Yelp.com Usability Study
Evaluating user efficiency and cognitive load when navigating multi-constraint search tasks — and finding the friction that a 63% success rate can't hide.
SESSIONS
SUCCESS RATE
TASKS 3 & 4
VARIANCE
The Study
Evaluate how easily users can find specific businesses on Yelp when they have multiple must-have requirements — a specific time of day, noise level, or pet policy. Success and efficiency (time on task) were measured across 5 real-world task scenarios.
Unmoderated testing on the Lyssna platform allowed us to capture authentic user behavior without observation bias. Every participant was assigned the same city per task, regardless of their physical location — ensuring search results, filters, and business profiles were identical across all 30 sessions.
| ATTRIBUTE | DETAIL |
|---|---|
| Participants | 6 participants, USA, age 18+ |
| Sessions | 30 unmoderated sessions (6 participants × 5 tasks) |
| Platform | Lyssna.com — desktop usability testing |
| Recruitment | Private links distributed to classmates, friends, and family |
| Controls | Fixed city per task — all participants saw identical results |
| Measures | Task success (completion) and efficiency (time on task) |
★ Location was fixed per task to eliminate local search bias and allow accurate time-on-task comparisons across all 30 sessions.
5 Real Situations. Real Stakes.
You're planning a retirement party for 12 people this Friday. Find a Mediterranean restaurant with a 4.0+ rating and locate where you'd start a reservation for the group.
You just moved to San Diego and need a hair salon that's open on Sundays and offers online booking. Find one and click the "Request an Appointment" button.
Your power goes out. You need a coffee shop to work from. Search reviews for a shop that's described as "quiet" and has free wi-fi, then click Get Directions.
You're traveling with your large dog and want to go to brunch. Find a spot with outdoor seating and confirm the reviews mention it's dog friendly.
You're on vacation with your partner. Find a massage therapist in Scottsdale with a 4.5+ rating and click the "Request an appointment" button to complete the task.
Results at a Glance
| TASK | SUCCESS RATE | AVG TIME | FASTEST | SLOWEST | PRIMARY FRICTION |
|---|---|---|---|---|---|
| T1: Group Dining | 83% | ~1.7 min | ~0.7 min | ~2.6 min | Filter awareness — used search bar instead of sidebar |
| T2: Service Hours | 67% | ~2.1 min | ~1.0 min | ~5.1 min | "Open At" looked static — users checked profiles manually |
| T3: Quiet Coffee | 50% | ~3.4 min | ~1.4 min | ~7.0 min | No "noise level" filter — manual review scanning required |
| T4: Dog Brunch | 50% | ~1.8 min | ~0.5 min | ~4.0 min | Over-reliance on filter — skipped verification step |
| T5: Booking | 67% | ~3.1 min | ~0.7 min | ~6.9 min | Inconsistent CTA label — users couldn't find the right button |
Out of 30 sessions, 19 were successful. 11 failures were caused by tools that were hidden, mislabeled, or not recognized as interactive.
Four Patterns Across 30 Sessions
Almost every task, users defaulted to the search bar, not because they preferred it, but because the interface's structured filters failed them. When a filter couldn't be found, the search bar became the escape valve. Users typed "quiet coffee shop" directly when no "noise level" filter existed, and used text search for "Mediterranean" even though it was a filterable sidebar attribute.
Observed across all 5 tasks50% of users failed Task 4 the moment they applied the "Dogs Allowed" filter — they considered the task complete. The checkbox felt like system-level confirmation, so they skipped the human-level verification step (checking a review). When filters returned no results at all, the opposite happened: users lost faith in the structured path and reverted entirely to manual methods.
50% failure rate — Task 4In Task 2, the "Open At" text appeared in blue — but with no dropdown arrow, border, or underline to indicate editability. Users scanned past it and spent most of their time manually clicking into profile after profile. The feature existed. The affordance didn't. This pattern repeated with "Open Now," which looked like a status label rather than a clickable filter.
Key friction point — Task 2Task 5 revealed that Yelp displays different CTA labels for the same business depending on how the user arrived. Searching "Massage Therapist" produced "Get pricing & availability." Searching "Massage Therapy" produced "Request an appointment." Users with a rigid mental model for the button they were looking for couldn't find it — because the button they were looking for kept changing.
7-min time variance — Task 5In Their Words
There are so many filters, it's difficult... I don't see a 'quiet' tag.
This is a lot of options to go through.
I know you can search user reviews.
I've already filtered by 'dogs allowed,' so the task is complete.
What Yelp Built vs. What Users Did
The "Open At" filter under Features is technically functional; clicking the blue text opens a date/time picker where users can specify a day and hour. The "Open Now" checkbox under Suggested also applies a real-time hours filter. Both tools exist. Both work.
- "Open At" appeared as static blue text with no dropdown arrow, no border, no affordance
- "Open Now" read as a status label showing current time, not a clickable filter
- Editing the time required a 3-step sequence users never discovered
- Users spent most of their time manually clicking into profiles to check Sunday hours
- 67% success rate; the tool existed, the signifier did not
Yelp places a "Search Reviews" bar directly within the reviews section of every business profile. This bar allows users to search for specific terms across all reviews, which is exactly what was needed to find mentions of "quiet," "dog friendly," or "wifi" without scrolling manually.
- Search bar was buried below the map, photos, and business info, mid-page
- By the time users scrolled to that point, they were already reading reviews one-by-one
- The bar's position made it appear supplementary, not essential
- 50% of Task 3 users never found it and spent up to 7 minutes reading manually
- The one user who succeeded on Task 4 explicitly knew to use it from prior experience
Yelp's primary CTA button changes label depending on how the user arrived at a business listing. Searching "Massage Therapist" displays "Get pricing & availability." Searching "Massage Therapy" or "Thai massage" displays "Request an appointment." The underlying action is identical (scheduling a session) but the label is not.
- Users built a rigid mental model around the button name from the task prompt
- When the label didn't match, they assumed the feature was unavailable
- Multiple participants returned to search with a different query to find "the right button"
- Time variance peaked at 7 minutes, entirely due to this label mismatch
- 67% success rate, with failure concentrated in users who arrived via specific search terms
Three Targeted Fixes
- Move the "Search Reviews" bar directly under the Overall Rating section
- Add placeholder microcopy: "Search for quiet, dog-friendly, wifi…"
- Apply a visible border and background to signal interactivity
- This single change addresses the 50% failure rate on Tasks 3 & 4
- Add a chevron arrow and border to the "Open At" time control
- Style "Open Now" as a toggle button, not a label; it should look pressable
- Reduce the time-edit flow from 3 steps to 1 (single click to open picker)
- Add a "Good for Work / Study" filter to replace the "quiet" search workaround
- Use one label: "Request an Appointment" across all search entry points
- Add a calendar icon to reinforce the action's meaning at a glance
- Never vary button text based on search query — users carry mental models between sessions
- This eliminates the 7-minute time variance gap entirely