We get a fair amount of feedback that home care isn’t moving boxes (a reference to Amazon). While that's true, obsessing around the customer experience is actually an important part of what employees learn at Amazon. They do this through obsessing around the metrics that matter to customers and coming into work every single day to improve these metrics.
The methodologies described below work and they are a significant part of what has allowed Amazon to become the company it is today, one that delivers millions of packages every single day throughout the world, while meeting the highest customer expectations. It is important to understand that this type of work takes time as there normally isn’t a single solution or a silver bullet. Long term solutions require time and it is the summation of structural solutions to problems that creates durable long term change. And that is what we are all about doing.
Technology is how we will create these long-term structural solutions that allow us to achieve scale while improving the quality of our service. People are at best 3 sigma machines1, so if you string together a series of human processes, they deteriorate quickly - three human-based processes strung together equates to “success” of 81.2% or an almost 20% failure rate. Machine-based processes are consistent, and this is a massive benefit. This consistency allows you to “count” defects, identify the underlying root cause of the defects and then fix the root cause, which allows one to improve quality durably overtime.
The defect framework
During our time together in Denver in September, we reviewed the work we have been doing to measure the client experience through the lens of experiences that don’t meet customer expectations, which we’ll refer to as “defects.” Defects are experiences we can measure with data that show a strong connection to the client’s duration on our service.
After looking at over 20 defects (e.g., call-offs, late arrivals, no call no shows), we identified the 3 most impactful to client retention as:
Unique Care Pros: the # of unique Care Pros visiting the home, often called “consistency”
Visit Cancellations: the rate of canceled hours due to staffing reasons relative to total hours
Unique Honor Reps: the number of different reps the client interacts with (we assume this also extends to different people on the agency side, but we don’t have visibility into these things)
The defects we identified in the data are very aligned with the feedback we have heard from you, from our care team, and directly from clients. But, what’s different is that we can now measure these experiences with a lot of rigor, so we know for any client at any moment in time exactly the state of their consistency, for example. In addition to the presence or absence, we can also understand the magnitude of the impact and the intersections of the defects. We are very excited to move from anecdote to structured and programmatic measurement. Now every single client at any moment in time can be assessed. For the avoidance of doubt, this doesn’t mean that there are not other things that we can work on; we are simply putting a stake in the ground that says “these are the ‘problems to solve’.”
Additionally, now we can measure how changes to the way we work improve the metrics that matter. It also allows us to optimize the system to minimize the occurrence of these defects. For example, our staffing algorithms can be optimized to maximize consistency for all clients in the market.
The Care Platform teams’ (this includes the data science, engineering as well as care delivery teams) top priority is to reduce the rate at which clients turnover during start of care (SOC), specifically the first 4 weeks of the client’s journey, when the data shows they are particularly sensitive to these defects. And our primary problem to solve to achieve that reduction is improving consistency from day 1 of the client’s journey on the platform.
It's important to understand that although we aspire to be perfect, even in the most controlled environments perfection is impossible - but these methodologies allow us to continue to improve every single day and gives us a common language and understanding of the opportunities that sit in front of us. So even as we aspire to get to perfect consistency, life still happens and issues still come up - clients’ conditions change, Care Pros’ life circumstances change. So our secondary problem to solve is the customer service experience, as measured in our defect framework by defect #3 - the number of Honor reps with which a client interacts.
How we are working
We want to share exactly how we are approaching solving these two problems, as it’s very different from how we’ve solved these problems in the past:
First, we are partnering really really deeply with several owners across several markets to understand what is happening on the ground at a very detailed level and working with them on developing solutions to reduce the occurrence of these defects. Over the course of the last two months, as an executive team (Ian, Juliet and Tamar), we are meeting weekly and in some cases daily with owners in 4 markets - Phoenix, Raleigh, San Antonio and El Paso to review every start of care, every service failure. Our technology team, the people who build the system that enables the start of care experience, are also partnering really deeply with these owners, specifically in El Paso and Raleigh for now, to focus intensely on these problems.
Second, we are taking bigger swings. We aren’t looking for incremental improvements to how we do things today. We are looking at fundamentally new and different approaches (more to come on this below).
Through our experience so far spending time with this group of owners, we have formed hypotheses (or ideas) about where we need to take some big swings and are in the process of testing/validating new approaches. We often refer to this approach as a “pilot.” A pilot is simply a way to test and validate a hypothesis around a new way of doing things, but doing it with just a small subset of your customers/users. It allows you to work quickly to validate the new approach and figure out what the system/technology needs to do to support. We often do this in what we refer to as “scrappy” ways. Then once we know something is working, we invest in robust software, team and process solutions that can scale and be operationalized in all 25 markets.
This is really important, because, as a platform that works across ~25 markets, and hopefully soon 50 and 100, to drive any type of fundamental change in how we do something, we need the technology to enhance the work of our people. And giving our technologists a front row seat working with owners and being involved in the day-to-day delivery of care is what allows them to viscerally understand the problem and creatively think about how technology can solve it.
One of the things to keep in mind is that our long-term goals haven’t changed. We want to build a platform that allows us to facilitate all of the products and services that are desired for an aging population to stay in their homes, but the order of operations is important. We aspire to first be exceptional at providing the service we do today, and then on top of this solid foundation move on from there.
So let’s talk about the hypotheses we are actively testing as pilots in Raleigh and El Paso:
Improving consistency
Care Pro (CP) Engagement Model
We often talk about how staffing is a constant game of tetris. And as we’ve shared, we’ve invested this year in building out an optimization model that at any moment recommends the ideal Care Pro (CP):client pairs at a market level that deliver consistency for our clients. But what we learned in the process of sending these recommendations out to CPs in the form of job offers through the CP app was that our acceptance rate of these job offers was very low. How can we create consistency if we can’t rely on our CPs to take the work we offer to them?
So we spent time interviewing and learning from owners both on and off platform about how you’ve historically approached your engagement model with Care Pros and how that ties to creating consistency for clients - everything from how you set expectations and enforce accountability, how you build relationships, how you staff.
And through this learning journey, we formed the hypothesis that if we adopted an approach where expectations are clearer and work is more or less assigned to CPs based on what we know about them, we could significantly improve consistency for clients. But first we needed to test out and learn how to operate2 in this new way that’s different from the current CP experience on platform (for example, today Care Pros can see all jobs they are eligible for and choose from that selection - this often leaves us with the hardest to staff schedules remaining open). We’ve partnered with the Pattons in El Paso, who were just joining the platform in October, to test out this new approach.
So specifically, what we are doing differently in El Paso compared to the rest of the platform:
Curated job offers: We have turned off all job offers in the CP app. While we have always had a curated, interactive component to our staffing, in this market we are offering work exclusively via phone calls and texts based on the CP's performance rating, availability and preferences based on the need of the market and our clients.
Accountability to expectations: We are setting the expectation that CPs accept work that’s offered to them within their availability and that repeatedly declining work will impact their eligibility for future work. So far, we are seeing substantially higher visit acceptance rates. We are also setting the expectation that they provide at least 2 weeks notice for any planned time off, and that last minute call offs that are not related to illness or other protected reasons will count against their performance scores.
Rewarding performance. We are ensuring that the highest performing CPs are recognized and prioritized in scheduling. We are factoring in client feedback on their delivery of care in the home into this prioritization.
We are still only a couple weeks into this pilot but are encouraged by the results so far, with consistency remaining stable post-IMPL in El Paso.
Care Pro Hiring
We operate in constantly changing external dynamics, which require that we be able to respond quickly to new circumstances. We see in our data and have validated through your experience on the ground that consistency suffers when we simply are undersupplied with Care Pros relative to client demand. In Raleigh and El Paso, we are testing some new approaches to Care Pro hiring to dramatically increase the volume of new CPs hired every week without sacrificing the quality of our hires. These approaches are focused on reducing time from application to first visit, and increasing the conversion rate of applicants to first visit.
Not every applicant is equal in their contribution to solving a supply problem in a market. Sometimes you are tight on Care Pros who can reasonably commute to a certain area, or tight on CPs who can do overnights. We’ve developed an applicant prioritization score that measures the source of our “undersupply” in a market. This score, which is automatically calculated based on the data in our system, like open shift rates and clients’ consistency rates, allows us to target the right candidates - fast track and put focus on converting candidates that solve our specific source of undersupply, customize our ad targeting to these specific needs3.
As we shared in Denver, we recently implemented the Fountain system, which is an applicant tracking software, to power our Care Pro recruiting. This system is very flexible, candidate friendly and allows for a lot of automation. In Raleigh, within a matter of a few days, we were able to set up an entirely new candidate flow in the system, integrate the new applicant prioritization score into this flow, and start funneling candidates through this new experience. Our team has regularly been flying out to Raleigh to spend time with Stephen Lair and his team to problem solve this together. We expect the pace of iteration on hiring practices to be high and as soon as we have validated these new approaches work in reducing time to hire and improving conversion rate, we will bring them to more markets.
Staffing Tools
We are rapidly iterating on tools that allow us to more easily play that tetris game behind the scenes to identify the optimal staffing scheme that delivers consistency to our clients. We loved hearing Debbie Gross’s analogy recently to staffing as the exercise of matching pairs of socks. These new tools, including our optimization algorithm, enable us to do things like more easily identify CPs for new SOC clients, to easily identify restaffing opportunities (reassemble the mismatched socks), to find more hours for our top CPs, etc. Using these tools, we are generating these staffing solutions and reviewing them daily with the Patton and Lair teams to get their feedback, and integrate their deep knowledge of the clients and CPs into staffing decisions.
We appreciate everything we have learned from the Patton and Lair teams!
Improving customer service
As we saw in the defect analysis, one aspect of the client experience that influences the client’s likelihood of turnover is their customer service experience. The best measure of this to date (though we expect to refine the measurement over time), is the # of different honor reps the client interacts with (defect #3). This was validated in the client research study4 we conducted earlier this year. Clients shared dissatisfaction with interacting with multiple different people to resolve their issues, repeating information and not receiving the personalization for which they were looking.
Kelly Cornelius and Mike Melinger have been passionate advocates around the opportunity to improve how we communicate with our clients and deliver exceptional customer service. A few weeks ago, we (Ian, Juliet and Tamar) got together with Kelly and Mike to map out, end-to-end, how they have traditionally approached engaging clients. We were particularly inspired by Kelly’s team’s mantra to be singularly focused on “anticipating and delighting our clients.”
Even when a client’s consistency is great and they love their CPs, “stuff” still happens. A CP could have a last minute emergency and call off a visit, a CP could resign, the client’s condition could change or the client could experience an event that results in hospitalization. And for those clients where we don’t initially achieve consistency or they don’t love one of their CPs, we can mitigate the impact with great customer service. How we handle these situations when they do come up can make all the difference with our clients.
We are currently in the process of designing a pilot that we’ll launch before the end of the year with Kelly and Mike where we take a dramatically different approach to client management. The details are still being worked out, but in concept we will be experimenting with how the local office can take the lead on managing their clients’ experience.
In parallel to testing this new way of working with Kelly and Mike, we are developing predictive tools that can be utilized across all owners on the platform to anticipate and get ahead of potential client turnover. What’s great about the Care Platform is that we “observe” nearly everything that’s happening with our clients as it’s all logged in our system. So we are taking all the various sources of data we have on each client, including…
Their CP consistency
History of recent cancellations
Volume and frequency of customer service touchpoints, number of reps they have interacted with
Escalations or other acute events
Analysis of their sentiment based on text and phone interactions that’s performed using GPT
Recent CP performance events associated with the client
… and now feeding this data into a client turnover prediction model. What this model then allows us to do is create nudges for our team and owners on the ground to intervene before it’s too late and save clients from leaving our service. We’ve been working on this for ~2 weeks and are now generating a daily report that highlights clients predicted to be at risk of turnover with all the detail on the recent events that put them at higher risk. We are still refining this report with a subset of owners and plan to make it broadly in the next couple months.
Other improvements across the platform
Through the recent time we’ve been spending working closely with owners, we’ve also been able to identify and deliver some additional improvements that have benefited all Care Platform owners:
LTCI Notifications
As shared previously, clients now receive automated notifications every time we submit an LTCI claim on their behalf, with the claim itself attached to the email.
Requiring outreach to team to cancel visits
A small portion of client visit cancellations are made in the family app. We’ve heard feedback from you that when AOs cancel visits from the app, we are often surprised and don’t have the opportunity to understand the why and address the root cause. So we have disabled the ability for families to cancel visits in favor of a phone call to our team.
Owner feedback on CPs
You and your teams are such a valuable source of feedback on how our CPs are performing as you interact with them yourselves, or receive feedback from clients on their experience. As of this week, you now have the ability to submit CP feedback directly into the system via portal. We’ll share more information on this in the coming week.
Improving CP Quality
As shared in Denver, we are raising our standards around CP performance. We’ve already implemented a number of systematic changes to accomplish this and others are in development:
(Done) We are executing on a stricter policy for no call no shows (NCNS). Any CP that no call no shows will be terminated, barring exceptions for extenuating or protected circumstances, while also taking into account the Care Pro's tenure and prior record.
(Done) We did a one-time review of CP performance as it relates to reliability (attendance) and removed ~6% of CPs who were most unreliable on the platform according to a new performance scoring.
(Done) We are taking swift action on Care Pros who prove to be unreliable. The system now scores every Care Pro on their reliability, as measured by their attendance record (<72 hour notice call offs and late arrivals). Care Pros who are highly unreliable are prevented from getting staffed on any new shifts and we are manually reviewing these scenarios to determine whether they are eligible to continue working for Honor or not.
(In development) We are in the process of developing a score that evaluates CP performance on the dimension of client satisfaction (or ability to deliver great care to clients). We have all experienced that some CPs have high client skills but are not necessarily reliable, so it’s important to separate these measures of CP performance and weigh them appropriately. We are leveraging all the data sources we have on CP feedback - from you, from clients directly via surveys, calls, text messages, to develop this relatability score. This is where providing CP feedback from the portal will be really helpful, as it is another important source of data to feed into this score. Once this is in place, we will take similar system-driven actions to start, to prevent those who demonstrate poor relatability to continue working with clients. We will also utilize this score, together with the reliability score, to ensure that systematically, we are fully utilizing our highest performing Care Pros and getting them the hours they want, and that we are retaining them at a high rate.
(In development) We are testing additional visit confirmation outreach - for example, testing in-app confirmations for new clients and testing weekly schedule confirmations, to mitigate the risk of NCNSs and last minute call offs.
1This is a quality measurement, stolen from 6 Sigma. 3-sigma is 93.3%, which means that a process is defect free 93.3% vs a 6-sigma process which generates 3.4 defects per million (defect free 99.99966% of the time).
2This is probably the most critical thing for everyone to understand: Although there might be evidence that a specific solution worked, we need to understand if that solution will work in the context of a scaled solution. For context, we have over 4K Care Pros on the platform today (~20x a typical agency), the point being to effect a change that works at scale, for everyone on the Care Platform and that is durable and repeatable.
3The larger point of these isn’t that any of these are new ideas (or particularly insightful) but they need to be implemented for scale. For context, on any given week we get over 3.5K started Care Pro applicants.
4We interviewed 26 clients, mostly still active on the platform, to understand the good and the bad of their experience on the honor platform.