How to Evaluate a B2B Marketing Agency

The fastest way to evaluate a B2B marketing agency is to ask questions that cannot be answered with positioning. Nine are below. Before them, here is our own delivery data, published so you have something concrete to benchmark any agency against, including us.

I run an agency, so read all of this with that in mind.

What delivery numbers should an agency be able to show you?

Most agency proposals lead with client outcomes: pipeline lifted, traffic doubled, cost per lead halved. Those numbers are usually real and almost never fully attributable, because the agency did not control the product, the pricing, or the sales team during the period being measured.

Delivery numbers are different. They are what the agency actually produced, they belong to the agency, and they are checkable. Ours, across eight client accounts and eleven brands:

What we shipped	Figure	Context
Largest programmatic content engine	255 pages	Single GEO programme, US reputation-management SaaS
Enterprise site rebuild	42 routes	Loyalty platform, rebuilt on Astro
CMS records migrated in that rebuild	387 entries, 545 assets	Full Contentful export
Content pages under an automated publish gate	44	Construction procurement SaaS
Ad markets run from one generated creative set	3 (US, UK, AU)	Australian taxi payments account
Newsletter issues shipped through a QA gate	11	Issues 20 to 30, one account
Delivery cadence	Weekly	Every account, no exceptions

Ask any agency you are considering for the equivalent table. The ones who can produce it in a day are operators. The ones who send case study PDFs instead are selling you the outcome of someone else’s product.

Who will actually write your first draft?

This is the single biggest predictor of how the engagement goes.

Agencies sell with senior people and deliver with junior ones. That is not automatically bad, junior people supervised well produce good work, but you should know the shape of it before signing rather than in month two.

Ask for the name and the seniority of the person who writes your first piece, and ask to meet them on a call before you sign. If the answer is vague, the answer is a junior you have not met.

What did you stop doing for a client last year, and why?

Good agencies kill their own work. A channel underperforms, they say so, they cut it, they move the budget.

If nothing was ever stopped, either nothing was measured or nothing was ever wrong. Both are worse than an honest cut. The specificity of the answer tells you whether measurement actually informs their decisions or just fills a monthly report.

Can you show me something that did not work?

Not a case study. A failure, with the diagnosis attached.

We had a run of generated ad creative fail completely because we tried to composite a product screen onto a phone in post rather than generating it into the scene. Zero usable images. Regenerating the same shots in one pass produced three usable out of three. The lesson cost us about a week and changed how we brief every image now.

An agency that cannot describe a failure at that level of specificity has either not run enough campaigns or is not being straight with you.

Specificity is the thing that cannot be faked in a sales conversation, and it is the same thing that shows up in the work.

What would you need from us to succeed?

The right answer contains uncomfortable asks. Access to your customers. Two hours a month from a subject matter expert. A decision-maker who can approve inside a week.

An agency that says “just a kickoff call” is either exceptional or planning to work without you, and the second is far more common. Content produced without access to anyone who knows anything is why so much B2B writing is interchangeable.

When does the first thing go live?

Ask for a date, not a phase. The first deliverable a customer of yours could see.

Three weeks is a reasonable ceiling. Beyond that you are buying discovery, which is a legitimate purchase but a different one from execution, and it should be priced and named accordingly.

What specifically ships in month four?

Months one to three are always busy because setup work is visible. Month four is where the difference shows, once the audits and workshops are done and the question becomes whether a repeatable weekly cadence exists underneath.

We ship weekly on every account partly for this reason, and the full reasoning is here. On a monthly cycle, a wrong assumption takes up to eight weeks to surface, which is roughly twelve corrections a year. Weekly caps that gap at seven days and gives you about fifty.

How do you price, and what changes the number?

You want a structure and a reason, not a figure that matches whatever budget you named.

An agency that reprices instantly when you mention your budget has just told you the first number was arbitrary. That is worth knowing before you are inside a contract.

Can I speak with a client who left?

The willingness matters more than the reference itself. Almost every agency has churned accounts. The ones who will connect you to one are telling you something real about how their endings go, and endings are where you learn what a firm is actually like.

What do you think is wrong with our current marketing?

Ask it on the first call, before any paid audit.

You are testing three things at once: whether they looked before the call, whether they can be specific, and whether they will say something you might not enjoy hearing. Vague flattery here is disqualifying. So is a generic list that would apply to any company in your category.

If you are weighing this against building in-house instead, the same constraint test is in your first marketing hire should not be a generalist. If the scope is specifically content, what a content marketing agency should actually do covers what to ask for.

The pattern underneath all nine

Every question above is a test of specificity. Marketing agencies are easy to start and hard to run, so the field is full of firms that can produce a convincing deck and cannot produce a repeatable week.

Specificity is the thing that cannot be faked in a sales conversation, and it is the same thing that shows up in the work.