An agency whose best work is a dentist, a law firm and a furniture retailer is not automatically wrong for your software company. Plenty of the craft transfers. Writing well transfers. Knowing how a search result behaves transfers. What does not transfer is the part that takes the longest: understanding a product well enough to explain it to someone who has never seen it.
So the useful question in a first call is not whether they have software logos. It is whether they know how to learn a product they have never used. That shows up in seven places.
1. Who actually writes, and what do they read first?
Ask for the name of the person who will write, not the name of the person on the call. Then ask what the first three things they read would be.
A good answer names your own material before anything external: your documentation, your changelog, your support tickets, recorded sales calls. A weaker answer starts with competitor blogs and keyword tools, which is how you end up with a post that could be about any product in your category.
2. How would you handle our documentation and changelog?
For a lot of software companies the docs and the changelog are the highest-intent pages on the site and nobody treats them as marketing surfaces. An agency that has only worked on brochure sites will not think to ask about them.
A good answer treats them as search assets with their own indexing, linking and structure questions, and asks who owns them internally. A weaker answer proposes a blog and stops.
3. Have you worked a sales cycle longer than a week?
A local service business converts in a session. Your buyer reads for six weeks, loops in two colleagues, and disappears into a security review. The content, the measurement and the patience required are different.
A good answer talks about multiple readers with different questions, and about measuring something other than last-click. A weaker answer talks about conversion rate optimisation as if there were one conversion.
4. How do you write for a developer and for the person who signs?
These are two audiences with opposite tolerances. A developer wants specifics and will close the tab at the first unsupported claim. The economic buyer wants the cost of the problem, in their own numbers.
A good answer says plainly that these need separate pages, and can describe what changes between them. A weaker answer proposes one message for everyone, usually the softer one, which loses the developer without winning the buyer.
5. What would you need from our engineering team, and how much?
Anyone honest will need engineering time: rendering and indexing questions, analytics events, page templates, a staging environment. The answer you want is a small, specific number with what it is for.
A good answer is something like a couple of hours in week one to get read access and answer questions about how the site renders, then an hour or two a month. A weaker answer is that they will not need any, which usually means they intend to work around your site rather than on it.
6. What would you do in the first 30 days?
This is the question that most reliably separates a plan from a pitch. You want a sequence with an output at the end of it, not a discovery phase with no artefact.
A good answer ends in something you can read and disagree with: a written diagnosis, a positioning document, a prioritised list. A weaker answer is onboarding, audits and strategy sessions with nothing named that you get to keep. For what it is worth, the milestones I work to are published, which at least makes them arguable.
This is also the argument for running a marketing audit before anything else, whoever ends up doing the work. An agency that resists producing a written diagnosis early is usually protecting a plan it has already decided on.
7. What would make you tell us to stop spending on this?
Every honest marketer has a version of this. Paid acquisition before the positioning is settled. Content before anyone can tell what the product does. A launch before there is anything to launch.
A good answer comes quickly and costs them money. A weaker answer is that everything they sell is always a good idea.
Two things not to hold against them
Not knowing your product in the first call. They have had your website and twenty minutes. What matters is the quality of their questions, not the accuracy of their guesses.
Not having a case study in your exact category. A specialist in your category may have learned one pattern and be about to apply it to you whether it fits or not. Breadth is not automatically worse; incuriosity is.
The general questions still apply on top of these. The ten questions to ask any marketing partner cover account ownership, reporting, contract length and who does the work, and the seven above are the technical-fit layer that sits over them.
If you are running a shortlist this month, pick the three questions above that scare you most and ask all of them the same three. The spread in the answers will be wider than the spread in the proposals.