Tanvi Mehta · Jan 03, 2026 · 3 min read
UX Research Isn’t About Testing Products, It’s About Preserving Meaning
Tanvi Mehta · Jan 03, 2026 · 3 min read
UX Research Isn’t About Testing Products, It’s About Preserving Meaning
Tanvi Mehta · Jan 03, 2026 · 3 min read
UX Research Isn’t About Testing Products, It’s About Preserving Meaning
The biggest risk in product work isn't building the wrong feature. It's misreading why users do what they do.
The biggest risk in product work isn't building the wrong feature. It's misreading why users do what they do.
The biggest risk in product work isn't building the wrong feature. It's misreading why users do what they do.
Product teams love building, shipping, and testing. Feedback is everywhere, and collecting it is easy. Keeping its meaning intact is the hard part.
Product teams love building, shipping, and testing. Feedback is everywhere, and collecting it is easy. Keeping its meaning intact is the hard part.
Product teams love building, shipping, and testing. Feedback is everywhere, and collecting it is easy. Keeping its meaning intact is the hard part.
Most usability tests record what users say and do. Very few record what users expected, how that expectation shifted, and how they felt when reality didn't match. That gap between prediction and outcome is where the truth lives.
Most usability tests record what users say and do. Very few record what users expected, how that expectation shifted, and how they felt when reality didn't match. That gap between prediction and outcome is where the truth lives.
Most usability tests record what users say and do. Very few record what users expected, how that expectation shifted, and how they felt when reality didn't match. That gap between prediction and outcome is where the truth lives.
Where UX Research Quietly Fails
Where UX Research Quietly Fails
Where UX Research Quietly Fails
A typical research workflow looks like this:
A typical research workflow looks like this:
A typical research workflow looks like this:
• Test with a handful of users
• Test with a handful of users
• Test with a handful of users
• Watch them do tasks
• Watch them do tasks
• Watch them do tasks
• Note their reactions
• Note their reactions
• Note their reactions
• Summarize later
• Summarize later
• Summarize later
The sessions aren't the problem. Synthesis is. That's when nuance gets crushed into bullet points:
The sessions aren't the problem. Synthesis is. That's when nuance gets crushed into bullet points:
The sessions aren't the problem. Synthesis is. That's when nuance gets crushed into bullet points:
• Expectations turn into quotes
• Expectations turn into quotes
• Expectations turn into quotes
• Emotions turn into labels
• Emotions turn into labels
• Emotions turn into labels
• Confusion turns into "usability issue"
• Confusion turns into "usability issue"
• Confusion turns into "usability issue"
• Context turns into a footnote
• Context turns into a footnote
• Context turns into a footnote
The insights are technically correct but missing the point. And half an understanding leads to confident, wrong decisions.
The insights are technically correct but missing the point. And half an understanding leads to confident, wrong decisions.
The insights are technically correct but missing the point. And half an understanding leads to confident, wrong decisions.
What Testing Actually Looks Like
What Testing Actually Looks Like
What Testing Actually Looks Like
If you've run real sessions, you know the chaos. Every participant touches dozens of elements. And for every single moment, four things happen:
If you've run real sessions, you know the chaos. Every participant touches dozens of elements. And for every single moment, four things happen:
If you've run real sessions, you know the chaos. Every participant touches dozens of elements. And for every single moment, four things happen:
• What they expect
• What they expect
• What they expect
• What actually happens
• What actually happens
• What actually happens
• How they feel about the gap
• How they feel about the gap
• How they feel about the gap
• How they explain it afterward
• How they explain it afterward
• How they explain it afterward
AI note-takers capture words well. They struggle with contrast, the difference between what users thought would happen and what did. That's exactly where meaning slips away.
AI note-takers capture words well. They struggle with contrast, the difference between what users thought would happen and what did. That's exactly where meaning slips away.
AI note-takers capture words well. They struggle with contrast, the difference between what users thought would happen and what did. That's exactly where meaning slips away.
The Shift: You're Testing Expectations, Not Screens
The Shift: You're Testing Expectations, Not Screens
The Shift: You're Testing Expectations, Not Screens
Every tap is a little prediction:
Every tap is a little prediction:
Every tap is a little prediction:
• "I think this will do X."
• "I think this will do X."
• "I think this will do X."
• "I expect this means Y."
• "I expect this means Y."
• "I expect this means Y."
• "I assume this leads to Z."
• "I assume this leads to Z."
• "I assume this leads to Z."
Good design matches those predictions. Bad design breaks them, even when the feature technically "works."
Good design matches those predictions. Bad design breaks them, even when the feature technically "works."
Good design matches those predictions. Bad design breaks them, even when the feature technically "works."
Picture a bookmark icon on an article page. The user taps it, expecting to save the article for later. Instead, it pins the article to a shared team board. The feature works perfectly, but the user feels exposed. A standard test logs that as "icon unclear." The real story is "I just shared something I meant to keep private."
Picture a bookmark icon on an article page. The user taps it, expecting to save the article for later. Instead, it pins the article to a shared team board. The feature works perfectly, but the user feels exposed. A standard test logs that as "icon unclear." The real story is "I just shared something I meant to keep private."
Picture a bookmark icon on an article page. The user taps it, expecting to save the article for later. Instead, it pins the article to a shared team board. The feature works perfectly, but the user feels exposed. A standard test logs that as "icon unclear." The real story is "I just shared something I meant to keep private."
Skip the expectation, and teams end up arguing about symptoms instead of causes.
Skip the expectation, and teams end up arguing about symptoms instead of causes.
Skip the expectation, and teams end up arguing about symptoms instead of causes.
Where Meaning Gets Lost: Two Examples
Where Meaning Gets Lost: Two Examples
Where Meaning Gets Lost: Two Examples
Example 1: The fix that didn't fix anything. A team launches a new reporting feature. Early research says users find it confusing. So the team cleans up the flow, cuts three screens, and simplifies the layout. On paper, it's much better. But adoption barely moves. Why? Users never had a flow problem. They expected the feature to answer "how is my team doing?" and it answered "what did my team do?" That mismatch never showed up in the task data. It lived in the meaning behind it.
Example 1: The fix that didn't fix anything. A team launches a new reporting feature. Early research says users find it confusing. So the team cleans up the flow, cuts three screens, and simplifies the layout. On paper, it's much better. But adoption barely moves. Why? Users never had a flow problem. They expected the feature to answer "how is my team doing?" and it answered "what did my team do?" That mismatch never showed up in the task data. It lived in the meaning behind it.
Example 1: The fix that didn't fix anything. A team launches a new reporting feature. Early research says users find it confusing. So the team cleans up the flow, cuts three screens, and simplifies the layout. On paper, it's much better. But adoption barely moves. Why? Users never had a flow problem. They expected the feature to answer "how is my team doing?" and it answered "what did my team do?" That mismatch never showed up in the task data. It lived in the meaning behind it.
Example 2: The form that was already easy. A team redesigns its signup form, cutting it from five steps to two. Super clean. Signups stay flat. The problem was never complexity. It was trust. The form asked for a phone number before users had seen anything worth handing it over for. No amount of streamlining fixes "I don't trust you yet."
Example 2: The form that was already easy. A team redesigns its signup form, cutting it from five steps to two. Super clean. Signups stay flat. The problem was never complexity. It was trust. The form asked for a phone number before users had seen anything worth handing it over for. No amount of streamlining fixes "I don't trust you yet."
Example 2: The form that was already easy. A team redesigns its signup form, cutting it from five steps to two. Super clean. Signups stay flat. The problem was never complexity. It was trust. The form asked for a phone number before users had seen anything worth handing it over for. No amount of streamlining fixes "I don't trust you yet."
Flatten moments like these into bullet points or click-through rates, and you lose the context that should drive design decisions. You end up treating symptoms, not causes.
Flatten moments like these into bullet points or click-through rates, and you lose the context that should drive design decisions. You end up treating symptoms, not causes.
Flatten moments like these into bullet points or click-through rates, and you lose the context that should drive design decisions. You end up treating symptoms, not causes.
A Framework for Preserving Meaning
A Framework for Preserving Meaning
A Framework for Preserving Meaning
To keep nuance intact, I use a simple grid.
To keep nuance intact, I use a simple grid.
To keep nuance intact, I use a simple grid.
• Columns are features, icons, pages, or interactions
• Columns are features, icons, pages, or interactions
• Columns are features, icons, pages, or interactions
• Rows are participants
• Rows are participants
• Rows are participants
• Each cell holds three things:
• Each cell holds three things:
• Each cell holds three things:
💭 What the user expected
💭 What the user expected
💭 What the user expected
⚡ What actually happenedssume this leads to Z."
⚡ What actually happenedssume this leads to Z."
⚡ What actually happenedssume this leads to Z."
🙂 / 😕 / 😡 How they felt about the gap
🙂 / 😕 / 😡 How they felt about the gap
🙂 / 😕 / 😡 How they felt about the gap
That's it. The grid keeps each person's thinking whole until patterns have time to surface.
That's it. The grid keeps each person's thinking whole until patterns have time to surface.
That's it. The grid keeps each person's thinking whole until patterns have time to surface.
Phase 1: Before the Product
Phase 1: Before the Product
Phase 1: Before the Product
Before showing anything, learn the user's world:
Before showing anything, learn the user's world:
Before showing anything, learn the user's world:
• How they solve the problem today
• How they solve the problem today
• How they solve the problem today
• What frustrates them
• What frustrates them
• What frustrates them
• What "success" means to them
• What "success" means to them
• What "success" means to them
• The mental model they walk in with
• The mental model they walk in with
• The mental model they walk in with
This baseline is the lens they'll use to judge everything else.
This baseline is the lens they'll use to judge everything else.
This baseline is the lens they'll use to judge everything else.
Phase 2: First Contact
Phase 2: First Contact
Phase 2: First Contact
Next, watch expectations form before anyone clicks:
Next, watch expectations form before anyone clicks:
Next, watch expectations form before anyone clicks:
• First impressions
• First impressions
• First impressions
• Assumptions
• Assumptions
• Assumptions
• What they think each page, icon, or feature does
• What they think each page, icon, or feature does
• What they think each page, icon, or feature does
• The words they use to describe it
• The words they use to describe it
• The words they use to describe it
Expectations usually form before the first tap. If someone calls your "Workspace" tab "the settings thing," you've already learned something.
Expectations usually form before the first tap. If someone calls your "Workspace" tab "the settings thing," you've already learned something.
Expectations usually form before the first tap. If someone calls your "Workspace" tab "the settings thing," you've already learned something.
Phase 3: The Expectation Alignment Grid
Phase 3: The Expectation Alignment Grid
Phase 3: The Expectation Alignment Grid
For each feature, capture:
For each feature, capture:
For each feature, capture:
• 💭 Expectation
• 💭 Expectation
• 💭 Expectation
• ⚡ What happened
• ⚡ What happened
• ⚡ What happened
• 🙂 / 😕 / 😡 Emotion
• 🙂 / 😕 / 😡 Emotion
• 🙂 / 😕 / 😡 Emotion
Phase 4: Spot Patterns with Red and Green Dots
Phase 4: Spot Patterns with Red and Green Dots
Phase 4: Spot Patterns with Red and Green Dots
Once the grid's full, zoom out and mark it up:
Once the grid's full, zoom out and mark it up:
Once the grid's full, zoom out and mark it up:
• 🟢 Green dots above features where expectations matched
• 🟢 Green dots above features where expectations matched
• 🟢 Green dots above features where expectations matched
• 🔴 Red dots where they broke
• 🔴 Red dots where they broke
• 🔴 Red dots where they broke
Now the big problems jump out without you having to over-synthesize. You'll see:
Now the big problems jump out without you having to over-synthesize. You'll see:
Now the big problems jump out without you having to over-synthesize. You'll see:
• Clusters of confusion
• Clusters of confusion
• Clusters of confusion
• Spots where emotions swing hard
• Spots where emotions swing hard
• Spots where emotions swing hard
• Features that work as intended
• Features that work as intended
• Features that work as intended
• Features that need a redesign
• Features that need a redesign
• Features that need a redesign
If four out of five participants have a red dot on Icon 2, you don't need a debate. You need a new icon.
If four out of five participants have a red dot on Icon 2, you don't need a debate. You need a new icon.
If four out of five participants have a red dot on Icon 2, you don't need a debate. You need a new icon.
Phase 5: Making Decisions
Phase 5: Making Decisions
Phase 4: Spot Patterns with Red and Green Dots
The grid doesn’t just record feedback. It steers decisions by showing you:
The grid doesn’t just record feedback. It steers decisions by showing you:
The grid doesn’t just record feedback. It steers decisions by showing you:
• Where people need more clarity
• Where people need more clarity
• Where people need more clarity
• Which issues are system-wide and which are one-offs
• Which issues are system-wide and which are one-offs
• Which issues are system-wide and which are one-offs
•Where the design is already doing its job
•Where the design is already doing its job
•Where the design is already doing its job
Why This Matters
Why This Matters
Phase 4: Spot Patterns with Red and Green Dots
This approach changes how each part of the team thinks:
This approach changes how each part of the team thinks:
This approach changes how each part of the team thinks:
• Design: Clarity beats cleverness.
• Design: Clarity beats cleverness.
• Design: Clarity beats cleverness.
• Product: Alignment beats feature count.
• Product: Alignment beats feature count.
• Product: Alignment beats feature count.
• Research: Synthesis is a design act.
• Research: Synthesis is a design act.
• Research: Synthesis is a design act.
Great products don’t remove all friction. They remove surprise where it matters most.
Great products don’t remove all friction. They remove surprise where it matters most.
Great products don’t remove all friction. They remove surprise where it matters most.
Final Thought
Final Thought
Phase 4: Spot Patterns with Red and Green Dots
Most UX research doesn’t fail loudly. It fails quietly. The data isn’t wrong. The meaning just gets flattened too early.
Most UX research doesn’t fail loudly. It fails quietly. The data isn’t wrong. The meaning just gets flattened too early.
Most UX research doesn’t fail loudly. It fails quietly. The data isn’t wrong. The meaning just gets flattened too early.
Lose the expectations, and you lose the understanding. Lose the understanding, and you ship confidently in the wrong direction.
Lose the expectations, and you lose the understanding. Lose the understanding, and you ship confidently in the wrong direction.
Lose the expectations, and you lose the understanding. Lose the understanding, and you ship confidently in the wrong direction.
UX research isn’t about testing products. It’s about protecting how users think long enough for the patterns to tell the truth.
UX research isn’t about testing products. It’s about protecting how users think long enough for the patterns to tell the truth.
UX research isn’t about testing products. It’s about protecting how users think long enough for the patterns to tell the truth.