Honor is a network and platform connecting clients, caregivers, and local agency owners to provide non-medical home care to aging adults. Honor’s operations team handles all visit logistics and customer service, and tickets in the administrative console are their primary mode of internal task tracking. Tickets are used for handling inbound issues from clients and caregivers, outbound tasks such as outreach and incident follow-up, or internal coordination across teams.

The principal designer for our pod did broad research to understand where our operations team spends the most time in an effort to reduce operating expenses, and she found that tickets was one of the more problematic areas. The ops team spends ~15% of their week in tickets, and I was tasked to own this product area and investigate how we might reduce this time.

Problems

I interviewed several roles across our operations team to understand ticket workflows and found several opportunity areas related to ticket triage: how the team prioritizes work and determines a ticket is ready to be worked on.

If a ticket or any sub-task within it is assigned to an individual, it will show up in their inbox, resulting in a queue of hundreds of tickets. With such a large queue, the ops team spends a significant amount of time manually managing their and/or their team’s inbox daily, outside of the actual ticket work to be done. This might include triaging which tickets to prioritize, handing off tickets to another teammate, or pushing due dates for overdue tickets. Inbox management is made worse by what we identified as “non-actionable tickets”.

Non-actionable tickets do not require action from you right now or any action at all. Tickets can have sequential tasks that must first be completed by other people or teams. Some tickets must be completed on a future date, like a client check-in or reminder, and other tickets have no action needed at all and serve as an internal note. One client manager I spoke to had 190 tickets in her inbox, but after triaging, only 3 needed to be actively worked on for the day.

These non-actionable tickets add noise to their inbox and force them to re-triage and re-investigate each ticket daily to determine if something needs to get done today. This process relies on the ops team to constantly check for updates but also leads to the team leaning on their memory and various workarounds. This eventually results in work falling through the cracks and a poor experience for our clients and caregivers.

“I feel like we almost miss things because we're kind of ignoring the top half of our tickets because, oh, well, yeah, I've already seen that one.” — Regional manager

The ops team created workarounds and out-of-product solutions, but these have created more touch points and taken even more time out of the ops team’s day.

Goals & metrics

If the system can identify and segment actionable versus non-actionable tickets, important tickets will be completed faster and are less likely to be overlooked while also reducing cognitive load for the ops team.

Success metrics

• Reduce the number of low value touch points in critical work, with fewer people involved in each ticket as a proxy metric

• Reduce time spent in tickets overall

• Reduce number of overdue tickets

Business goals

• Increase client satisfaction around issue resolution

• Improve Care Platform economics

Solution

In the future, we see tickets becoming a personalized queue of work that tells you when a ticket is ready for your input, with many tasks and processes becoming automated. For this project, engineering work was scoped for 4-6 weeks, and we were dealing with a legacy ticketing system with considerable tech debt. We agreed as a team to start with managing a subset of non-actionable tickets in the existing inbox. In addition to reaching the success metrics above, specific goals for this solution include:

• Serving the immediate need to stay on top of tickets and increase accountability

• Building trust between the users and the system with a smaller automation and visibility into the logic, as a step towards the future

• Visualizing the mental model and distinction between actionable and non-actionable tickets by creating different spaces for them

Some discarded ideas that didn’t make the cut due to higher effort and lower impact:

MVP solution

Tickets are divided into two tabs within the inbox— “Action items” and “Everything else”

• New and existing tickets are automatically sorted into the “Everything else” tab if they are primarily for documentation (based on key words in ticket titles), if the due date is more than 90 days out, or if they are assigned to a known placeholder/dummy account

• If any of the above attributes are changed, the system informs the user where the ticket has been moved and why.

• The ops team can explicitly move tickets to or from either tab, and we gather additional data on why to improve the sorting logic.

• This separate tab allows the ops team to audit the “Everything else” tab as they get used to this automation, though we anticipate the need to do this will decrease with time.

Results and next steps

The project was rolled out to a test group in Dec 2022 and fully rolled out in January 2023. Over a 4 week period post-launch, we noted that out of 1900 tickets auto-sorted, only 1 ticket was manually placed back into action items tab, affirming our initial sort parameters. The average time to close non-actionable tickets decreased by about 49%, or 131 days.

However, our target metrics showed little movement. Despite the mental footprint the non-actionable tickets occupied, they may not affect the time spent in the inbox drastically, as the remaining action item tickets still require heavy triage. In regard to the metric for reducing overdue tickets, often times tickets are overdue for reasons out of a team member’s control, such as waiting on a client or owner to complete a task.

Regardless, this phase of work was still critical to building out ticket automation infrastructure and building trust between the system and user, and we have a great sense of where to focus in the next phases:

• Focus on the high volume of the other types of non-actionable tickets, i.e. those that are pending other work to be done first

• Almost half of tickets are manually created, so we must continue to increase the use of ticket templates to standardize and enable system identification and automations