Published 18 September 2026
There is a predictable moment in a company's life when someone buys an enterprise marketing automation platform, and a second predictable moment, roughly nine months later, when everyone agrees it is not being used properly.
The tooling is rarely the problem. The mismatch between tooling and process is. Here is what is worth running at each stage, and what to deliberately not buy yet.
Pre-seed and seed: one list, one source of truth
You need somewhere to put people, somewhere to see deals, and the ability to send an email that does not come from a personal inbox. That is genuinely it.
A lightweight CRM and an email tool are enough. The single highest-value thing at this stage is deciding what a lead, an opportunity and a customer mean, and writing it down. Every automation you build later inherits those definitions, and re-deriving them at Series A is far more expensive than agreeing them now.
Deliverability setup — SPF, DKIM and DMARC — belongs here too. It is an afternoon of work and it prevents a category of problem that is miserable to diagnose later.
Series A: the lifecycle becomes real
This is where flows earn their keep: onboarding sequences, trial nudges, re-engagement, and handover rules between marketing and sales that both sides actually agreed to.
It is also where lead scoring usually gets introduced too early. Scoring is only useful once you have enough closed-won and closed-lost history to know which behaviours predict anything. Before that it is a formula someone invented in a meeting, and it will route your best leads to the wrong place with great confidence.
Pick the platform you can staff. A capable tool your one marketer knows well beats a powerful one nobody can configure without a contractor.
Series B and beyond: the integration problem
By now the constraint is rarely the marketing tool. It is that product data, billing data and CRM data disagree, and every report requires someone to reconcile them by hand.
The fix is architectural: a warehouse as the source of truth, models that define the metrics once, and tools that read from it rather than each maintaining a private version of the customer. This is the point where the automation platform stops being the centre of the stack and becomes one consumer of it.
It is also the point where a migration may genuinely be warranted — but for integration and data-model reasons, not because a competitor's tool has a nicer campaign builder.
The mistakes that repeat at every stage
Buying ahead of process. An enterprise platform does not create lifecycle discipline; it makes the absence of it more expensive.
Automating a broken process faster. If the handover between marketing and sales is unclear, automation distributes the ambiguity at speed.
Building flows on unclean data. Deduplication and field discipline are unglamorous and they determine whether any of the rest works.
No owner. A stack with no named owner degrades within two quarters — fields drift, flows conflict, and nobody is sure which of the three welcome emails is live.
A reasonable default
Start with the smallest stack that covers capture, storage and sending. Write the lifecycle definitions down before you automate against them. Add scoring only when you have outcome history to calibrate it. Move to a warehouse-backed model when reconciliation starts costing real hours.
And whenever someone proposes a new tool, ask which existing tool it replaces. If the answer is none, you are adding an integration surface, not a capability.