Automation projects fail quietly. Nothing crashes. The platform is configured, the workflows exist, the invoice is paid — and six months later nobody is using it, the sequences are switched off, and the CRM has gone stale. Almost always the cause is the same: the agency implemented software instead of fixing a process.
The gap between a partner who understands that and one who does not is enormous, and it is essentially invisible in a pitch. Here is how to see it anyway.
Ask what they will do before they touch the software. The right answer involves mapping your current process, sitting with whoever handles enquiries, and understanding what happens today — including the messy bits. If the first deliverable is a configured platform rather than a process map, you are buying installation, not marketing automation and CRM implementation.
Ask who writes the emails. This is the single most revealing question. Automation is mostly content, and content is where the value is. If the answer is "we will need copy from you" with no further support, budget for a copywriter or expect the sequences to sit unfinished. If the answer is "we write it, in your voice, and you approve it", you are talking to someone who has done this before.
Ask what happens if you leave. Who owns the account, the data, the workflows, the domain? Can you export everything and continue without them? A good partner answers this comfortably. Hesitation here is the loudest warning you will get.
Ask them to describe a project that went wrong. Everyone has one. The useful information is not the failure, it is whether they can talk about it honestly and what they changed afterwards. An agency with only triumphs is either very new or not being straight with you.
Ask what they will not automate. There should be a list. Anything involving judgement, negotiation, or bad news should stay human. A partner who wants to automate every touchpoint has misunderstood the job.
Ask how they will measure success. "More leads" is not a measure. Response time, follow-up completion rate, conversion by stage, and pipeline value are. If nobody proposes numbers you will look at monthly, nobody will look at them.
Ask about platform independence. Are they recommending this platform because it fits your business or because it is the only one they build in, or because there is a partner commission? Both are acceptable answers if stated openly. Neither is acceptable if hidden.
Ask what the first ninety days look like, week by week. Vagueness here predicts vagueness later. You want to hear a sequence: discovery, process map, foundation build, content, testing, launch, review.
Ask who you will actually deal with. The person in the meeting is often not the person doing the work. Ask to meet whoever will be in your account.
Ask what happens after go-live. Automation is not a project that finishes. If there is no plan for review, iteration and reporting, expect the system to decay.
Ask for a reference you choose, not one they choose. Ask that reference one question: what has actually changed in your business since?
And ask about the small things — response times, how change requests are handled, what happens when something breaks on a Friday. The answers tell you what working together will feel like.
Now the red flags. A quote produced without anyone asking how you currently handle enquiries. Guaranteed lead numbers, which nobody can honestly promise. A proposal that lists features rather than outcomes. Reluctance to put ownership of accounts and data in writing. A contract with a long minimum term and no clear exit. And the biggest one: an agency that talks entirely about the platform and never about your customers.
Two more things worth saying plainly. First, cheap implementations are usually expensive, because the cost is not in the configuration, it is in the thinking and the writing — and that is precisely what gets cut to hit a low price. Second, size is not the signal people assume. A small team who have done thirty of these for businesses like yours will beat a large one where your account is the training ground.
Choose on process, content and honesty about ownership. Those three predict the outcome better than any credential.
If you are shortlisting, we are happy to be one of the names on the list — and equally happy to tell you if what you need is not what we do.