Minimum Viable Data 2.0

From ‘swamps’ to ‘smarts’

Over-collection has swollen lakes into swamps, while generative-AI tools now spin results so fast that errors masquerade as truths. Bad data ‘in’ still equals bad data ‘out’. To break the cycle, we argue that you need less data, better context and continuous trust. Rather than hoard information ‘just in case’, capture only the facts that matter, enrich them with business context, and govern them continuously.

The answer to your business problems isn’t more data, it’s ‘smart’ data… in context. That is the ethos of Minimum Viable Data (MVD) 2.0.

Revisiting the MVD idea

We first wrote about the MVD concept while steering government digital-transformation initiatives in 2021, arguing (early) that capturing, storing and governing every byte was both impractical and counter-productive. Four short years later that warning feels understated. The volume of data, including dark and unused data, has radically swollen to dominate many data estates, fuelling operational cost, creating cyber risk and analytic noise. Meanwhile, generative AI systems have exploded onto the scene, promising breath-taking ways of working, and faster insight, which simultaneously exposes a brutal dependency - without small, well-curated, context-rich data sets, even the smartest AI model simply hallucinates at scale.

That tension, between the unstoppable surge of raw data and insatiable appetite from ever smarter algorithms demands a sharpened response. Organisations are in a conspicuous dilemma trying to decide where proven business value ought to be.

Minimum Viable Data 2.0 is our pragmatic answer. By focusing governance and enrichment on the smallest, highest-trust slice of information required to power critical decisions, we ensure that both humans and machines operate on a foundation that is lean, reliable and continuously fit-for-purpose.

From big data hype to smart data reality

For the last decade the mantra was simple – store everything. Storage was cheap, and perhaps one day a clever algorithm would unearth hidden gold. In practice, the mountain grew faster than our ability to catalogue it. Querying became slower, governance slipped, and confidence nosedived.

By contrast, smart data begins with a clear mission question and deliberately curates only the facts that matter. These facts are validated on entry, enriched with relevant business context, and stored in a schema-agnostic hub so they can serve any application or AI model. The approach echoes the ‘old’ (but often forgotten) design principles of flexibility, speed, interoperability, resilience and trust, which were once used, back in the ‘olden days’, when we did not have access to that tsunami of data. We used what we had and used it well.

Five propositions to align stakeholders with MVD

These propositions act as a compass when enthusiasm to ‘collect everything’ resurfaces.

1. Good data beats more data every time - Volume without veracity slows the organisation.

2. Gen AI is an accelerator, not an alchemist - It won’t turn ideas into gold by mixing vectors. It only amplifies the quality of its inputs, good or bad.

3. Context is king - Relationships between people, places, and events convert raw facts into actionable intelligence.

4. Governance should be embedded at ingestion - So that security, lineage and privacy flow downstream automatically.

5. Data value decays - Which means MVD must be a continuous loop, not a one-off project.

The Minimum Viable Data 2.0 loop

• START: By framing the business challenge and intended outcomes. Only with that purpose defined do we select the without which the decision cannot be made.

• LOCATE AND ASSESS: High-value facts, essential entities, attributes, and relationships, scoring each source for relevance, quality, lineage and sensitivity.

• CONTEXTUALISE AND HARMONISE: Ingesting data ‘as is’ into a semantic hub where automated enrichment tags time, location and business meaning, creating a single, feature-rich asset.

• GOVERN: For trust, applying attribute-based controls, encryption, and immutable provenance.

• ACTIVATE AND OBSERVE: Through APIs, dashboards, event streams and AI agents, with feedback loops measuring accuracy and impact.

• ITERATE AND RETIRE: Reassessing each asset’s contribution. Data that no longer ‘earns its keep’ is archived or purged, maximising investment where data drives the greatest strategic return.

Architectural guardrails

A successful MVD platform ingests data as it arrives and indexes it immediately, favouring speed over premature perfection. Because new sources appear constantly, the repository must be schema-agnostic, yet relationship-aware. It employs RDF knowledge graphs so analysts and algorithms can traverse connections fluidly. Event-driven APIs and data streams decouple producers from consumers, ensuring that a change in one system propagates everywhere without brittle point-to-point integration. Resilience is non-negotiable and a zero-trust security posture protects the data layer itself, not merely the applications that sit above it.

MVD 2.0 is not a theoretical manifesto, it is a practical concept, rooted in both proven logic and tested data logistics. It aims to prove (on real systems and real users) that sharply focused, context-rich data consistently outperforms vast, un-curated stockpiles.

Turning concepts into competitive advantage

LMC Digital turns the Minimum Viable Data 2.0 concept into a competitive edge through a laser-focus on the decisions that matter; a schema-agnostic spine that grows without re-engineering; and governance robust enough to satisfy both executives and algorithms.

We help our clients to:

• Pinpoint the bottleneck: Surface the single decision-flow whose fix unlocks the greatest value and fastest confidence.

• Stand up the backbone: Deploy a secure-by-design smart-data hub, event-driven, zero-trust, and ready to scale domain by domain.

• Prove-and-extend: Each production slice lands in live hands, cuts rework, and earns sponsorship before the next domain joins, spreading lineage, enrichment, and security services automatically.

• Embed lifelong governance: Lineage, privacy, and auditability are baked in from day one, ensuring every future expansion inherits the same rock-solid trust model.