The Numbers Person: Creative Ability for Demand Gen and Performance Marketers
Part of a series on creative ability by marketing persona. The overview is in every marketer is now expected to be creative. Earlier posts covered the writer who can’t design and the founder doing the marketing. This one is about the marketer who thinks in funnels.
What the demand gen marketer has
A demand gen or performance marketer is the person on the team who can tell you, from memory, what the demo request conversion rate was last quarter and what moved it. They read a funnel the way a writer reads a paragraph. They know what a channel costs, what it returns and when to stop feeding it. They set up the test, name the hypothesis, wait the right number of days and call it honestly. They are comfortable with a budget and a dashboard in a way most marketers are not.
They are also the person who connects marketing to revenue, which makes them the persona a CEO trusts most. When the board asks what marketing produced, the demand gen marketer has the slide. We wrote the benchmarks they live by in demand generation vs demand capture and the B2B SaaS funnel conversion benchmarks.
What the demand gen marketer lacks
Look at the tests.
Every test needs an input: a new message, a new offer, a new image, a new page. The demand gen marketer can set up the test in an hour. The creative for it takes a week, or two, because it comes from the writer, who has the launch, or the designer, who has a backlog, or an agency, which has a queue. So the test waits. Or, more often, the test runs on a variant the marketer made themselves in twenty minutes, which is last month’s email with a different subject line and the same stock image.
The copy, when the demand gen marketer writes it, is competent and sounds like every other B2B email. It is written to the metric rather than to the reader: the subject line that tests well, the CTA that converts, the five bullets that cover the features. It does not sound like the company, because the marketer is not the keeper of the voice and was never asked to be. We wrote about the drift in why AI content sounds generic, and the same thing happens to hand-written copy optimised purely for a number.
The images are stock, or they are AI images from a prompt typed in a hurry, and across a quarter’s tests they look like six companies. The visual inconsistency is invisible to the marketer, who sees the conversion rate, and visible to the buyer, who sees the brand.
And the landing page is the slowest part of all, because it is a build, in a page builder, by someone with access, and the test cannot start until it exists.
None of this is a failure of ability. The demand gen marketer is doing the hardest part, the measurement and the money, well. The creative is the input to their machine, and it comes from outside, slowly, and in whatever shape it comes.
Why it is worse than it looks
The cost of the creative bottleneck is not the week the test waits. It is the tests that never run.
A demand gen program learns at the rate it tests. If creative takes two weeks per variant, the program runs perhaps two tests a month and learns twenty-four things a year, most of them small, because big tests need big creative. If a variant takes a day, the program can test the offer, the audience, the message and the concept, in parallel, and the second half of the year runs on what the first half learned.
The budget environment sharpens this. Gartner’s 2026 CMO Spend Survey puts marketing budgets at 7.8 percent of revenue, flat, with 56 percent of CMOs saying they lack the budget for their strategy. A demand gen marketer who cannot test fast enough is spending the budget on last quarter’s best guess. The benchmarks for what that costs in paid channels are in Google Ads benchmarks for B2B SaaS and LinkedIn Ads benchmarks for B2B SaaS.
The fill: hypotheses in, variants out
The reframe for this persona is that they already write briefs. They call them hypotheses. “Buyers in operations roles respond to a time-saved claim more than a cost claim.” That is a creative brief with a metric attached. The line below turns it into finished pieces.
Station one, the hypotheses are the brief. The marketer briefs the campaign in chat with the campaign agent: the audience, the offer, the two or three claims to test, the channel, the date. The agent asks what a producer would ask and holds the answers alongside the business context captured once: who the company sells to, what it can claim, how it sounds, what it looks like.
Station two, the drafts, in the voice. Agents draft the email, the landing page, the social posts and the follow-up sequence, and the message variants the hypotheses call for, in the brand voice. The variants differ where the hypothesis says they should, in the claim or the hook, and nowhere else. The review agent reads every draft for language, brand and claims before the marketer does, which is the checklist the marketer would never have applied to their own twenty-minute draft. Why drafting in a captured voice beats prompting from nothing is in AI copy generation in a brand voice.
Station three, approve the copy. The marketer reads the variants, kills the one that would not survive the sales team, edits the line that overpromises, approves. One gate, before any image exists.
Station four, the images, in one system. From the approved copy, the image agent makes the visual for each variant inside the brand look, so three concepts for one campaign look like three concepts from one company rather than three companies. Logo overlaid. Sized per channel. The method for keeping a set consistent is in consistent AI images across a campaign. The marketer approves each image. Nothing is used that they did not approve.
Station five, send from the same place. The emails and the sequence send from the workspace. The landing page is live from the approved copy, in a template, without a build ticket. The posts go on the calendar. The follow-up automation is armed. And because revenue is traced to the exact message, the test reads out at the level of the variant, which is the level the hypothesis was written at. Where this fits in the demand gen workflow is in how demand gen teams use Marqeable.
The demand gen marketer’s creative ability is the hypothesis. Everything downstream of it, the copy, the variants, the images, the page, is production, and production is what agents draft and the marketer approves. The test starts the day the hypothesis is written.
What the demand gen marketer ships this week
A hypothetical B2B company, Northwind, testing whether operations buyers respond to a time-saved claim or a cost claim for a new feature.
| Piece | Before | After |
|---|---|---|
| Two message variants | Last month’s email with two subject lines | Two drafts from the brief, differing in the claim and the hook, in Northwind’s voice, claims checked |
| Landing page per variant | One page, one build ticket, two weeks | Two pages from the approved copy, in a template, live the day the copy is approved |
| Visual per variant | Stock photo of a person at a laptop | A concept illustration per claim, in the brand look, with the logo placed, which the sales team can also use in the deck |
| Social posts supporting the test | Skipped, no time | Two posts per variant on the calendar, same look as the pages |
| Follow-up | Set up separately, after the launch, if remembered | Armed as part of the campaign |
| Readout | Channel-level conversion rate | Revenue traced to each variant’s messages |
The test started on Tuesday instead of in three weeks, and the readout is at the level the hypothesis was written at.
The two images at the top of this post are this test’s visuals. Both were generated in the Marqeable image studio from each variant’s brief, in Marqeable’s own brand look as a stand-in for Northwind, with the logo placed automatically. They differ in the claim and nowhere else: same palette, same illustration style, same office. The cost image took a second attempt. The first came back as a generic person at a desk with no cost idea in it, so it went back once with a change request, which is what the review gate is for.
What the demand gen marketer should not do
Do not test the brand. Variants should differ in the claim, the offer, the hook or the audience. They should not differ in the color palette, the illustration style or the voice, because then the test is measuring brand consistency against brand inconsistency, and inconsistency always loses with the buyer even when it wins the click.
Do not skip the review gate because the metric will tell you. The metric tells you which variant converted. It does not tell you that the winning variant made a claim the company cannot substantiate. The review agent’s claim check, and the marketer’s approval, are the step between a good number and a problem. The rules for substantiating AI-written claims are in FTC substantiation for AI marketing copy.
Do not confuse volume with learning. Ten variants that differ trivially teach less than two that differ in the claim. The hypothesis is the thing. Write fewer, sharper ones.
Frequently asked questions
What is the creative bottleneck in demand gen?
Every test needs a new message and a new image, and the demand gen marketer can run the test faster than anyone can make the creative. So the test waits, the variant is a tweak to last month’s, and the image is stock. The channel math is fine. The input to it is the constraint.
Can a performance marketer write their own copy?
They can write a competent version, and it usually sounds like the last campaign because they are optimising for the metric rather than the reader. Drafting from a brief in a captured brand voice, with a review agent checking claims, gives them copy that sounds like the company and a checklist they would not have applied to their own draft.
How do you make several creative variants that still look like one brand?
Capture the brand look once and generate every variant inside it, from approved copy, one image at a time. Variants should differ in the concept, the claim or the hook, not in the color palette or the illustration style. Then a person approves each one before it is used.
How does Marqeable help a demand gen marketer?
The marketer briefs the campaign and the hypotheses once. Agents draft the email, landing page, posts and the message variants in the brand voice, the review agent checks the claims, and the image agent makes the visuals in the brand look after copy is approved. The marketer approves and sends from the same workspace, and revenue is traced to the exact message.
The bottom line
The demand gen marketer is the best measurer on the team and runs fewer tests than they could, because every test waits on creative from outside, and the creative they make themselves sounds like last month and looks like stock. The fix is to treat the hypothesis as the brief, draft the variants in the brand voice with the claims checked, make the images inside one visual system after the copy is approved, and send from the same place the revenue is traced.
That is the line Marqeable runs: brief in, every piece and every variant drafted in your voice and your look, one gate, sent and published from the same workspace, revenue traced to the exact message. A marketer who can write a hypothesis becomes a marketer who tests every week. That is the outcome the factory exists for: more leads generated, and more of them won as customers.
See each station of the factory: build campaigns from one brief, review and approve content, run follow-up with automations, answer visitors with website chat, and tie revenue to the exact message with attribution.
Marqeable runs your campaigns, answers every visitor, text, and email in seconds, and turns them into booked jobs and meetings - even at 9pm on a Saturday. We’re in private beta with a small early cohort. Get early access
