Creative Intelligence: How to Compare Creative, Messaging, and Campaigns Before Buying Media
In brief: Before signing off on a campaign, compare creative directions using consumer evidence. Creative intelligence brings that evidence into creative decisions before media spend. Synthetic persona responses are simulated and cannot predict campaign outcomes. With Nexus: Galaxies Lab, teams can compare alternatives and understand why each audience might favor or reject them.
A campaign can reach the approval meeting with strong creative and one question still unanswered. Why does this option deserve the budget? When the discussion revolves around personal taste, the team hears plenty of arguments for the work and learns little about how the audience understands it.
In 2017, Nielsen and Nielsen Catalina Solutions analyzed about 500 consumer packaged goods campaigns. Creative accounted for 47% of advertising's contribution to sales, more than reach and targeting combined. Source: Nielsen. Yet choosing between two ads often still comes down to taste.
Creative intelligence helps teams ask questions that can change a decision, gather evidence, and give creatives something specific to act on. A recurring objection might call for a better demonstration of the benefit. A message with several interpretations might need a more precise headline. The output is a decision the team can explain and a plan for what to test next.
What is creative intelligence in marketing?
Creative intelligence can mean analyzing ads already in market, identifying performance patterns, or using data to guide creative development. Here it refers to decisions before media spend, while the team can still revise the alternatives.
Using AI to generate twenty headlines gives the team more options. Assessing what each headline communicates, who it speaks to, and where it creates friction provides evidence for choosing among them. Both activities can work together when the team knows which question it is answering.
Look at the execution and the audience's interpretation. A video may introduce the product and brand early. That tells you how it was made. You still need to find out whether the promise feels relevant to the intended audience.
What decision comes before the test?
Start with a choice the team can still change. “See whether the campaign is good” leaves too much open. “Choose which benefit should lead the campaign for consumers unfamiliar with the brand” identifies the audience, the material, and the decision.
Specify what you will do with likely findings. If the message is clear but hard to believe, add proof. If it resonates only with current customers, consider whether it serves the acquisition goal. If two options look equally promising, both can move on to research with consumers or an in-market experiment.
Set your criteria before preferences emerge. A team with a favorite can easily look for excerpts that support its choice. Writing down hypotheses and criteria for advancing reduces selective interpretation.
How do you build a creative decision matrix?
Use this planning matrix to connect the material, the question, and the decision. It is an example for structuring a round of research. Nexus does not use it as a certified scale or benchmark.
Material · Business question · Possible decision
Campaign concept · Which idea expresses the benefit most clearly to this audience? · Prioritize a concept and revise how it demonstrates the benefit
Core message · What does the audience think the brand is promising? · Choose the argument and clarify the promise
Image and headline · Does the image reinforce the copy? · Adjust the pairing or test the elements separately
Video · Does the sequence explain the offer and make the brand recognizable? · Revise the script or the order in which information appears
Complete ad · What questions or objections could stop someone from taking the next step? · Select finalists for an in-market test
Keep a short audience description alongside the matrix. Category familiarity, purchase situation, and relationship to the brand often explain more about a response than demographics alone. A convenience claim may appeal to people who know the product and sound vague to someone unfamiliar with what it does.
Our guide to AI market segmentation goes deeper. Avoid treating an overall average as a recommendation for every segment.
How do you keep variables from getting tangled?
If two ads change the image, promise, price, and CTA at the same time, the comparison can help choose an execution as a whole. It cannot isolate what drove the difference. Keep other conditions comparable when the goal is to learn about one element.
The creative testing guide covers hypotheses, alternatives, questions, and a checklist for running a round.
Google's ABCD framework focuses on Attention, Branding, Connection, and Direction. Kantar analyzed 11,000 ads and found that videos aligned with the principles had, on average, a 30% higher modeled likelihood of driving short-term sales. Source: Kantar.
How does Nexus support the research?
Nexus: Galaxies Lab is Galaxies' platform for investigating choices and supporting decisions before spend. Persona Chat explores motivations, Questionnaire compares responses to the same questions, and Stimulus Analysis compares up to 10 assets with up to 10 personas in one study.
At Galaxies, we call this work Marketing Decision Intelligence. We help marketing leaders make decisions about creative, claims, audiences, and messaging before committing budget, using synthetic personas built from real research and human review.
Persona Chat lets teams follow up on objections. Questionnaire shows how people might respond to alternatives under the same questions. Stimulus Analysis puts the actual creative in front of the personas. The Nexus Agent can guide teams toward the format that fits their question.
Give the creative team feedback they can act on. “The message didn't work” leaves too much to guess. Explain that the offer was read as a subscription when the product is a one-time purchase, and identify the elements that may have contributed. The team can revise the ad and check whether the confusion persists.
How should you read the results?
Persona responses are simulated. They support analysis alongside research with people and cannot predict campaign results. Galaxies measures agreement against a reference group of real respondents in the same segment, using questions held out when building the profiles. Conversations are also evaluated continuously with the open-source DeepEval framework across five dimensions: relevance, bias, hallucination, profile fidelity, and toxicity.
How do you interpret preferences and objections?
An option can attract interest because it promises something the product cannot deliver. Another may seem less compelling at first while communicating a benefit that matters to a priority audience. Read rankings alongside comprehension, credibility, and fit with the objective.
Pay attention to disagreement. If one segment values convenience and another wants proof of quality, the team may need message variants. If both misunderstand the offer, fix the underlying explanation before adapting it by segment.
Base the recommendation on preference, comprehension, and credibility. When interest depends on a feature the product lacks, revise the promise before selecting the ad.
How do you connect pretesting to media?
Pretesting helps clear up ambiguity, identify barriers, and select candidates. An in-market experiment measures behavior under delivery conditions, such as exposure, clicks, or conversions. Each stage contributes a different kind of evidence.
Mahta Bio compared three ads in Nexus before launching its campaign and recorded a customer acquisition cost 35% below its historical campaign average. Read the Mahta Bio case study.
Record the initial hypothesis, what you revised, and what you will observe in market. If the recommended ad falls short, examine context, distribution, and the limits of the hypothesis. Keep the original question on record so you can see what you actually learned.
For media experiments and scaling criteria, see the guide to ad testing and in-market A/B testing.
An example from campaign development
Imagine a fictional food brand launching a breakfast product. One direction emphasizes convenience for people who leave home early. The other highlights the ingredients. The first round should explore what each idea communicates and what proof the audience would expect.
In this hypothetical example, convenience draws interest, but some people assume the product replaces a full meal. The ingredients angle feels more specific, but leaves questions about preparation. The feedback points to different revisions.
The team can clarify the everyday use case in the first option and simplify the demonstration in the second. It can then compare the revised ads under the same conditions. An option that better serves the objective without creating a false expectation can move on to in-market validation. These signals are illustrative and do not come from a Galaxies study.
How do you build it into the approval process?
Fit creative intelligence into the production schedule. Identify which decisions need evidence before work moves forward and who will frame the question. Involve the creative team because it knows what can still change and what each revision will cost.
Flag the most consequential decisions at kickoff. If the concept is still open, a CTA test can wait. If production is already booked, the promise or information sequence may still be adjustable. Match the study to a decision the team can act on at that stage.
Make time to interpret the findings together. Product can explain offer conditions, creatives can locate the source of an ambiguity, and marketing can assess whether the feedback matters for the objective. That gives isolated comments context before they become change requests.
Agree on who approves revisions and when to test again. A straightforward factual correction can be checked directly. A change to the promise or meaning deserves a new look because audiences may interpret it differently.
What if there is no clear winner?
The alternatives may be understood in similar ways, or the difference may be too small to matter. That can save the team a long debate over a minor detail.
A round may also have combined audiences or criteria that should be read separately. One ad could explain the offer better while another fits the brand better. Decide which job matters most for this campaign before combining results.
If every option runs into the same barrier, revisit the shared issue. When none makes the benefit relevant, wording changes alone may have little effect. Look at the offer, the situation in which it would be used, or the audience before producing more variants.
What belongs in the approval meeting?
Bring the alternatives, the criteria, and the reasons for your recommendation. Add the main objections by audience, the revisions made, and what still needs validation. That gives decision makers a basis for choosing.
Carry the learning into the next brief. If successive ads raise the same question about how the product works, the explanation may need revision. If the main promise always requires a long explanation, reconsider the value proposition.
Before your next media buy, identify one open decision and two comparable options. Decide what the audience should understand and what evidence would justify moving ahead. Talk to Galaxies about exploring that choice in Nexus.
At Galaxies, we call the time between a consumer signal and a brand decision the Signal-to-Decision Gap. In creative work, that gap narrows when teams can compare options between briefing and approval.
Frequently asked questions
Is creative intelligence a tool for generating ads?
It can incorporate AI tools, but the focus here is choosing with evidence. Generating assets gives you options. Assessing comprehension, relevance, and credibility helps you decide which to revise and which to validate.
Can you test an idea before producing the video?
Yes. You can explore a concept or script before final production. Interpret responses as feedback on that material. Reactions to a script do not establish how the finished video will be received.
Can pretesting predict CTR or conversions?
Simulating how an audience might respond cannot measure actual CTR or conversions. Those metrics depend on media delivery and other conditions. Pretesting helps narrow the hypotheses to verify in market.
What is the role of human judgment?
The team sets the objective, evaluates the evidence, and makes the decision. Brand constraints, feasibility, risk, and responsibility for the promise still require human judgment. Our guide to the limits of synthetic personas explains why.
About the author
Carol Yoshida is Head of Product at Galaxies, a Marketing Decision Intelligence company. She leads development of Nexus: Galaxies Lab, a platform that helps marketing leaders make decisions about creative, claims, audiences, and messaging before committing budget, with synthetic personas built from consumer research and human review.



