OFFICIAL FOI 25/26-0424

DOCUMENT 1

Executive brief

Ref: EC25-001758

Public data: Tables 1, 2, 3

To : Matthew Swainson

What : Requesting approval not to publish the three average payments public data tables whilst a review of their purpose is undertaken, and stronger governance controls are developed.

An alternative approach to the tables is also presented for your consideration.

Why : As DCEO you have responsibility for the public data and the controls to produce results and analysis.

Please act by : 2 June 2025, noting that the non-payments datasets have already been uploaded to the public data website.

Executive sponsors : Jessica Lilley

—|

Recommendation

That you:

Agree not to publish payments data tables 1, 2, 3 at 31 March 2025 on the Data and Research Microsite (Option 1 - preferred).

Decision: Agree to option 1
Alternatively, agree to publish tables 1 & 2, and not publish table 3 (Option 2)

Decision: N/A

Decision comments

Agreed to option 1 – To not publish payments tables and to post a holding statement on the website.

Date: 4/6/2025

Page 1 of 5

OFFICIAL

OFFICIAL FOI 25/26-0424

1. Key reasons for the recommendation

1.1. The primary reasons are:

The end use of the average payments tables by the public is not known. An internal review can establish the intention of these data sets and design tables that are fit for purpose, using feedback from external users via the Scheme Actuary inbox and in consultation with other parts of the Agency (e.g. Providers and Markets, Finance).
There is a recent history of replacing published tables. Errors have been found after investigations prompted by public questions. It is possible that the proposed redesign may necessitate another replacement of published tables, so the preferred approach is to hold off on publishing the datasets while this is worked through, rather than having multiple refreshes of the datasets.
Given the size and complexity of the public datasets, work is required to improve consistency in the calculations across each publicly available source, and to strengthen governance and controls prior to publication, to reduce the risk of errors.

There are two options proposed for Tables 1 to 3, with option 1 being preferred -

Option	Comments	Risks
1	Do not publish payments tables 1 to 3 for the Mar-25 quarter. Post a holding statement on the website	Tables 1 & 2 are published to Dec-24. Table 3 is published to quarter Jun-24 but has errors with the count of participants. A mathematical solution to the errors in Table 3 has been developed, but, due to the very granular level of the item data, the participant definition differs to that of Tables 1 and 2.
2	Publish payments tables 1 and 2 only, do not publish table 3 for Mar-25. Post a holding statement on the website	As above, it is likely that continuing to not publish Table 3 will lead to complaints.

Page 2 of 5

OFFICIAL

OFFICIAL FOI 25/26-0424

2. Background

2.1. The Data and Insights (now known as Data and Research) Microsite National Disability Insurance Scheme (NDIS) was established in 2019 as part of the commitment from the Insights forums (2019). The microsite states the Agency’s public position that “The NDIA is committed to publicly sharing NDIS data. This information can help you better support NDIS participants, identify new business opportunities, understand the NDIS marketplace around Australia, and help inform your decision making.”

2.2. Quarterly data uploads span participant demographics, plan budgets, utilisation, payments and market statistics. With the exception of payments, data as at 31 March 2025 has now been uploaded. There are currently three Payments datasets | NDIS:

Table	Average payments by:	Website includes:
1	Support class (3)	June 2023 to December 2024
2	Support category (~15)	June 2023 to December 2024
3	Line item (~1200)	December 2022 to June 2024

2.3. Approval for release of the quarterly data uploads is provided by the Branch Manager, Actuarial Insights and Monitoring. There is currently no formal business owner of public data uploads or releases (see also 2.9).

2.4. The end use of the public data is not well understood. Users of the site include:

Advocates
Consultants
Data providers
Disability service providers
Government
NDIA
Property consultants
Researchers

Given the wide range of users, there is the opportunity to reconsider the data being produced and how best to balance the needs of end users with the resources of the NDIA.

2.5. On average, these datasets are downloaded around 200 - 300 times each for December and June datasets, and less than that for the March and September quarters. Despite these low numbers, there are a small number of vocal users who regularly contact the Agency with questions and provide feedback of use and design of the data. These are primarily consultants, focused on SDA data and the NDIS market.

2.6. Each table is further subdivided by disability, age band, state and (for Table 3) service district. Table 3 has more than 500,000 rows, of which ~70% are masked due to the small sample size.

A small extract from the tables is shown below:

Page 3 of 5

OFFICIAL

OFFICIAL FOI 25/26-0424

2.7. There is a recent history of replacing published payments tables. Errors have been found after investigations prompted by public questions. When an error is identified, it will often necessitate the replacement of the entire history of the table on the data website.

2.8. There are challenges in defining the calculation of average payments over a period. The definition of participant at the class/category/line-item level is not intuitive and several different approaches for counting participants are currently used across publicly available datasets, including:

a. Persons with a transaction in the month, averaged over the months in the calculation period

b. Unique persons with a transaction over the calculation period

c. Persons with the support category (for the given class or line item) included in the budget.

There are additional considerations to the calculation, such as:

Participant counts where the person’s demographics change over the report period
Payment reversals (negative payments) and the count for a payment reversal.
Censored results at the granular cell level.

2.9. In the development of the Agency’s public data offering, the Markets team were the original business owner. However, the intended business purpose and business owner of Tables 1 to 3 has become lost over time, and the production of Tables 1 to 3 requires stronger controls on the reasonableness of the results. Work is required to re-establish an appropriate business owner, with support from ADA, across all public data, and improve governance and controls.

3. Next Steps

3.1. On approval of one of the options, the steps are:

a. Release a holding statement on the website. The draft holding statement for Option 1 is:

The NDIA is committed to releasing downloadable data that improves the understanding of the NDIS. A review is being undertaken on the format of the payments datasets, and it is anticipated that releasing payments datasets will recommence in early 2025-26.

b. (If Option 2 selected) Publish Tables 1 and 2

c. Led by ADA:

Page 4 of 5

OFFICIAL

OFFICIAL FOI 25/26-0424

Consult across the NDIA on the purpose of the data and establish the business owner
Establish the report owner within ADA for reporting on payments by class / category / line item
Agree the purpose of the tables, being guided by the business owner and by historical complaints and requests
Prepare tables that meet the agreed purpose
Develop a quarterly governance process that includes appropriate checks, and incorporates continual feedback via the senior data officer.

4. Attachments

Nil

Page 5 of 5

OFFICIAL