Kroger Pharmacy IVR - vaccine scheduling

Tested and refined Kroger’s IVR system to encourage vaccine scheduling and reduce calls to the pharmacy
Service design
UX strategy
Conversational UX
Header image

Phones were ringing, associates were drowning

At peak vaccination season, pharmacy associates were spending a huge chunk of their day handling inbound calls for vaccine scheduling.

They were doing what they could, but it came at a cost:

  • Less time for in-store patient care
  • Long call queues and frustrated callers
  • Inability to scale

The process “worked” but only because humans were filling the gaps.

What was actually at stake

This wasn’t just an efficiency exercise. The stakes were real:

  • Missed or delayed vaccine appointments
  • Frustrated patients and associates alike
  • No scalable solution to handle call volume surges

Without intervention, call handling relied entirely on memory and follow-up, a fragile, high-stakes system.

What we did

We stopped treating IVR like a phone tree and started treating it like a capacity strategy.Instead of routing callers to stores faster, we asked: What if the system could actually finish the job?

So we:

  • Built a dedicated vaccine scheduling path designed specifically for containment
  • Integrated IBM Watson to handle natural language and edge cases
  • Smoothed friction points (especially date-of-birth failures) through clearer prompts and A/B testing
  • Designed intentional system handoffs so transitions didn’t feel like errors
  • Closed calls with confirmation, reducing repeat dials

The goal wasn’t perfection. 
It was containment without frustration.

My role

I owned the end-to-end strategy for redesigning the vaccine scheduling IVR:

  • Mapped call patterns and pain points from existing IVR reports
  • Interviewed pharmacy associates to capture real-world behaviors
  • Designed and prototyped a dedicated vaccine scheduling IVR flow
  • Conducted usability testing and A/B experiments to refine prompts and transitions
  • Partnered with engineering, analytics, and IBM Watson teams for implementation

The results (3-month pilot)

In three months, the pilot handled 22,734 calls.

14,163 callers
Successfully moved through intent recognition and scheduling
5,539 appointments
Were fully scheduled inside the IVR

Discovery: understanding the real system

Before redesigning anything, we needed to see reality, not assumptions.

What we learned:

  • High-call-volume bottlenecks existed at repeat scheduling points and date-of-birth prompts
  • Existing IVR flows mirrored competitor patterns defaulting to store transfers or web URLs, which often increased drop-offs
  • Pharmacy associates spent considerable time repeating information already captured in-system
  • Caller frustration spiked when the system couldn’t handle edge cases or unfamiliar responses

Design moves: how we solved it

Smoothing transitions

Telecom systems and Watson had different tones and response styles.

  • We designed a handoff message to prepare callers for the switch
  • Tested it against a version without the message
  • Result: fewer drop-offs and smoother experiences

Instead of email chains, we created structured workflow.

Ending calls well

Previous IVR flows ended abruptly callers hung up, often needing another call.

  • Added early task identification and closing prompts
  • Ensured all caller needs were addressed before disconnect
  • Reduced repeat calls and boosted caller confidence

Solving the date-of-birth drop-off

We noticed a major friction point when the IVR couldn’t recognize DOB inputs.

  • Introduced an example format prompt
  • Ran an A/B test to measure improvement
  • Outcome: significant reduction in hang-ups at this step

Design moves: how we solved it

Smoothing transitions

Telecom systems and Watson had different tones and response styles.

  • We designed a handoff message to prepare callers for the switch
  • Tested it against a version without the message
  • Result: fewer drop-offs and smoother experiences

Instead of email chains, we created structured workflow.

Ending calls well

Previous IVR flows ended abruptly callers hung up, often needing another call.

  • Added early task identification and closing prompts
  • Ensured all caller needs were addressed before disconnect
  • Reduced repeat calls and boosted caller confidence

Solving the date-of-birth drop-off

We noticed a major friction point when the IVR couldn’t recognize DOB inputs.

  • Introduced an example format prompt
  • Ran an A/B test to measure improvement
  • Outcome: significant reduction in hang-ups at this step

What this project really demonstrated

This project reinforced the principle: automation is an ally, not a replacement.

  • Pilots reveal what’s possible within real operational constraints
  • Testing and iteration matter more than assumptions
  • Preserved familiar workflows to minimize adoption friction
© 2026 Andrea B