What a Marketing Automation Strategy Actually Decides
Ask what the automation is for and most teams answer with a channel: email nurture, lead scoring, onboarding drips. Those are outputs. A marketing automation strategy sits above them and answers four questions first.
- Whose attention is worth automating for? Not "everyone in the database". A defined set of companies and roles that can actually buy at your size.
- Which moments matter? The points in a buying journey where something useful can happen: an enquiry, a burst of research, a stalled deal, a new customer's first month.
- At each moment, does a machine act or does a person? And if a person, how fast, and who.
- What is it judged on? Agreed with sales before anything is built, not reverse-engineered from whatever the dashboard reports.
Everything after that is execution. If those four answers are written down and sales has signed them, almost any modern platform will do the job. If they are not, the most expensive platform on the market will automate a guess.
Start From the Accounts, Not the Database
The default unit in most automation tools is the contact. One email address, one score, one journey. That is a poor fit for how B2B buying works. A serious purchase is approved by a committee of six to ten people, and most of them never fill in a form. Automation built around individual contacts will happily nurture one of them for months while the other seven decide without you.
So the first job is the list. On our account-based marketing work, that means a finite target account list, tiered by what each account is worth, with the fit criteria written down and signed off by sales. The tier then sets how much of the work a machine should do:
| Tier | Accounts | Automation's job | A person's job |
|---|---|---|---|
| One-to-one | The handful your board can name | Alerts, reminders, record-keeping | Research, outreach, every conversation |
| One-to-few | Clusters sharing an industry, use case or trigger | Sequenced content per cluster, signal alerts | Personalising the opener, working warm accounts |
| One-to-many | Hundreds or thousands | Most of it: nurture, scoring, routing | Picking up accounts the signals surface |
That table is the opposite of how automation usually gets deployed. The biggest accounts get the least of it, because they deserve a human, and the long tail gets the most, because no team can work a thousand accounts by hand. Without the list, there is no way to tell the two apart.
If you sell small, self-serve deals where one person decides in a week, skip the tiers. Your list is a segment, and the rest of this still applies.
Map the Moments That Matter
Next, write down the moments. Not the workflows you could build, the things that actually happen to a buyer between first contact and renewal. For most B2B companies there are about seven, and each one needs a trigger, an automated action and a named human action.

| Moment | What triggers it | Automation does | A person does |
|---|---|---|---|
| New enquiry | Form, chat, ad lead form, call | Instant reply, record created, routed | First real conversation, inside an agreed window |
| Active research | Pricing page read twice, a contact returning in one week | Alert to the account owner with context | Decides whether and how to approach |
| Early interest | Content download, event sign-up | Stage-matched nurture | Nothing yet, unless the account is tier one |
| Stalled deal | No activity on an open opportunity past a set time | Flags it; sends material to the wider committee | Re-threads the deal through another contact |
| New customer | Deal marked won | Onboarding sequence, review request | Kick-off, first check-in |
| Expansion window | Usage, renewal date, a new division appearing | Alert, relevant case material | The expansion conversation |
| Gone quiet | No engagement for a set period | Release from active sequences | Nothing, until a new signal appears |
Two rows deserve more attention than they get. "Gone quiet" is the one teams forget: most flagged accounts will not become customers this quarter, and without a rule for releasing them the list grows until the same companies are contacted repeatedly by people who do not know it has happened before. And "new customer" is where the strategy usually stops, even though the first signature is the smallest revenue event in a good account.
Decide What Stays Human
This is the decision that separates a marketing automation strategy from a software project. For every moment in the map, someone has to say, in writing, which signals reach a person and which trigger something automatic, and what the response time is. On our intent and signal work that rule is written down rather than held as a judgement, because a signal that decays in two days and reaches somebody in five has been bought and thrown away.
Where a Marketing Automation Strategy Hands Off to a Person
A useful default: automate the parts the buyer does not want a human for, and hand over everything where they expect judgement.
- Automate acknowledgement, scheduling, reminders, record updates, routing, the release of cold contacts, and the first reply to an enquiry. Nobody wants to wait for a person to confirm a meeting time.
- Keep human anything involving price, scope, a tier-one account, a stalled deal, an unhappy customer or an expansion conversation. An automated email at those moments tells the buyer they are being processed.
- Automate the nudge, not the conversation. When a target account starts researching, the machine's job is to put the right context in front of the right rep, quickly. Account-level alerts routed to the right person with context attached beat a dashboard nobody opens.
The test for any automated step: would the buyer be glad it was instant, or annoyed it was not a person? If the second, it is a human step with an automated reminder behind it.
Five Programmes, in the Order to Build Them
With the decisions made, the build list is short. Build these in order, and do not start the next until the one before it is working. Each one needs entry and exit criteria: what puts a contact or account into it, and what takes them out. A programme that cannot be described that way should not run.
- Enquiry response. Every source, from forms and chat to the Google and Meta ad lead forms, lands in one CRM record and gets an instant first touch. On the CRM systems we build, a text and email go out within seconds of a form fill or an ad lead, and the lead is routed to a named owner. This is the cheapest revenue in the whole strategy, because these people have already raised their hand.
- Signal alerts. First-party behaviour, the pricing page read twice or a named contact returning several times in a week, raises an alert to the account owner. It costs nothing to collect, and most businesses capture none of it because nobody connected the site to the record.
- Stage-matched nurture. Only now the emails. One sequence per stage, not one sequence for everybody, with a clear exit when the contact moves stage, books a call, or goes quiet.
- Customer onboarding and expansion. A won deal starts its own programme: onboarding communication, a review request, and alerts when an expansion window opens. Most companies automate acquisition and leave this to whoever remembers to ask at renewal.
- Closed-loop reporting to the ad platforms. Push closed-won deals back to the ad platforms as offline conversions, so LinkedIn and Google optimise toward revenue rather than form fills.
Notice where nurture sits. Third. Most teams start there because it is the visible part, and spend months polishing emails to people who were never going to buy while the hottest enquiries wait a day for a reply. If you are fixing an existing setup rather than starting fresh, the common mistakes in marketing automation follow the same order.
Write Nurture for the Committee, Not the Contact
When you do get to content, write it for the roles in the buying group rather than for a generic persona. The economic buyer, the champion, the technical evaluator and the blocker are each asking a different question, and one sequence cannot answer all four.
- The champion needs material they can forward internally: a business case written so they do not have to rewrite it.
- The economic buyer needs the cost of doing nothing and what changes if it works, in a page.
- The technical evaluator needs specifics: how it integrates, what it replaces, the security and procurement answers.
- The blocker needs the objection answered before it is raised, usually a comparison against the incumbent or the status quo.
This is also where buyer personas earn their keep, if they describe roles in a decision rather than demographics. Tag each piece of content by role and by stage, and let the automation choose by both.
How to Judge a Marketing Automation Strategy
Automation platforms report what they can see: sends, opens, clicks, contacts moved between stages. A marketing automation strategy should be judged on what it was built to change, and those numbers move at different speeds.
- In weeks: median time from enquiry to first human contact, and the share of signals actioned inside the agreed window. If these do not improve, nothing downstream will.
- In a quarter: for account-based programmes, how many target accounts are engaged and how many members of each buying committee you have reached. For volume programmes, sales acceptance rate.
- Over the sales cycle: pipeline created and influenced, win rate by source, and expansion revenue from existing customers.
Be honest about timing. On a nine-month sales cycle, revenue in month two is not a signal, it is noise; coverage and response times move first and they are the fair early evidence. And state the attribution window on every report, so the numbers survive a question from the board.
A 90-Day Rollout
Do not launch the whole thing at once. A strategy switched on across every account on day one fails everywhere at once and teaches you nothing, because you cannot tell whether the list, the message or the follow-up was wrong. Start small enough to learn.
- Weeks 1 to 3: the decisions. The target list or segment, the moment map, the human-versus-machine rule with response times, and the measures. One document, signed by sales.
- Weeks 4 to 6: the foundation. Every source into one record, enquiry response live, signal alerts routed to named owners.
- Weeks 7 to 10: one pilot. One stage-matched nurture for one tier or segment, with entry and exit criteria and success measures agreed in advance.
- Weeks 11 to 13: the review. Response times against the target, what the pilot produced, and which steps to cut. Then scale what worked to the next tier.
Write down what you cut, too. A programme where nothing was ever stopped is a programme nobody was reading.


