Mable payments

Speeding up payment timeframes and reducing invoice rejections

Mable is reimagining aged care and disability support for Australians.

Challenge

Improve the payment experience for all stakeholders processing claims for NDIS plan-managed clients on Mable.

Objectives

  • Ensure support workers on Mable set rates that are under the appropriate price limit for the service they provide.

  • Allocate accurate support items to hours on support worker invoices sent to NDIS plan managers. 

  • Increase the speed of payments for workers.

  • Reduce the number of invoices rejected by plan managers.

My role

Product Design Lead

Duration

8 months (2023)

Team

The Payments Squad: Product Manager, Engineering Manager, Data Analyst, and a delivery team of engineers and QA.

Taking on a complex design challenge

I joined the Payments squad with the daunting realisation that previous designers who had attempted to solve this challenge had parked it due to the uncovered complexity.

Why is it so complex?

Support items are chargeable codes that can be invoiced to the NDIS. There are over 840 codes, and NDIS participants only have some codes available specific to their NDIS plan. The most common types of support overlap many different codes.

This problem is unique to Mable as an unregistered provider. Registered providers can access this information through the NDIA Portal, apps or APIs.

Prototype discovery testing with clients

Could clients select the charge categories?

In short, no. We discovered that most clients are unfamiliar with the budget they have available or how it works. Clients who have tried to understand their budget have found the NDIS rules complex and overwhelming. It’s a big reason this cohort use plan managers. 

Image: screens from a prototype I created to test with Mable clients

“I feel physically sick just seeing this question. I’ve put so much energy and effort into trying to understand these categories. I would have no idea what to select here, and I would be so so scared of getting it wrong and affecting my budget.”

— Research participant, March 2023

Pivot number 1: Could plan managers share their client’s budget information on their behalf?

Testing our riskiest assumptions: clients will consent for their plan managers to share their budget information with Mable.

We found that clients were reasonably comfortable with the concept of Mable contacting their plan manager for budget category information.

Image: screens from a prototype I created to test whether clients would be comfortable giving consent for us to contact their plan manager and request budget information

Testing our riskiest assumptions: plan managers will be willing to do the upfront effort to provide client details.

Yes, but with mounting technical and experience complexity requirements. Interviews revealed that Mable would not be able to facilitate the request. A client would need to take proactive action with informed consent, and we would need a secure portal for document upload.

Pivot number 2: How accurately could Mable determine the codes with our existing data?

Asking the right questions:

  • An interview with our finance Head of Collections uncovered that ~20% of rejected invoices were for allied health services. The NDIS support items for these services are one-to-one, making it simpler for Mable to use logic to determine the service based on the worker's qualifications and client service request.

  • The NDIS released an industry quarterly report detailing how funds were being allocated to providers. Some key statistics built confidence in some bets we could make:

    • 78% of all NDIS budget allocation is core category (AKA the flexible funds), across the whole scheme.

    • The report showed that unregistered providers like Mable provide services for core funding around 99% of the time.

Defining our vision (and our bet)

We can allocate codes for all of Mable’s plan-managed NDIS invoices automatically, with at least 95% accuracy.

Once we learn the volume of rejections from automated codes, we take a reactive approach to specialised support. Our bet is that rejections will be less than 5% and therefore not increase the workload of collections teams, while improving 95% of cases.

Redesigning the rates UX, logic, and structured data

We created new logic to default services to NDIS codes based on existing job and worker data. We updated the rates input to structure the data around the days and times to help infer the correct price limits for rates earlier in the customer journey. Our principle was that workers don’t need to understand NDIS support items, but they can access and edit them if they would like to.

Image: an example flow documenting a worker's default rates created from their profile then edited for one client

Setting defaults for workers, and an escape hatch to request specialised support types.

Our escape hatch went to a Qualtrics form. This was a hack approach to test and learn the size of the problem. We tracked accordion and link clicks to collect a signal, plus captured and actioned survey service requests.

Image: Painted door inspired interaction to collect a signal on the specialised service problem size

Getting workers to keep rates under the price limits

We do not prevent workers from setting any rate amount because they are independent contractors. But, if they submit hours for NDIS clients that are over a price limit then those hours will not be paid. This is a huge pain point for affected workers and takes a lot of Mable resources to try to rectify.

4-5% of plan-managed invoices were being rejected for being over a price limit, and there was a risk this problem would scale with support items added to our invoices.

Image: Rate guidance Info alert banner

Iterating the alert banner for impact

We put an alert banner in (figure 1) and were surprised to find it only had a 5% impact on people adjusting their rates. So, we included the banner in a quick round of testing and discovered some key issues:

  1. People were reluctant to head to an external website (confirmed in our click tracking).

  2. The NDIS information was confusing, and workers often misunderstood the rules and came back confident that they could charge the rate they had nominated.

Iteration (figure 2): we updated our banner component to handle multiple accordions. We believed that if we could give workers more specific information while in the context of their rates, they would be more likely to take action. The updated banner performed more than 10x better, resulting in 54% of workers updating their rates.

Figure 1

Figure 2

Images: The evolution of our mobile alert banner to increase impact

Support items released onto 100% of invoices

The initiative had a huge impact on streamlining payments, improving external stakeholder relationships with plan managers, and positively impacting all the metrics we targeted.

Some headline metric outcomes

  • Support codes onto 100% of invoices

    Support item codes onto all plan-managed invoices, with less than 2% rejected for an incorrect support code.

  • Rejections reduced by nearly 60%

    The overall invoice rejection rate went down, and rejected invoices are quicker and simpler to rectify.

  • Payments cases reduced by 44%

    Cases that had to be managed by an operations team member reduced.

  • Call volume reduced by over 30%

    Phone call volumes to our operations team were reduced because of fewer calls about payment issues.