Havish M Consulting
Customer Case Study · Marketing & Creative Intelligence
Explore the analyzer
Housing Finance × Havish M Consulting

Great creative deserves
a clearer review.

From “I like it” to “here’s why”. An AI-powered ad analyzer brings colour, copy and call-to-action into one structured marketing conversation.

Client Housing FinanceFunction MarketingSolution Ad Analyzer + API
Enter the creative lab See the original tool
Core creative lenses
3lenses
Colour · copy · call-to-action
Shared review format
1framework
Consistent, actionable feedback
Original tool screens
2views
Upload experience + analysis
Live lab integration
APIpowered
OpenAI Responses API

Case narrative and original screenshots supplied for this study. New concepts and rubric scores demonstrate the experience; they are not campaign performance results.

The Challenge

Creative reviews need more than a personal preference

The objective: bring consistency and objectivity to ad evaluations, using AI to examine colour, copy and call-to-action through a repeatable set of questions.

Different opinions

A visual can trigger many reactions. The team needs a shared way to explain what works and what needs attention.

Unclear messages

Strong design can still leave the audience unsure about the main message or the benefit being communicated.

Missing next steps

A memorable headline needs a clear action. Reviewers should be able to identify what the audience can do next.

Inconsistent reviews

Feedback scattered across conversations makes it harder to compare iterations and retain the reasoning behind a change.

HMC Approach

Give every creative the same thoughtful questions.

Turn an uploaded visual into a structured first review, then give the marketing team the context to refine it.

The Solution

Upload a creative. Get a more useful conversation.

An AI-powered internal tool lets teams upload ad visuals and receive structured insights on design tone, message clarity, CTA effectiveness and brand alignment.

01 · The original tool

From a blank upload area to structured insight

Explore the supplied screens of the Ad Analyzer.

Original screenshots
A simple starting point. Upload or drag an ad into the tool, preview it and request a review.
01Upload→02Analyze→03Refine

The supplied original interface uses Google Gemini. The live interactive analyzer below now uses OpenAI, with human review throughout.

A structured response. Colour palette analysis and ad copy feedback are visible; the CTA section continues beyond the supplied capture.
02 · Interactive creative lab

See the creative. Inspect the reasoning.

Explore two sample concepts, scan an image’s colours locally, or connect OpenAI for a new AI review.

Illustrative assessment

Initial concept

Illustrative visual
Illustrative initial home-loan marketing concept

New demo artwork. Not an active advertisement or offer.

Image720 × 900
Palette6 colours
ReviewDemo rubric

PNG, JPG or WebP · up to 5 MB · drag and drop supported

Connect OpenAI for a live review ↗

Send the selected image and your brief to OpenAI only when you choose Analyze with OpenAI.

Your key is held only for this tab and is never saved or exported. Live calls require internet access and use your OpenAI API account, with API usage billed separately. Use an API key here, not your ChatGPT password. For a shared production tool, keep the key behind your own server.

Creative review

A strong idea. A less obvious next step.

Hand-authored demo rubric

The concept has a recognisable housing theme, but its broad copy and missing CTA leave the next step unclear.

Review scores guide discussion. They do not predict clicks, leads or conversions.

Priority for the next iteration

Add one explicit CTA with a clear destination.

Adjust what this review prioritises 4 dimensions

Weights are normalised to 100%. They change this review index, not the underlying observations.

Local colour scan works offline · OpenAI is optional
03 · Creative anatomy in 3D

Pull the creative apart. Keep the message together.

Explore how colour, copy and CTA work as distinct layers of the same composition.

Interactive 3D

Drag to rotate · choose a layer to focus · conceptual visualisation

04 · Colour & legibility studio

A colour choice you can actually measure.

Try a foreground and background pair. The contrast ratio updates as you change either colour.

Calculated in your browser
YOUR NEXT CHAPTER

A clearer path
to a new home.

Explore your home loan options.

Take the next step ↗
Illustrative copy · colour-pair preview

—:1
Contrast ratio

Checking colour pair…

Try a pair

WCAG AA reference: 4.5:1 for normal text; 3:1 for large text. This checks the selected solid colour pair, not the entire ad’s accessibility. W3C guidance ↗

05 · The refinement canvas

Move from feedback to a visible change.

Drag the divider to compare a constructed initial concept and a refined version.

Illustrative redesign
Illustrative initial ad concept
Illustrative refined ad concept with clearer message and explicit CTA
↔
InitialRefined
A focused creative brief

Three changes. A clearer handoff.

  1. 01
    Lead with one message

    Use a concise headline and one supporting line instead of several competing ideas.

  2. 02
    Make the action explicit

    Give the CTA a visible label and connect it to a defined next step.

  3. 03
    Strengthen the hierarchy

    Use contrast and space to separate brand, message and action.

Scores are hand-authored examples of this rubric. The concepts have not been tested with customers.

06 · Behind the experience

A simple interface. A structured API conversation.

Follow the journey from an image and brief to reviewable observations.

Begin with the creative and its context

A visual tells only part of the story. The intended audience, message and brand guidance make the review more useful.

Workflow demonstration
Workflow playback is a simulation. A live request is made only with the “Analyze with OpenAI” control above. Credentials and image bytes are excluded from these examples and exports.





The Outcome
Faster reviews. More evidence in the discussion.
The tool enabled faster, insight-driven creative reviews, helping the marketing team refine campaigns with evidence-based feedback rather than subjective opinion.
A shared structure for a better first review.
Colour, copy and CTA feedback gives the team something specific to question, improve and carry into the next iteration.
Qualitative outcome supplied for this case study
Ready for the next creative review?
0 / 5
Use this checklist to guide your handoff.



Before vs After
From open-ended reactions to actionable feedback
Illustrative workflow comparison.
DimensionUnstructured reviewAI-assisted review
Starting pointPersonal reactions to a visualColour, copy and CTA as consistent review lenses
Message clarityBroad comments such as “make it stronger”Specific observations about the headline and supporting copy
Call-to-actionAssume the next step is obviousCheck whether an action and destination are actually visible
Brand discussionRely on memory and visual familiarityCompare visible cues against the supplied brand context
IterationChanges without a shared rationalePrioritised recommendations that explain what to revise
DecisionFeedback spread across conversationsStructured insights for the marketing team’s final judgement



Tools & Tech Used
A practical AI tool, with an interactive story around it
OpenAI API live multimodal creative analysisREST API + JSON structured responsesD3.js new interactive chartsThree.js / WebGL new 3D layer explorerCanvas new local colour scanEmbedded assets one standalone HTML


Key Takeaways
Make the feedback useful enough to act on
01
Standardise the questions
A consistent review structure makes it easier to compare creative ideas and explain a revision.
02
Connect evidence to action
Pair every observation with a practical recommendation for colour, copy or call-to-action.
03
Keep the team in control
AI provides a first review. People validate brand alignment, claims and the final creative decision.
Less “make it pop”. More “here’s the next change”.
Explore the creative. Inspect the feedback. Refine with purpose.
Return to the creative lab ↑





Reference 1 of 2


Sources & methodology
Know what each part represents
The case study and source images
The objective, solution and qualitative outcome come from the supplied case brief. Two screenshots are embedded as supplied. The screenshots document the original Gemini-based tool; the optional live analyzer on this page has been adapted to OpenAI. The analysis screenshot is cropped at the CTA section. Any offer text inside the historical screenshot is part of that reference, not a current offer reproduced by this page. No measured time saving, lead uplift or conversion change was supplied.
Demonstration concepts and scores
The initial and refined concepts are newly constructed examples. Their observations and 0–5 scores are hand-authored to demonstrate a repeatable review structure. They are not a transcription of the original tool’s output and were not generated by a live API call. The weighted index is 100 × Σ(score × weight) ÷ [5 × Σ(weight)] only when all four dimensions are assessed and total weight is greater than zero. If any dimension is unassessed, the index is withheld and coverage is shown. All-zero weights also make the index unavailable. A higher index means a higher assessment under this illustrative rubric, not better campaign performance.
Local image colour analysis
Images are decoded in the browser and sampled on a small canvas. Transparency is composited over white. RGB values are grouped into colour buckets; the six largest buckets and a remainder category describe approximate pixel shares. Palette extraction does not infer message meaning, audience response or brand compliance. Local uploads stay in the tab unless you explicitly run the OpenAI option.
Live OpenAI review
The optional integration sends the selected image and brief to the OpenAI Responses API using your entered API key, only after you choose Analyze with OpenAI. It requests structured JSON observations for colour, copy, CTA and brand cues. Unreadable or insufficient evidence should yield an unscored dimension. AI output can be incomplete or mistaken; check the source creative. The key is held only in the page’s memory and excluded from downloads. A shared production implementation should route requests through a server that keeps its credentials private.
Contrast calculation
The checker uses sRGB relative luminance and (lighter luminance + 0.05) ÷ (darker luminance + 0.05). The displayed WCAG AA reference points are 4.5:1 for normal text and 3:1 for large text. Large text is generally at least 18 pt regular or 14 pt bold. This is a colour-pair check, not a whole-ad accessibility certification. See W3C’s contrast guidance.
3D scene, comparison and workflow
The 3D composition separates conceptual creative layers; it does not simulate eye tracking. The divider compares two illustrative visuals. API workflow playback is a labelled simulation; live requests use the dedicated OpenAI control. D3 and Three.js are embedded with their licences, with Canvas rendering available when WebGL is unavailable.
Technical references
OpenAI: Responses API · OpenAI: image understanding · Structured outputs · API key guidance · W3C: contrast minimum