This document summarizes a workshop about using predictive analytics to identify people at risk of homelessness. The workshop discusses how predictive analytics works, identifying early predictors of homelessness using household data, engaging with at-risk households early through programs like in Luton and Croydon Councils, and measuring the success of prevention efforts by tracking outcomes. Attendees then broke into groups to discuss challenges and opportunities around early identification, engagement, and measuring impact.
1. Jade Alsop
Ben Fell
Policy in Practice
How predictive analytics can
help you identify people at
risk of homelessness
2. Agenda
1. Introduction
• to our workshop
• to Policy in Practice
2. Homelessness - what do we know?
3. What is predictive analytics?
4. Section 1: Identifying early predictors
5. Section 2: Early engagement
6. Section 3: Measuring and Tracking successful prevention
7. Take away action plan
5. A team of
professionals
with extensive
knowledge of the
welfare system
who are
passionate about
making social
policy work
We help local
authorities use
their household
level data to
identify
vulnerable
households,
target support
and track their
interventions
We develop
software that
engages people.
We identify the
actions people can
take to increase
their income,
lower their costs
and build their
financial
resilience
8. The triggers that immediately
precede a homelessness
application are shown in the
table opposite.
Causes include:
• An increase in demand for
affordable homes, not
matched by supply
• The impact of welfare reforms
• Personal factors that can
cause homelessness
The drivers of homelessness
9. • Around half of all homelessness
applications are found to be in
priority need
• Under the new responsibilities,
demand could double
The challenge of HRA
10. There are around 5,000 rough
sleepers in England and 8,000
across the UK.
A 15% increase on a year ago,
and the seventh year numbers
have been rising.
Rough sleeping
11. The NAO concluded that
government efforts to tackle
homelessness could not
demonstrate value for money
The government needs to:
• Evaluate effectiveness
• Help local authorities share best
practice
• Help ensure housing supply
meets housing need
• Monitor the impacts of policies
and interventions on
homelessness
NAO: A change in approach
12. 131313
Predictive analytics:
Advanced analytics which are used to make
predictions about unknown future events.
Predictive analytics uses many techniques such as data
mining, statistics, modeling, machine learning and
artificial intelligence, to analyse current data in order
to make predictions about the future.
14. How we work with household-level data
Housing Benefit / Council Tax data, household
level arrears / debt data from local authorities
Data is processed by our Benefit and
Budgeting Calculator
Detailed view of household-level financial circumstances
now and in the future
Councils identify and engage households at risk before
a crisis occurs
16. 5 minute breakout
Identifying early predictors of homelessness
1. How are you doing it now?
2. What measures are you using?
3. If you’re not doing it how could a data approach help?
18. Luton Council: context
• Luton Council has received funding for the
Homelessness Trailblazer
• As part of a service re-design, a new homelessness
prevention team has been created
• The use of household level data a key feature of
new team
• Identifying homelessness risk early through data to
drive prevention at 56+ days
• New service estimated to be three times as
effective in preventing homelessness
19. How is Luton Council using data?
• Tracking pathways into homelessness to
understand biggest local risk factors
• 78 households identified as at risk but not yet
presenting as homeless
• All households contacted early and offered
support through Early Action Network; 22 agreed
• Early support included benefits/budgeting
support, rent deposits, income maximisation etc.
• Prevention team staff designed a manual for
process post data-led identification
• Outcomes of 22 households tracked through
data to determine impact
22. Croydon Council
• Croydon’s award winning Gateway programme
focused on using data tackle hardship proactively
• As a result of the programme, 2,003 families at high
risk of homelessness have avoided being moved
into temporary accommodation.
• 217 of these were previously unemployed
households helped into stable work
• Overall, Croydon Council estimates that using the
LIFT Dashboard to guide early intervention has
achieved costs avoidance savings of over £4m in
one year
23. 5 minute breakout
Early engagement
1. What challenges have you encountered when getting people to engage?
2. What has worked for you?
26. 5 minute breakout
Measuring successful prevention
1. What impact would tracking have on how you deliver your service?
2. Have you had challenges with this?
3. Could a data approach help you overcome them?
Session outline - Discussing challenges and solutions for early prevention
Intended outcome - Come away with ideas for how to overcome these challenges
What's worked for us
- What's worked for you
DOES ANYONE HAVE ANYTHING THEY WOULD LIKE TO ADD TO THIS
Analytics – Software – Policy
A poll on a webinar of around 60 councils this is what they said
Priority for many authorities particularly due to the introduction of HRA
HRA: The Homelessness Reduction Act 2017 is one of the biggest changes to the rights of homeless people in England for 15 years. It effectively bolts two new duties to the original statutory rehousing duty:
Duty to prevent homelessness
Duty to relieve homelessness
https://assets.publishing.service.gov.uk/media/5a969da940f0b67aa5087b93/Homelessness_Code_of_Guidance.pdf
2018 stats
INCREASE AND UNKNOWN DEMAND= PREVENTION OF THOSE WE DO KNOW
NATIONAL AUDIT OFFICE
= PREVENTION BETTER THAN CURE- PREDICTIVE ANALYTICS
(million possible predictors, which to choose, what is covered by the data? what has the biggest effect?
DEMO DB- ---- IN CRISIS PRIVATE RENT
Governments may know how one policy affects many people. We can show how all policies combined affect one person.
We work with household level data from over 40 different local authorities to
Welfare reforms we model, and how accurate we are.
Traditional approach doesn’t always work.
engage around the problems that ARE relevant e.g., benefit take up, childcare, DHPs
REF TO BEN TAKE UP SCREEN/DHP SCREEN AND CALC
THEN DEMO BEN TAKE UP SCREEN DEMO
THE RESULTS FOR CROYDON AND MENTION HUBS
How do you track cases, evidence a successful service, record 'preventions' for statutory requirements of the Homelessness Reduction Act
Demo of tracking screen
Download the cohort from step 1 and put in the sankey/import pre-prepared cohort
How do you track cases, evidence a successful service, record 'preventions' for statutory requirements of the Homelessness Reduction Act
Demo of tracking screen
Download the cohort from step 1 and put in the sankey/import pre-prepared cohort