to six sub-segments and the language that goes with each one. This is where they become campaigns, and where most teams undo the work by collapsing them back into one list.

A filter tells you someone matches a demographic. It does not tell you they have the problem you solve, and it says nothing about whether they have it right now. Relevance is pain and timing, not attributes.

An angle is a sub-segment plus the signal that makes it detectable at scale. A decision maker newly in seat, a hiring spike, a stack change, a department scaling, a capability present or absent in house. The signal is not the insight. The signal is how you find the people your insight already describes.

The part most people miss is that the angle changes the message, not just the list.

A CTO who arrived two months ago is reassessing what they inherited and has no loyalty to the incumbent.

A CTO of four years suddenly hiring five engineers is not reassessing anything, they have a delivery problem. Same title, same company profile, two different emails.

Or take a software development agency selling offshore capacity.

If the prospect already has developers in Eastern Europe, most teams treat that as a filter and exclude them. It is not a filter, it is an angle that inverts.

They have already solved that problem, so you either say nothing or you sell something else. A prospect with an all-US in-house team is the one you can talk to about speed, cost, and integrating without a multi-year commitment.

Same fact about the market. One team uses it to shrink a list, the other uses it to write two campaigns.

Then tier what you have built. Accounts with a stack of signals firing at once earn multi-channel effort. Accounts with one strong signal get email plus LinkedIn. The large remainder with no extractable signal gets email at scale, and its job is to teach you something rather than to close.

That gap between knowing your market and translating it into outbound is where most in-house programmes stall. Companies have the market knowledge. Almost none have the translation, and nearly every team we talk to that has used an agency found the agency did not have it either.

Not sure your angles are real angles?

Send me two of your segments and the signals behind them. I will tell you which ones will change a message and which ones are just filters wearing a costume.

→ ilya.growth.band

In September the clock is running and every day spent rewriting is a day of lost sending. Copy is an August job.

The test we apply to every sequence, and the only one that matters: could this email be sent to a different company without changing a word? If yes, it is not ready. Rewrite it around the signal.

Three technical rules that are pure deliverability rather than style:

→ Spintax every line, not just the greeting. Pattern detection flags identical bodies, and greeting swaps stopped being enough a while ago.
→ Plain text. No HTML decoration, no images in a cold first touch.
→ Open tracking off. It costs more in placement than it returns in data, and open rates lie anyway.

If you want the toolchain we run, the 30 Claude Code skills we use internally are open source, and the copywriting chain is in there: copy generation tied to a specific signal, a stress-test skill that scores drafts and iterates until they pass, subject lines, spintax, follow-ups.

→ github.com/ilya-a11y/growthband-outbound-skills

The final piece is the one that separates a campaign from a system.

You do not run one big campaign and then argue about it. You run many small ones, each with a fixed shape, and you decide the verdict rules while nobody is emotionally attached to anything.

The unit is one segment, one signal combination, one angle, 100 to 500 contacts, six working days. Then two numbers decide: reply rate, and the share of those replies that are actually interested.

→ Under 1 percent reply, or under 17 percent interested: kill it. No debates, no one more week.
→ High reply, low interest: the segment is right and the angle is wrong. Keep the list, rewrite the message.
→ Low reply, high interest: the message works and delivery does not. Check placement before you touch the copy.
→ Over 30 percent interested: scale it. More volume, adjacent segments, multi-channel.

Across a vertical this looks like dozens of campaigns, of which a small minority produce most of the results. Nobody can tell you in advance which ones. The job of the system is to find them faster than your competitors find theirs, which is why the real goal of outbound is clarity rather than a lead count. Leads are what clarity pays out.

Print it, split it between whoever owns infrastructure and whoever owns the market, and work down it.

I have been your GTM lead for the length of one email. If you want the version where I look at your actual situation instead of a hypothetical one, that part is real.

I am keeping August open for strategy reviews, because it is the only month where changing the plan is cheap.

Book 30 minutes and bring whatever you have: the calculator output, your segment list, your current sequences, your infrastructure setup, or just the goal and the confusion. I will tell you what I would cut, what I would resequence, and what will break first.

No pitch unless you ask for one. If your plan is sound you should leave knowing that, which is worth something on its own in August.

New issue every Monday, 3pm CET.

Ilya

growth.band