What a GTM Engineering Engagement Delivers, Week by Week

Why a week-by-week view matters

Most agency pages describe their work in 30, 60 and 90 day blocks. That is too coarse to hold anyone to. If you are about to hand your outbound to an outside team, you should know what exists at the end of each week, and what you are being asked to sign off.

Below is the sequence we run. Your timeline can move a week either way depending on how fast approvals come back and how clean your CRM is. The order should not move. Sending before the list is verified, or before inboxes are warm, is how programs burn their domains in the first month.

The 12 weeks at a glance

Each stage ends with a gate. Nothing moves to the next stage until you approve it.

Week 1: the plan comes before any tool

The first week is spent on decisions, not setup. Who exactly are we contacting? Which job titles, which company sizes, which regions? What is the one offer we lead with? What counts as success at week 12: meetings booked, qualified replies, pipeline created?

You approve a short written plan. If the plan names more than two or three personas, push back. Each persona needs its own list, its own copy and its own reporting line, and splitting a small program too many ways leaves every segment too thin to learn from.

Weeks 1 to 4: domains, warm-up and the verified list

Sending domains are bought and set up in week one, because warm-up takes time and cannot be rushed. Each domain gets SPF, DKIM and DMARC records and a few inboxes. The inboxes then send and receive a slowly rising volume of normal mail so providers learn to trust them.

While inboxes warm, the list gets built. Accounts come from a data source, contacts are enriched, and every address goes through verification. Risky and invalid addresses are held back. You see a sample early so you can say "these are not our buyers" before the whole list is built around a wrong assumption.

For a B2B SaaS client in fleet safety, this stage produced five persona campaigns to 9,962 verified contacts across 35 inboxes. The inbox count came from the volume, not the other way round.

Week 5: copy approval

AI researches each account and drafts the sequences. A person edits them. Then you read every email in every sequence before anything is loaded. This gate is not a formality. You know which claims your company can make, which customers can be named, and which words your buyers hate.

Expect one or two rounds of changes. After that, the rule stays the same for the life of the program: a human approves every send.

Weeks 6 to 7: first sends and reply triage

First sends go out at low volume and rise as the numbers stay healthy. Hard bounces and unsubscribes are suppressed between every send, so the next step of the sequence never reaches them.

Reply triage goes live on the first day of sending, not later. AI reads and categorises every reply: interested, not now, referral, out of office, unsubscribe, bounce or question. Ambiguous replies go to a human review queue. Interested replies go to your team the same working day.

Suppression is where early results show. For the fleet safety client, hard bounces fell from 1,637 on the first send to 75 on the third send of the same campaign wave, and unsubscribes fell from 76 to 8.

Weeks 8 to 9: CRM sync and tracking

Replies and their categories sync to your CRM, so your sales team works from one place. You approve the field mapping and who owns each type of reply.

This is also when we check that site tracking counts what it should. For the fleet safety client, rebuilding GA4 lead tracking surfaced 42 form fills that had never been counted. Outbound often drives people to the website rather than to a reply, and if the form is not tracked, that pipeline is invisible.

Weeks 10 to 12: reporting and iteration

By now there is enough data to make decisions. The weekly report joins email, site and CRM data and answers a small set of questions:

Iteration is deliberately narrow. Change one thing at a time, such as a persona's list, a subject line or the first email's offer, so you can tell what caused the result. Volume only rises when bounce and complaint numbers stay low. If a segment produces nothing after a fair test, it is cut and the sending capacity moves to the segment that is working.

• Which persona and segment produced the most positive replies?
• Which sequence step drew the replies, and which step only drew unsubscribes?
• How many replies became meetings and pipeline?
• Are bounce and complaint numbers still low enough to add volume?

What you should own at the end

At week 12 you should be able to see and name every part of the system: the domains, the inboxes, the list and its source, the suppression file, the sequences, the triage rules and the reports. If an agency cannot hand you that inventory, it has built something you are renting, not something you understand. Ask for it in writing before you sign.

The same goes for accounts. The domains, inboxes and CRM should be registered to your company, with the outside team given access, not the other way round. That way, if you part ways, nothing has to be rebuilt.