One home. One tenant.
One version of the truth.
Foundation Link lands your housing management, finance and repairs data in a governed lakehouse — conformed in a mapping studio, gated by a quality rule engine, and masked column by column before anything downstream can query it.
Connects over PostgreSQL · Snowflake · MySQL · REST API · S3 / ADLS Gen2
A housing data platform, in one sentence.
Foundation Link is the governed layer between the systems you already run and the answers you need out of them. It lands the data, conforms it, proves it is fit to use, and masks what should not be seen — three things, kept current, that together answer almost every question anyone will ask about your data.
Ingest
Every source system — HMS, finance, repairs API, file drops — landed on a schedule you control, with schema evolution and row-level change tracking from the first run.
Govern
Quality rules gate what reaches reporting. Column-level masking decides who sees a real email address, a real date of birth, a real postcode — and who never does.
Model
A conformed gold layer across six canonical housing entities, queryable in plain English and available to the pre-built models in the marketplace.
Bronze, silver, gold.
Nothing reaches a report unproven.
Every field arrives through a mapping you can read as generated SQL and passes a rule you can read as a predicate. Both are on screen before either one runs.
Bronze
Raw and source-shaped, one schema per connected system. Full load or incremental against a watermark column, on a cron you set — with run-now, forced reload, pause and archive per job.
Silver
Mapping Studio conforms source columns to canonical housing entities — properties, tenants, repairs, tenancies, appointments, voids — with chained transforms and the generated SQL shown before it runs.
Quarantine
Records failing a rule are routed to an entity quarantine table tagged with a violation code, not silently dropped and not silently passed. Fix, then release.
Gold
Curated, masked, governed. The only layer analysts, conversational queries and AI models are allowed to touch.
● The only layer AI seesControls a data team would build,
already built.
Data Quality & Quarantine
A rule engine that runs before the merge, scoring accuracy, completeness, consistency, timeliness, validity and uniqueness into one grade your board can read.
- Not-null, pattern, allowed-values, composite and referential-integrity rules
- Visual rule builder or custom SQL, with the live predicate always visible
- Quarantine management with violation codes and a release action
- A DQ Copilot that explains why a pillar score moved
Governance & PII Studio
Column-level masking driven from metadata and applied as generated gold-layer views, so two people looking at the same table can legitimately see different values.
- Full mask, email mask, partial mask, age and postcode generalisation, salted hash
- Sensitivity tagging, with AES-256-GCM on high-sensitivity columns
- Unmasked-for grants by group — admins, engineers, analysts, auditors
- Destructive changes require explicit confirmation
Mapping Studio
Source-to-silver mapping per business entity, with a confidence indicator on every field so nobody has to trust a fuzzy match they never saw.
- One silver destination per entity, not per field
- Chained transforms: cast, trim, case, conditional defaults
- Generated SQL rendered in real time before anything runs
Talk to your data
Plain-English questions resolved to SQL against the gold layer only — never against raw or half-cleaned data, and never around your masking rules.
- Same masking and access rules as direct table access
- Row-level security applied automatically
- Admins can inspect exactly which tables a query space exposes
“One home, one tenant, one version of the truth. Ingested, cleaned, governed and AI-ready — built for social housing, not bolted on to it.”Foundation Link · an accelerator by Lashan Digital
Models built for housing outcomes,
not generic ML.
A catalogue of pre-built models. Each one binds to an inference table and a scheduled scoring job over the gold layer at the point you deploy it.
Act upon poor workmanship
A repeat-repairs model that identifies recurrent poor workmanship.
Identify ASB and domestic violence
Identifies customers who may be experiencing anti-social behaviour or domestic violence, so a person can act proactively.
Reduce contact volumes
Surfaces trends in contact centre volume to inform process redesign.
Reduce complaint volumes
Identifies the recurrent issues driving complaints, to intervene pre-emptively.
Each deployed model is an AI system in its own right, and the ASB and domestic-abuse model in particular touches a sensitive, high-risk use case. Before production use, a deployment needs a recorded risk classification, a DPIA and a named human decision point — output that drives action without meaningful human review would be automated decision-making under UK GDPR. Foundation Link gives you the governed, masked data a score was computed from; it does not certify the decision, and it is not a substitute for legal advice.
Answers for the people
who have to give them.
Foundation Link is built around three audiences inside a housing association, not one generic dashboard everybody ignores.
IT & data leads
One place to register a source, schedule its ingestion and hold its masking policy, instead of a point-to-point integration per request — across separated dev, staging, sandbox and production environments.
Finance
The finance warehouse landed alongside housing data in the same governed layer, and questions answered against it in plain English rather than raised as a report request.
Executive & board
One data quality grade across every active rule and scorecard pillar, and a quarantine list showing exactly which records were held back and which rule stopped them.
Governance by design,
not by policy document.
Five controls that are part of how the platform is configured, rather than a policy document written alongside it.
See it running on your own data.
Book a walkthrough against a live dataset and see the quality grade, the quarantine list and the masking rules on real records.