Problems

Brief DA: Data Freestyle

Your call: no starter data required.

Read the Data Asset Register, or otherwise, pick a dataset and use it to build a model (a classifier or regressor, using predictive ML to extract trends from data), visualise a map, or tell a story.

Brief DB: Where should London actually build?

The question: demand for housing isn't evenly spread across London, and neither is supply. Can you make the mismatch legible, and actionable, for someone who isn't an economist?

Why it matters: the Warwick WhereToBuild work puts a number on this (demand outstrips supply roughly 5:1 in places), based on Rightmove search and listing data. That's a strong headline, but a map alone doesn't tell a councillor or a developer what to do with it. There's also a custom WhereToBuild-style extract already in the data warehouse, worth cross-checking against the original.

Starter data: WhereToBuild (Warwick CAGE) plus the custom WhereToBuild-style extract already in the data warehouse, GLA residential completions and pipeline dashboards, Census 2021 household data, borough local plan housing targets.

Possible directions: a borough-level scorecard comparing demand pressure to delivery rates; a tool that overlays demand hotspots with sites that already have planning permission but haven't started; a plain-English explainer of why a specific ward is under-supplied.

Audience: policy campaigners, local councillors.

Heads up: if you're using WhereToBuild's bespoke data for this hackathon, there is a data use agreement in the drive. The most salient point is to avoid uploading records publicly or sharing outside the event. Speak to the hackathon organisers if unsure about data governance for this dataset.

Brief DC: Can you build the planning-data pipeline London doesn't have?

Jam's white whale: ambitious, but a working prototype counts as a win.

The question: PlanX/BOPS is submission-only, planning.data.gov.uk's bulk export is real but thin (six data providers, stale since September 2025), and UK PlanIt aggregates all 33 London boroughs without ever turning that into a bulk download. Can you architect the pipeline that finally produces one clean, bulk-exportable, London-wide planning applications dataset?

Why it matters: this is the blocker sitting behind half the other data briefs here. Every "predict or analyse planning outcomes" idea eventually hits the fact that nobody's assembled the underlying dataset. Solve it well and the output isn't just a hackathon demo, it's something #1 (and every civic-tech project after it) can actually reuse.

Starter data: UK PlanIt (planit.org.uk, the closest thing to a London-wide index, built and maintained by one person); planning.data.gov.uk's planning-application dataset (a working example of the target schema and output format); the Open Digital Planning digital-planning-data-schemas repo (target shape for individual applications).

Possible directions: a rate-limit-respecting scraper or ETL against planit.org.uk with caching, so you're not re-fetching what you already have; an ML or LLM classifier that filters the output down to housing-relevant applications, the same relevance-filtering problem solved by hand for the house repo (see Brief DD); ship a documented, queryable output (CSV, Parquet or DuckDB), not a raw HTML dump. A realistic scope for one day is a working pipeline against one or two boroughs end to end, not all 33.

Audience: data engineers, future House London organisers, the wider open planning-data community.

Heads up: UK PlanIt is run by one volunteer (Andrew Speakman) and it rate-limits. Lean on backoff and caching rather than proxy rotation, and check the terms of use before scraping at volume.

Brief DD: What does the approvals data actually say?

Handle with care: taking it further than the original scrape did.

The question: planning outcomes look inconsistent: similar-looking applications get different results depending on borough, committee or officer. Can a model predict, or at least characterise, that variation?

Why it matters: planning consultations are dialogic, not mechanistic, so treating this as a simple rulebook tends to fail. That doesn't mean outcomes are unpredictable in aggregate, though, and this is one brief where the hard data-acquisition problem is already solved: roughly 350,000 scraped and relevance-filtered housing planning applications already exist, probably the largest dataset of its kind outside a local authority. The job here isn't finding the data, it's doing something sharper with it than there was time for originally.

Starter data: the pre-scraped dataset (around 350,000 housing-relevant planning applications, relevance-filtered, available as-is from the shared Google Drive).

Possible directions: a classifier trained on outcomes for one borough, tested for transfer to another; a simple similarity search ("here's what happened to applications like yours"); an analysis of approval-rate variance across committees, controlling for application type.

Audience: developers, planning consultants, planning officers, policy campaigners.

Heads up: this is a self-collected scrape, not an official release, so expect inconsistent fields between boroughs.

Brief DE: Who's sitting on land that could be homes?

Handle with care.

The question: permissions are granted, but homes don't always follow. Land banking is one of the more politically loaded explanations. Can you find and characterise the gap between "permitted" and "built" with data rather than assertion?

Why it matters: the lag between a permission being granted and homes actually completing is one of the concrete, measurable levers in the "add to housing stock" policy landscape, distinct from, and less politically contentious than, land value tax debates about why it happens.

Starter data: London Development Database export (scheme-level history, frozen at a 2020 extract), GLA starts, completions and pipeline dashboards, brownfield land registers.

Possible directions: a "stalled sites" tracker showing time-since-permission for undeveloped sites; a distribution of permission-to-start lag times by site size or borough; flagging sites that have had permission renewed or resubmitted multiple times without building starting.

Audience: policy campaigners, journalists.

Brief DF: What happens after the outline application?

The question: a single outline application can spawn twenty or more related applications through post-determination amendments. What does that cascade actually look like, and does it slow delivery down?

Why it matters: this is one of the more underexplored parts of the planning process. Most public attention goes to the headline decision, not what happens in the following eighteen months of amendments, reserved matters and conditions discharge.

Starter data: Digital Planning Register data (application relationships and references), planning.data.gov.uk planning-application-document and planning-application-type datasets.

Possible directions: a network visualisation of one large scheme's application family tree; a measure of "amendment burden" per scheme; a comparison of cascade length between boroughs.

Audience: planning officers, national government (process reform angle).

Brief DG: Is anyone actually using this house?

Handle with care.

The question: empty homes and under-occupied stock are one of the "redistribute existing stock" levers in the policy landscape. Can open data actually find them, at any useful level of confidence?

Why it matters: this is one of the more contested asks (clearing empty homes) but also one of the more concretely measurable. EPC and Land Registry data leave a trail.

Starter data: EPC register (assessment dates as a usage proxy), Price Paid Data (transaction gaps), Census 2021 vacancy and tenure data.

Possible directions: a borough-level estimate of likely-vacant stock; a check on whether EPC-based proxies for vacancy roughly agree with Census figures; honest treatment of the false-positive problem, since a home with no recent EPC isn't necessarily empty.

Audience: policy campaigners, local councils.

Brief DH: Does "barriers to housing" mean something Census doesn't already capture?

The question: the English Indices of Deprivation 2025 has a "Barriers to Housing and Services" domain, brand new as of October 2025 and effectively untested against other need measures. Does it track with Census 2021 overcrowding, or is it catching something different?

Why it matters: if it just restates overcrowding, it's redundant context. If it diverges, that's a genuinely new lens on housing need that nobody's really used yet.

Starter data: English Indices of Deprivation 2025 (Barriers to Housing and Services domain, LSOA-level); ONS Census 2021 (household composition and overcrowding); Housing in London annual report (GLA) for context.

Possible directions: an LSOA-level scatter of the Barriers domain against Census overcrowding, flagging areas that diverge; a plain-English explainer of what's actually inside the Barriers domain for boroughs that score badly on one measure but not the other.

Audience: policy campaigners, local councillors.

Heads up: this domain isn't directly comparable to the 2019 Indices release, since the methodology changed, so don't try to show "change since 2019" on it specifically.

Brief DI: Do London's two affordable-housing datasets agree with each other?

The question: the GLA publishes two datasets that both claim to measure affordable housing delivery, and they don't obviously reconcile: the one-off Affordable Housing Open Data snapshot (2015 to 2021, broader coverage, frozen) and the GLA Affordable Housing Programme Outturn (quarterly, current to March 2026, but GLA-funded schemes only). Do they tell the same story where they overlap?

Why it matters: the Housing Delivery Test carries real statutory consequences. Under 95% delivery triggers an action plan, under 85% adds a buffer requirement, under 75% triggers a presumption in favour of development. But its latest measurement (2023) was only published in December 2024, so it's already lagging. If the two affordable-housing datasets diverge, that's a live methodology finding, not just a data-cleaning problem.

Starter data: Affordable Housing Open Data (GLA/DLUHC, 2015 to 2021); GLA Affordable Housing Programme Outturn (quarterly to March 2026); Housing Delivery Test (MHCLG, latest 2023).

Possible directions: a borough-level comparison of affordable share in the overlapping years between the two datasets; flag which boroughs were closest to a stricter Housing Delivery Test threshold band and check whether more recent GLA Datahub starts and completions suggest they've since crossed it.

Audience: policy campaigners, journalists.

Brief DJ: Are planning delays about the borough, or the workload?

The question: MHCLG's live tables carry statutory timeliness measures (P152/P154). Which boroughs miss decision deadlines most, and is that about the borough itself, or just how many applications it's handling?

Why it matters: "slow planning department" is a common complaint, but it might just be a volume problem, which points to a completely different fix: resourcing, not process reform.

Starter data: live tables on planning application statistics (MHCLG, P152/P154 timeliness, quarterly); UK PlanIt (application volume, cross-check); planning.data.gov.uk (constraint layers: green belt, conservation areas, article 4, flood risk, listed buildings).

Possible directions: a scatter of application volume against timeliness by borough; for the worst Housing Delivery Test performers, how much of their undeveloped land is actually constrained by a designation versus genuinely buildable.

Audience: planning officers, national government (process reform angle).

Brief DK: Does the empty-homes headline survive contact with the official count?

The question: MHCLG's Live Table 615 already gives an official empty-homes figure: London's up 81% since 2016, to 105,138 in 2025. Brief DG asks teams to estimate vacancy from EPC, Price Paid and Census proxies. Do those proxies actually agree with the official Council Tax-based count?

Why it matters: if the proxy methods and the official series disagree, that's worth knowing before anyone builds a policy pitch on either one.

Starter data: vacant dwellings Live Table 615 (MHCLG, Council Tax records, annual to 2025); Domestic EPC register; Price Paid Data; Census 2021 vacancy and tenure.

Possible directions: a borough-level comparison of Live Table 615 against an EPC/Price-Paid-based proxy estimate; characterise the direction and size of any gap.

Audience: policy campaigners, local councils.

Brief DL: Is overseas-owned land actually the land that's stalling?

Handle with care.

The question: OCOD/CCOD shows roughly 250,000 property titles owned by overseas companies. Cross-referenced with London Plan Opportunity Areas and brownfield land registers, how much of the undeveloped land inside Opportunity Areas is overseas-company-owned, and does that ownership structure correlate with permission-granted-but-not-started sites?

Why it matters: this sharpens Brief DE's land-banking question by isolating one specific, testable variable, ownership structure, rather than asserting a cause.

Starter data: OCOD/CCOD (HM Land Registry, corporate ownership only, no private individuals); London Plan Opportunity Areas; brownfield land registers (via planning.data.gov.uk).

Possible directions: an Opportunity-Area-level breakdown of overseas versus UK-company ownership on undeveloped or brownfield parcels; a check on whether overseas-owned stalled sites look different (older permission, more resubmissions) from UK-owned ones.

Audience: policy campaigners, journalists.

Heads up: OCOD/CCOD only covers corporate owners. It cannot show individual or personal ownership, and doesn't reveal beneficial ownership without linking to Companies House or the Register of Overseas Entities.

Brief PA: Freestyle

Follow your own curiosity. Think up a new policy from scratch, and support or disconfirm the effectiveness of the policy using the existing data.

Brief PB: Answer a live consultation

Real deadline. Needs a London lens.

Three UK government consultations touch on housing right now, closing within weeks of the event. Teams write evidence toward a real, submittable response instead of a hypothetical question.

Option 1: Quid pro quo leases and the ground rent cap

MHCLG, closes 27 August 2026. The government is capping ground rents at £250 a year (a peppercorn after 40 years), and consulting on exempting "quid pro quo" leases, where a leaseholder took a higher ground rent for a lower premium. It's explicitly seeking evidence on how common genuine quid pro quo arrangements are.

• How prevalent are leasehold titles with variable or escalating ground rent terms in London, using HM Land Registry OCOD/CCOD data?
• Can ownership and price records show whether a premium-for-ground-rent trade-off is real, or is this simply not answerable from open data?
• What would a London leaseholder actually gain or lose under each version of the exemption?

Option 2: More and better homes

RSH, closes 30 September 2026. RSH is rethinking social landlord regulation, shifting toward "increasingly targeted" oversight that uses data and modelling to spot risk and under-delivery: a direct hook for a data-first submission.

• Which London social landlords look highest-risk on a combination of stock condition (EPC data) and delivery against targets (Housing Delivery Test)?
• Where would data-led, targeted regulation change the picture compared to today's approach?
• What data gaps would make this kind of targeting hard to do reliably?

Suggested datasets (Options 1 & 2): HM Land Registry OCOD/CCOD; get-energy-performance-data.communities.gov.uk; Housing Delivery Test; English Indices of Deprivation 2025.

Option 3: Heathrow Expansion National Policy Statement

DfT, closes 1 September 2026. Buried in an aviation consultation is a real housing story that cuts both ways. The government's ANPS review recommends renaming it the Heathrow Expansion National Policy Statement (HENPS), covering the Northwest Runway, terminals and surface access.

Loss side: independent body CISHA puts homes at demolition risk at around 750, with around 5,500 properties in the wider impact zone (older campaign estimates for the full-scheme runway ran to 3,750), and finds Heathrow's compensation lags HS2 and Crossrail.

Growth side: Frontier Economics estimates +0.43% GDP by 2050, on top of the 80,000 jobs Heathrow already supports, with 60% of the benefit landing outside London and the South East.

Missing from the debate: if expansion creates local jobs, is anyone building homes for those workers near Heathrow, or does supply-constrained west London just eat the gain in higher prices and longer commutes? Already hyper-local (Hillingdon and neighbours), so no scoping needed.

• Which figure fits the current draft HENPS, CISHA's ~750 or the older ~3,750, and what explains the gap: downsizing, different assumptions, or stale campaign data?
• Using HM Land Registry price data, does CISHA's "prolonged limbo" show up in sale prices or time-on-market across the 5,500-property impact zone, versus similar areas further out?
• How does Heathrow's proposed 125%-of-market-value compensation compare with what HS2 and Crossrail actually paid? Does CISHA's criticism hold up?
• If expansion creates real local jobs (Heathrow's claim, or the rival "just displaces jobs from other regions" critique), does the Housing Delivery Test / Opportunity Areas pipeline in Hillingdon, Hounslow and Slough come close to housing that demand, or does growth just add pressure?

Suggested datasets (Option 3): HM Land Registry price paid data; House Price per Square Metre dataset; UK PlanIt planning application data; Housing Delivery Test; London Plan Opportunity Areas; London Datastore borough boundaries (Hillingdon, Hounslow).

Brief PC: Is "not enough supply" really the story?

Handle with care.

The default explanation for London's housing crisis is undersupply. That's contested: a December 2025 LSE blog (Hilber & Mense) argues the price-to-rent ratio doesn't support a pure supply-shortage story, while Centre for Cities' London-Paris comparison and IPPR's housing commission argue supply constraint is exactly why London diverges from comparable cities.

• Using London Datastore completions and permissions data alongside price and rent data since the mid-1990s, does the trend fit a supply-shortage story better than alternatives (financialisation, interest rates, demand-side pressure)?
• Where does London diverge most from the rest of England, on supply metrics, or on price/rent metrics?
• What would have to be true for the supply-shortage story to be wrong, and does the data available this weekend let you test it?

Suggested datasets: London Datastore housing completions and permissions; HM Land Registry price data; House Price per Square Metre dataset; Housing Delivery Test.

Data-literacy footnote: a 2019 Missing Numbers investigation flagged that the UK had no good open rent data, only a relative index, not what tenants actually pay. ONS partly fixed this with the Price Index of Private Rents (PIPR, launched March 2024), giving local-authority rent estimates for England and Wales. Sanity-check whether PIPR holds up at London borough level before leaning on it.

Brief PD: Is temporary accommodation bankrupting London's boroughs?

Arguably London's fastest-moving housing-crisis indicator. Select committee evidence puts around 183,000 Londoners, one in fifty, homeless in temporary accommodation, including more than 85,000 children, with councils' aggregate TA spend running into millions daily. The Institute for Government's October 2025 Performance Tracker names rising TA spend as a leading driver of local government financial distress, crowding out prevention spending. The RSA, Centre for London and the London Assembly have each proposed fixes over the past decade, worth checking which were actually taken up, and whether the numbers moved.

• Which London boroughs are most exposed: highest TA caseload relative to housing stock or population, and fastest-growing?
• Is there a visible link between TA spend and council financial-distress signals, such as Section 114 notices?
• Of the fixes proposed in the last decade of reports (RSA 2016, London Assembly 2019, Centre for London 2022), which were actually implemented, and did TA numbers respond?

Suggested datasets: DLUHC statutory homelessness live tables (homelessness/TA by local authority); MHCLG local government finance statistics; London Datastore borough profiles; English Indices of Deprivation 2025.

Format idea: School Cuts (NEU/ASCL/NAHT) lets anyone look up their own school's real-terms funding change, with a built-in "email your MP" action. The same shape could work here: a postcode or borough lookup for TA pressure, paired with a pre-written message to the local MP or councillor.

Brief PE: Will £340m end rough sleeping this time?

Fresh this week.

Andy Burnham became PM on 20 July 2026, and ending rough sleeping was his first pledge, backed by £340m for 1,200 new homes and support for 3,000 people. St Mungo's called it "a strong and welcome signal of intent" but stressed "there is significant work to do," citing his Greater Manchester record: he pledged to end rough sleeping in Manchester by 2020, missed it, though numbers fell 57% before rising again to a multi-year high. Rough sleeping (around 4,800 nationally, mostly single adults) is a smaller, more visible slice of the crisis than Brief PD's TA population (183,000+ in London): conflated in headlines, but different problems with different data.

• Using CHAIN (London's official rough sleeping database, run by St Mungo's for the GLA), has London followed the same rise/fall/rebound pattern Greater Manchester saw under Burnham?
• How far would 1,200 homes and 3,000 support packages go against London's own rough sleeping numbers, distinct from the wider TA population?
• What would count as "working" on CHAIN data a year from now, and is that a fair bar given the funding announced?

Suggested datasets: CHAIN (London rough sleeping snapshot and flow data); DLUHC rough sleeping snapshot statistics.

Brief PF: What does migration actually do to London's housing supply?

Handle with care. Needs a London lens.

FCDO has an active Area of Research Interest (a standing research priority, not a time-limited consultation), live since 11 May 2026: "What are the benefits and costs of migration… pressure on public services and housing supply? How do these relate to different categories of migrant?" Genuinely contested, politically live territory, written deliberately narrow, asking teams to test what the data shows rather than argue a position. Keep output evidence-first: name what the data can't support, and be explicit about correlation versus cause.

• Using ONS mid-year population and migration estimates alongside London housing completions and stock data, is there a visible relationship, in either direction, between migration flows and housing pressure at borough level?
• Does any effect differ by migrant category (students near universities, work-visa holders, asylum seekers in contract accommodation): very different housing markets that shouldn't be treated as one population?
• FCDO is asking because the evidence base is thin. What data would London need to answer it properly, and how much of that gap can this weekend close?

Suggested datasets: ONS mid-year population and international migration estimates; Housing Delivery Test; London Datastore; HESA higher education student numbers (as a proxy for the student migrant category).

Brief PG: Would Burnham's own tax reform help or hurt London?

Fresh this week. Handle with care.

Andy Burnham has repeatedly backed a land/property value tax, telling The Times on 22 May 2026, shortly before becoming PM: "I've long been persuaded of the argument for a land value tax… I see a big case for land and property and business taxation to be changed." He backs the Fairer Share campaign, which proposes replacing council tax and stamp duty with a flat 0.48% Proportional Property Tax (PPT), claiming 75% of households nationally would pay less. As PM, though, he's ruled out scrapping council tax or stamp duty "at this stage": a personal position and a campaign he backs, not policy. A national "75% win" claim can hide sharp regional splits, and London, with the highest property values and council tax bands frozen since 1991, is where that split is likeliest to bite.

• Using HM Land Registry price data and current VOA council tax bands, what would London households actually pay under a flat 0.48% PPT versus today's council tax: winners and losers, by borough?
• Council tax hasn't been revalued since 1991. How much of any London losses is just property values catching up for the first time, rather than the PPT design itself?
• Is a flat national rate the right design for London specifically, or would a banded or regionally-adjusted PPT change who wins and loses?

Suggested datasets: HM Land Registry price paid data; House Price per Square Metre dataset; VOA council tax band data; London Datastore borough profiles.

Brief PH: How much of London could actually get a street vote?

Street votes let two-thirds of residents on a street vote themselves planning permission for more homes: extensions, extra storeys, or small blocks, without the normal planning process. MHCLG has consulted on the rules and drafted around 90% of the legislation, but hasn't implemented it; it needs a final government push. An April 2026 modelling exercise (Centre for British Progress / Labour Together) put central-case impact at 28,200 homes a year nationally, 17,800 a year in London alone, on a take-up rate of just 0.17% of households a year.

Street votes exclude pre-1918 homes, listed buildings, conservation areas, and streets under ten homes.

• Given how much of London is Victorian or Edwardian terraced housing, how much stock would actually be eligible?
• The modelling assumes homeowners need £2,500/sqm profit to bother. Using the House Price per Square Metre dataset, where does that threshold clear, and does it match where the 17,800 homes a year is assumed to land?
• Does eligible, profitable street-vote territory overlap with where the TA and rough-sleeping pressure in Briefs PD and PE is actually worst, or would this policy mostly add homes somewhere else?

Suggested datasets: UK PlanIt planning application data; London Plan Opportunity Areas; House Price per Square Metre dataset; HM Land Registry OCOD/CCOD (ownership fragmentation: street votes need coordination among many separate owners); London Datastore borough/ward boundaries.

Brief PI: Which of these should your borough actually fight for?

No data skills required.

Every other brief here assumes some data work. This one doesn't: pure policy judgment, good for a policy-minded person without a data partner, or as the synthesis role on a mixed team. All the source material already exists in this document; nothing new needs pulling or analysing.

The scenario: you're Head of Policy for a London borough of your choice. Housing attention is unusually high: a new PM whose first pledge was on rough sleeping (Brief PE), a live consultation on the ground rent cap (Brief PB), RSH rethinking regulation (Brief PB), a PM sympathetic to land value tax (Brief PG), and a stalled-but-ready street votes reform (Brief PH). Your Leader has a meeting with a national decision-maker next month and wants one page on what to push for.

• Read the context paragraphs for Briefs PB, PE, PG and PH (no dataset needed) and pick one or two policy moves your borough should prioritise.
• For each option, weigh how much it would actually help, how politically live it is (a real deadline versus a personal view that isn't policy), and what the borough risks by engaging or staying quiet.
• Where the evidence in this pack conflicts or is incomplete, say so explicitly rather than picking the version that suits your recommendation.

Suggested output: one page, written as if for a council Leader ahead of a real meeting: plain English, a clear recommendation, and a short "what could go wrong" list.