Chegg · Product Quality
Approval isn’t adoption: making quality something every team planned for
I’d already won funding for a design system. But nothing on the roadmap made teams use it. So I took the problem to our VP of Product, turned quality into a company goal, and built the system to measure it.
The outcome
Quality became part of how Chegg planned, not something design asked for
What changed in how teams work
Roadmaps set aside time for quality, design system and accessibility reviews, and every product brief put design system and accessibility on the RASCI.
01 · The challenge
The President had funded the system. No team was required to use it.
I’d convinced Chegg’s President to fund Horizon, our design system, by showing him the “Frankensite”: a product that looked different on every page. But approval to build it wasn’t a mandate to use it. Product teams had their own key results, and migrating to Horizon wasn’t one of them.
The PM leading our CMS team was measured on page turnaround, conversion tests and a translation integration. Her pushback was reasonable: Horizon kept losing to work that counted. Meanwhile, inconsistent flows, accessibility gaps and avoidable support contacts kept shipping.
02 · The insight
Teams build what they’re measured on. Adoption needed a place in the goals, not a better pitch.
03 · My approach
I tied quality to our VP of Product’s goal, not design’s
01
What I saw
Erin Johns, our VP of Product, was moving Chegg from code freezes to continuous deployment. Teams were about to ship faster than ever, with no shared way to say whether something was good enough to ship.
02
What I showed her
Inconsistent flows across one journey, accessibility gaps, the issues driving support contacts, and the adoption pushback itself.
03
What we built together
I asked for Horizon adoption on the roadmap. Together we turned it into a quality OKR: a way to measure quality, with design system adoption scored next to ease of use, performance and accessibility.
Why it worked
Faster shipping without a quality bar would multiply the problems I was showing her. Erin sponsored it, and Product Quality became foundational work in the company OKRs. We built the standards space together, with her as the executive voice and me building the system.
04 · The system I built
Automated data where we had it, expert review where we didn’t
Every step in a flow got a score from 1 to 3 on each standard. Expert review used PURE, a method from Christian Rohrer and Nielsen Norman Group. I extended it past ease of use to design standards and accessibility, and combined it with performance data teams already tracked.
Automated data
- System uptime
- Errors and crashes
- Core Web Vitals
- Task completion
- Site incidents
Expert review
- Design standards (Horizon adoption)
- Accessibility (WCAG 2.1 AA)
- Ease of use (PURE)
Every step scored 1–3
Reviewers never scored a flow they designed.
Each flow gets a color and a deadline
Every score had an owner
I defined who owns what at three levels. An experience is led by its program lead and decided by the OKR. A surface has a “conservator” who guards it over time. A component belongs to the design system. Each flow also had one point of contact responsible for its scorecard.
05 · The pilot
We piloted on Chegg’s four most-visited flows
Most students pass through these flows, so fixing them reached the most people. Together they cover the whole subscription: joining, getting back in, and leaving. Each flow had its own PM, design lead, engineering manager and point of contact.
01
Search to subscription
Q&A page from Google → sign-in → interstitial → payment → onboarding
02
Cancellation
Orders → tell us why → give us another shot → before you go → confirmation
03
Device management
Register a device → my devices → device limit → are you sure → success
04
Reset password (2 flows)
Security → reset dialog → check your email → email → new password
Cancellation: what I pushed on
Cancellation is where retention pressure collides with a principle we’d written down. I partnered with our Customer Ops leader, so what students contacted support about became the target.
“No dark patterns: We do not ‘trick’ users and are upfront and clear on what they are purchasing and provide simple ways to cancel.”
From Chegg’s product principles, under Trustworthy
Account
Cancel is easy to findNo hiding the button or burying it in menus.
Offer alternatives
Real optionsPause the account, see offers, or contact support.
Exit survey
Listen, then respondOffer what fits the reason the student gave.
Cancelled
Keep what they paid forAccess until the period ends, and an easy restart.
Confirmation email
A person, not a receiptA note from support or the CEO they can reply to.
Follow-up
Stay in touchTimed emails that invite students back.
Result
The redesign shipped, and Customer Ops reported fewer cancellation-related contacts.
06 · What the scores exposed
The steps where students pay scored worst
About 80% of Chegg’s traffic came from Google search, so the path from a search result to a subscription mattered most. Its first scorecard showed where it broke.
13Q&A page
13Sign-in
12Choose a plan
16Payment
The answer
| Standard | Signed-out Q&A | Sign-in | Interstitial | Payment | Onboarding | SQnA | Signed-in Q&A |
|---|---|---|---|---|---|---|---|
| Ease of use | 1 | 1 | 2 | 2 | 1 | 1 | 1 |
| Accessibility | 1 | 3 | 1 | 3 | 1 | 1 | 1 |
| Design standards | 3 | 3 | 2 | 3 | 1 | 3 | 3 |
| System uptime | 1 | 1 | 2 | 1 | 1 | 2 | 2 |
| Errors and crashes | 1 | 1 | 1 | 1 | 1 | 1 | 2 |
| Core Web Vitals | 3 | 2 | 2 | 3 | 2 | 2 | 2 |
| Task completion | 3 | 2 | 2 | 3 | 2 | 1 | 1 |
| Step score | 13 | 13 | 12 | 16 | 9 | 11 | 12 |
Accessibility
Sign-in and payment both scored 3. Sign-in’s form fields had no heading structure, and images were missing alt text.
Design standards
Five of seven steps scored 3: no Horizon adoption. The Q&A page still ran on Fanta, the old system. Adoption was now visible, step by step.
Performance
The signed-out Q&A page and payment failed on both speed and task completion. Those are the first and last steps to a subscription.
07 · Making it stick
From design’s request to how teams worked
The PM whose team had pushed back became one of our biggest advocates. Her team built Chegg’s global header and footer on Horizon, the parts students see on every page.
“Kiley was an influential voice for design standards and accessibility… providing the much-needed support that my team required in creating the footer and header in the CMS.”
Julia Goodwin, Product Manager, CMS · Chegg
08 · What I learned
One lesson, and what I’d do differently
Getting approval is the start. Adoption happens when quality is part of how teams are measured.
Measure against yourself
Scores weren’t normalized, so a longer flow always scored higher. A scorecard compares a flow with its own past, not with other teams.
What I’d do differently
Agree on the measure before launch
Support told us cancellation contacts went down, but we hadn’t agreed on the metric and its baseline first, so the result is a report, not a number. Now I set the success metric with partners at kickoff.
I’d also calibrate the review panel before the first real score. We brought in reviewers from marketing, support and QA, and their scores were inconsistent until they learned the rubric.
As AI speeds up how fast teams ship, a quality bar you can measure matters more, not less.