How to 10x your product teardown with ChatGPT

Never say "make it pop" again

On the long list of things PMs need to be good at, product design is among the most important.

While designers are responsible for creating high-fidelity designs, product managers must be able to give thoughtful and helpful critiques to improve these designs, and of course their product overall.

But doing it right is nuanced and takes practice. Less experienced PMs (and most executives) will rampage through a designer's hard work, saying "make it pop!" which is more effective at making designers want to pull their hair out.

So let’s figure out how to never do that and instead make designers actively seek out and appreciate your feedback.

So it feels less like this.

Today we’ll level up your product teardowns by looking at:

  1. Overcoming impostor syndrome – who is a PM to critique a designer’s designs anyway?

  2. How to lead a product design teardown as a product manager

  3. Powering up this process with ChatGPT’s image input

  4. Resources to learn more

A quick word from this issue’s sponsor: Zeda.io

Zeda.io is an AI-powered product discovery and strategy software that helps product teams uncover problems that matter to customers, decide what to build next, and create strategies to drive outcomes.

  • Boost product-led sales by up to 50%

  • Free up over 90 hours every month for team collaboration

  • Cut customer churn rates by 20%.

  • Craft product strategies in just a third of the time.

Overcoming Impostor Syndrome

It’s easy to feel intimidated critiquing the work of an expert designer. Would we critique a civil engineer’s bridge design?

But remembering your unique value and perspective as a PM can help overcome this impostor syndrome.

Why PMs can add unique value when providing design feedback:

  • You're a voice of the customer. I won’t say you’re the voice of the customer because hopefully everyone on your team understands your users. But PMs tend to have a unique understanding of user needs from research and being familiar with exactly how past products have performed.

  • You have the most business context. Fundamentally, the PM role is a business function more than a technical one. (Maybe a spicy take 🥵.) PMs are uniquely suited to gauge how well a feature’s designs map to core business goals and metrics.

  • You can help descope features. PMs have a critical role in helping designers find ways to simplify their features and prioritize features that deliver the most value for the least engineering effort.

And as one final point, remember: design is not art. Unlike art, which is purely subjective, one design can be objectively better than another. Products exist to solve a problem, and whichever design solves a problem the most effectively is objectively the best design.

At the end of the day, you and the designer share a common goal of delivering the best possible experience. Offer feedback in that spirit, and trust that your unique expertise can provide value.

The Art of the Critique

Alright, you’re ready to give the critique. How should you do it?

Understand context

Before diving into the details of a design, make sure you understand what the designer needs right now. Are they looking for validation on a rough concept or final sign-off? Do they want you to focus on copy right now, or is that just placeholder text and you should focus on the overall flow?

Giving feedback on parts of the experience your designer hasn’t worked on is the fastest way to make them hate you. Meeting them where they are is crucial.

Ask your designers which area they’d like you to focus on:

  1. Overall concept and direction

  2. User experience and flow

  3. Functionality and usability

  4. Content and copy

  5. Technical feasibility

Start broad

Run through these 5 areas when evaluating a design at a high level:

  1. Business goals: Does this meet our key business goals? Will it clearly move the 1-2 core metrics that matter most (e.g. conversion rate, retention)?

  2. User needs: Does it address the top 1-2 user needs we've validated through research? Or are we relying on assumptions vs insights (be specific on the needs)?

  3. Simplification: Can we simplify and streamline the happy path while retaining most of the value? This reduces engineering complexity upfront.

  4. Information architecture: Is the information architecture intuitive and naturally aligned to how users will think? Or is it modeled too much on our internal structure?

  5. Visual hierarchy: Are we promoting the right actions through visual hierarchy, placement, and clear calls to action?

It’s important to start broad because these changes often make the smaller changes irrelevant. You don’t need to make copy suggestions on a page that won’t even be in the design.

Then get tactical and detailed

Now it’s time to get granular and think through these 4 areas:

  1. Copy improvements: How can we improve the copy to be crystal clear and scannable for our users? Are there any confusing terms or vague calls-to-action we should reword?

  2. User flows: Walk through the flows step-by-step. Are there any sticking points where users may get confused or frustrated? How can we smooth these out? Be specific about the friction you’re seeing?

  3. Visual design: Are there opportunities to simplify and tighten the UI design itself for greater efficiency? Can we align spacing, reduce clutter, increase clarity to make every element earn its place?

  4. Prioritization: Can we break complex features into smaller MVP milestones to reduce risk and test what resonates (provide suggestions on what to break out)?

Leave no stone unturned here! It’s easy for a designer to miss some aspects here because they’re so close to the work.

Your fresh eyes (especially when you try looking through the lens of your users) will help catch lots of improvements.

ChatGPT to the Rescue

Alright, now let’s look at how you can superpower this process with ChatGPT’s computer vision.

ChatGPT is much more helpful at giving more tactical suggestions than overarching comments on related to technical implementation or alignment with strategy. So you’ll still need to do that part.

Quickly, check out how insanely good ChatGPT is at understanding images.

Too good, right?

For this design example, we’ll use the most amazing app ever: GoodRx.

(Yes, I work there.)

We’ll be looking at the “My Best Pharmacy” feature, which helps users find their overall lowest-priced pharmacy based on the prescriptions in their account.

Yes, my team built this feature.

1) Load up your screenshots and ask Chat GPT to interpret them.

Your goal here is just to see if ChatGPT seems to get the idea of the feature.

Correct any misconceptions it has.

Pretty darn good.

2) Give ChatGPT context

Now before you ask ChatGPT for critiques, you need to give it context.

Here’s the prompt:

You are an expert product designer with a deep understanding of product management and excellent product design. Please help me analyze this design to find improvements.

Context: The goal of this feature is to [non-metric goal]. The primary metrics we’re trying to drive are [metrics].

Before we begin this analysis, do you have any questions or need any clarification on this design?

Example:

If ChatGPT has more questions about your product, answer them.

3) Time for the critique

Here’s the prompt:

Still acting as an expert product designer with a deep understanding of product management and excellent product design. Please consider the following questions. Be extremely specific, detailed, and tactical with your suggestions. Implementation should be directly understood, not general ideas or high-level advice.

What potential fiction do you see in these designs? How might we reduce those points of friction?

What would we do to simplify this experience from an engineering perspective while preserving its value for the overall goal.

Playing devil’s advocate, what are some reasons this feature might fail to achieve its goals?

Can you think of lightweight ways we could test the concept in these designs through an experiment that would be cheaper and faster than building the whole feature?

How might we incorporate feedback loops in the design to ensure continuous improvement?

Example:

From here, you can continue to have the conversation however you’d like. You can ask for follow-ups on any of that feedback, or continue with other questions.

Here’s a list of tons of other things you might want to ask:

  1. Information Architecture:

    • How is information structured across the screens, and is it optimized for user understanding and decision-making?

    • Are there ways the data presentation can be improved to facilitate quicker user actions?

  2. Interactive Elements:

    • How can interactive elements be optimized to enhance user engagement and make the user's journey toward the goal more efficient?

    • Are there opportunities to streamline or automate selections based on user behavior patterns?

  3. User Flow Optimization:

    • Does the sequence of screens suggest an optimal user flow, and are there any steps that could be eliminated or combined to reduce friction?

    • Could the flow benefit from more progressive disclosure—revealing information progressively as a user needs it?

  4. Contextual Assistance:

    • Are there opportunities within the screens to offer contextual assistance or smart tips that could anticipate user questions or errors?

  5. Actionable Feedback:

    • Is there a mechanism within the design that allows for quick user feedback which is actionable and directly feeds into improving the service?

    • How can the design encourage users to provide feedback that's constructive and specific?

  6. Feature Discoverability:

    • How discoverable are the core features and how might the design encourage exploration and discovery of additional features?

    • Are there unobtrusive ways to highlight new or unused features to users?

Further reading