Understand NDIS demand in your area

The NDIS Demand Map is a new data analytics tool that provides information to help providers to grow in the NDIS.

Use the NDIS Demand Map to help understand the services and locations NDIS participants may choose by 2023.

What is the NDIS Demand Map?

The NDIS Demand Map provides an up to date forecast of the NDIS demand by postcode across Australia.

Use the map to find out:

  • How many NDIS participants are expected to live in a postcode.
  • How much participants are expected to spend, and on what types of support.
  • How many workers may be required to meet participant needs and preferences.

What are the forecasts based on?

The forecasts use de-identified NDIS and other data (census, Department of Social Services data) to predict what the NDIS will look like by 2023, when it is fully operating and the growth of participant numbers starts to ease.

The forecast will be updated quarterly based on new NDIS data as it becomes available.

It is important to note that this model is based on participant spending patterns to date. This pattern may change in the future if their preferences change.

Demand map forecasts

How should you use the forecasts?

The Demand Map provides a forecast range for future participant numbers, spending and workforce. This range reflects the inevitable uncertainty in making forecasts, particularly in areas where the NDIS has not been operating for long or areas with low populations.

Demand map screenshot 1

We expect 90% of postcodes to fall within the reported ranges; but not necessarily the midpoint of the range.

Website users should not sum categories or sum totals across postcodes, as this will produce inaccurate results.

Users should make their own assessments when using this data.

Tell us your views

We welcome any feedback you have on how we can improve the Demand Map. Please send your views through our feedback form.

Important information in interpreting forecast ranges

In interpreting forecast ranges, Demand Map users should note:

Forecast certainty

  • These ranges are forecasts for the scheme once it is fully developed (expected to be by 2023). The ranges are not a measure of actual participants in the scheme currently.
  • Forecasts, by their nature, have an element of uncertainty. The ranges reflect the uncertainty about the participants that will access the scheme, what their needs are, where they will live and what they will choose to spend on what types of support and services. This uncertainty is particularly prevalent in less populated areas, or in areas where the NDIS is not yet available or has not been operating for long. This uncertainty may reduce over time, and the ranges may narrow, as the scheme develops and more data becomes available. The forecasts do not take account of unforeseen changes e.g. sudden changes in demand due to government action
  • We expect 90% of postcodes to fall within the reported ranges; but not necessarily the midpoint of the range. This means, 10% of the time, the actual number of participants, participant spending and workforce will be different to what is predicted. It is not possible to say with certainty which postcodes will be different to what is predicted.

Low population areas

  • For regions with low populations, caution should be taken when interpreting the ranges. Where the number of participants is expected to be low, an incorrect prediction could be the difference between no spend in a postcode and substantial spend in a postcode. In these cases, confidence could be increased by combining neighbouring postcodes.
  • For postcodes with very low populations (less than 50) or small postcodes where the forecast is not reliable, forecast ranges are not provided at this stage. Forecasts may be provided in the future, as more data becomes available, subject to meeting data privacy requirements.
  • To ensure privacy, forecast ranges will not be provided for postcodes with participant and workforce populations below 10.

Adding postcodes

  • The ranges for individual postcodes will not sum to the estimated ranges if multiple postcodes are selected due to greater confidence when estimating uncertain factors for a greater number of people.
  • Website users should not sum categories or sum totals across postcodes, as this will produce inaccurate results.

What is and isn't captured in the forecasts

  • There are some disability services currently funded outside the NDIS that are expected to become part of the NDIS over the coming years. This “in-kind” support does not present an opportunity for NDIS providers at this time but may do so in the future. The forecasts assume that some of this in-kind support becomes part of the NDIS. As this in-kind support actually becomes part of the NDIS, the forecast will become more reliable.
  • The model has been adjusted to, as far as possible, capture participants choosing to live in Shared Support Accommodation (SSA). As scheme data is updated and more SSA spend is recorded in NDIS data, the market model will improve the accuracy of spending forecasts associated with SSA.

Workforce estimates

  • There is currently no actual data on the number of workers in the NDIS to inform a forecast on the workforce that will be needed in the future. To produce this forecast, the model therefore estimates the workforce through analysis of participant spending, using assumptions on the share of NDIS payments paid as labour costs. These assumptions are based on stakeholder interviews.
  • Workforce estimates are provided for a number of occupation groups. Some occupations with low forecast numbers and/or more uncertainty about the specific occupation have been grouped together into “other”. Other may include occupations such as Nurses, Social workers, Podiatrists, Human resource managers, Nutrition Professionals, manufacturers, fitness instructors.

Participant spending categories

  • Forecasts of participant spending are broken down across 12 categories. These categories group the 38 NDIS registration groups for simplicity and modelling robustness. Details on which registration group maps to which website category can be found in appendix A.


  • The forecast utilises the most recent date postcode data provided by Australia Post. For some postcodes that are situated on state borders, the Australian Post dataset does not recognise postcodes directly bordering in adjoining states. In these instances you will need to directly click on the postcode required to see NDIS demand data.

Technical information

The Demand Map uses a ‘Random Forest Regression’ statistical technique to forecast NDIS participation, spend and required workforce to deliver these services. The model draws on census data, NDIS administrative data and Department of Social Services data from over 2,500 postcodes in Australia to project the following dependent variables: number of participants in each postcode, and value of spend in each postcode (job estimates are inferred from these spending figures). Only summary statistics at a postcode level are used in modelling, such that individual data observations are not linked across sources.

To estimate the values of dependent variables, the model observes NDIS data from postcodes where the NDIS is currently rolling out and links these figures to underlying postcode demographic and aggregated Department of Social Services administrative data.

The model estimates relationships between these data sources and NDIS outcomes, by generating multiple decision tree models to assign dependent variable estimates to each postcode after considering demographic and other variables in those postcodes. The ranges produced for the model are based on the variation in predictions using these multiple decision trees.

To ensure robustness, modelling is evaluated out of sample through tenfold cross-validation. This procedure splits the available data where the NDIS is currently rolling out into ten subsamples, and then trains a model on nine subsamples, and evaluates performance against the remaining subsample. The procedure is repeated ten times, such that each subsample is tested against, and the average performance of the model is compared to a naïve mean prediction model across two criteria:

  1. Accuracy: The model trained on the subsamples is used to generate predictions for out of sample data, and the difference between true and projected dependent variables (R2 scores) is examined to evaluate accuracy. A positive score indicates the model outperforms a naïve predictor using the mean observation of the whole sample.
  2. Precision: To evaluate precision, the model uses a more complex methodology of standardised log probability. The random forest model trained on the subsamples is used to generate predictions for out of sample data. A normal distribution is then estimated around predicted values, using the standard deviation of predictions of individual decision trees that make up the forest. Standardised log probability compares the probability density of the true value for the predicted distribution, compared to probability density using a naïve mean and standard deviation of the population. If the density of a true observation in the predicted distribution is higher than density of a true observation in the naïve distribution, the model precision is better than a naïve predictor.


These forecasts have been produced by AlphaBeta Advisors Pty Ltd. Every effort has been made to provide the most current, correct and clearly expressed information possible on this site. Nonetheless, inadvertent errors can occur and applicable laws, rules and regulations may change. The information contained on this site is general and is not intended to serve as professional advice. No warranty is given in relation to the accuracy or reliability of any information. Users should not act or fail to act on the basis of information contained herein. Users should not rely on the information for any business, commercial or other purpose, and are strongly encouraged to seek professional advice concerning the information provided on this site before making any decision. Users should not rely on sum categories or sum totals across postcodes, as this will produce inaccurate results. Users should not use this tool for the purpose of re-identification. All contributors to this site disclaim all and any liability to any person or organisation in respect of anything, or in consequence of anything, done or omitted to be done by any person, organisation or other user in reliance, whether in whole or in part, upon any information contained herein.

No part of the text or graphics on this site may be reproduced or transmitted in any form or by any means, electronic or mechanical, including by photocopying, facsimile transmission, recording, re-keying or using any information, storage and retrieval system. Certain links on the site lead to resources located on servers maintained by third parties. As such, no representations are made as to the accuracy, currency or any other aspect of the information contained on such servers or the timely, accurate or complete transmittal of such information.