Walk into almost any boardroom and you'll hear that data is a strategic asset. Walk into the same company's data warehouse and you'll find three conflicting figures for "active customers," a dashboard nobody trusts, and an analyst spending Tuesday reconciling spreadsheets. The gap between the sentence and the warehouse is where most data strategies quietly die. Closing it doesn't take a bigger budget, it takes a few deliberate choices about what your data is for.

The quick version

  • "Data as an asset" is a claim with a bill attached. Assets are valued, governed, and held to a standard. If you say it, you owe the data the same care you'd give cash or property.
  • Every data strategy is a trade-off between offense and defense. Offense uses data to grow (revenue, customer insight, better decisions); defense protects it (compliance, security, one trusted number). You can't max both, pick a tilt and say so.
  • Data isn't oil. It's non-rival: many teams can use the same data at once without using it up. So the strategic question is rarely "do we have enough?", it's "is the right data trusted, findable, and pointed at a decision?"
  • The move: name one or two decisions you want data to improve, declare a single source of truth for the numbers behind them, and pick your offense/defense tilt on purpose.

The idea in depth

The phrase "data as an asset" is older and more demanding than its boardroom use suggests. The person who made the case most rigorously is Doug Laney, the former Gartner analyst who coined the discipline he calls infonomics. In Infonomics (Routledge, 2017) he argues that if information genuinely is an asset, organisations should do the unglamorous things asset ownership implies: measure its value, manage its quality, and account for it the way they account for inventory. His uncomfortable observation is that most firms treat data as exhaust, a by-product of operations, while claiming in public that it's treasure. An asset you never measure, govern, or improve is an asset in name only.

The move, then, is to make the word "asset" cost something. Before you repeat it, ask: which data, valued how, held to what quality bar, owned by whom? If you can't answer, you don't have a data asset, you have a data liability that hasn't billed you yet.

Offense and defense: the choice you're already making

The most useful single frame for data strategy comes from Leandro DalleMule and Thomas Davenport in "What's Your Data Strategy?" (Harvard Business Review, May–June 2017). Their argument is that data work splits into two opposing aims. Defense is about minimising downside: regulatory compliance, security, fraud detection, and, critically, a single, trusted version of key numbers. Offense is about generating upside: using data to win customers, improve products, and decide better, which rewards speed and flexibility over perfect consistency. The two pull in different directions, and the authors are blunt that no company can be equally aggressive on both at once. The job of a data strategy is to choose where on that spectrum you sit, and to expect the answer to differ by function, a finance team leans defensive, a growth team leans offensive.

They put a sharp number on the cost of getting this wrong: in their account, "less than half of an organization's structured data is actively used in making decisions, and less than 1% of its unstructured data is analyzed or used at all." Treat that as a vivid illustration from a 2017 practitioner article rather than a precise law of nature, but the direction is hard to dispute: most organisations are sitting on data they never turn into decisions.

So the move is to stop arguing about tools and name your tilt. Tell the organisation, in plain words: "For the next two quarters we lean offensive in growth and defensive in finance and privacy." That single sentence resolves a hundred downstream arguments about whether the priority is moving fast or being sure.

flowchart LR
  D(["Your data"]) --> O("Offense, grow
revenue, insight,
better decisions")
  D --> F("Defense, protect
compliance, security,
one trusted number")
  O --> T(["Strategy = a deliberate
tilt, by function"])
  F --> T
Every data strategy resolves to a tilt between offense and defense, and the right tilt differs by team. After DalleMule & Davenport (HBR, 2017). Leaders Loop

The single source of truth, and its honest limit

The other half of the DalleMule–Davenport framework is the single source of truth (SSOT): one authoritative, governed copy of each core data element, with clear provenance, so "revenue" or "active customer" means one thing across the company. From that trusted core, teams can derive their own "multiple versions of the truth", a marketing definition of an active user, a finance definition, as long as everyone can trace them back to the same governed source. This is where data strategy meets data governance, quality and lineage: the SSOT is only as good as the stewardship behind it.

Be honest about the limit, though. A single source of truth is expensive to build, and chasing a perfect one for every field is a classic way to spend two years and ship nothing. The pragmatic read is to declare an SSOT for the handful of numbers in board decks and customer-facing decisions, and to tolerate messiness elsewhere until it earns the cleanup. Governance is a means to better decisions, not a monument to tidiness.

Why data isn't oil, and why that changes the strategy

"Data is the new oil" is the most repeated and most misleading line in the field. The flaw is economic. Oil is rival: the barrel I burn, you can't. Data is non-rival, the economists Charles Jones and Christopher Tonetti put it precisely in "Nonrivalry and the Economics of Data" (American Economic Review, 2020): "a person's location history, medical records, and driving data can be used by many firms simultaneously." Their peer-reviewed analysis shows that because data can be copied and used by many teams at once, broad use of it can create increasing returns, yet firms often hoard it, which is inefficient.

That points somewhere specific: the binding constraint is almost never the raw quantity of data. It's whether the data you already hold is trusted, discoverable, and connected to a decision. Hoarding "just in case" creates governance cost and risk without value. The leadership question shifts from "how do we collect more?" to "what is the most valuable decision our existing data could improve, and what's stopping it?" That reframe, from accumulation to activation, is the whole game, and it pairs naturally with how you think about probabilistic versus deterministic systems when you decide how much certainty a given decision actually needs.

The strategic question is rarely "do we have enough data?" It's "is the right data trusted, findable, and pointed at a decision?"

A worked example

Consider Mara, who runs operations at a 200-person subscription software company. (The scenario is illustrative; the figures below are invented to show the shape of the problem, not real benchmarks.) Her CEO has just told the board that "data is our biggest untapped asset" and asked Mara to build a data strategy. The instinct in the room is to buy a warehouse, hire data scientists, and "collect everything."

Mara resists the collect-everything reflex and applies the frame instead. She asks one question: which decisions would better data actually improve? Two surface immediately. First, the finance team and the growth team report different churn numbers every month, roughly 4% versus 6%, and the board has stopped trusting either. Second, customer success can't predict which accounts are about to cancel until the cancellation email arrives.

That diagnosis sorts cleanly into the framework. The churn-number conflict is a defense problem: she declares a single source of truth for "monthly churn," defined once, owned by finance, with the calculation documented so both teams trace back to it, no new technology, just a definition and an owner. The cancellation-prediction gap is an offense problem: the existing usage data, already sitting unused, could feed an early-warning signal, which connects to machine learning concepts and utility once the data is trustworthy enough to model. She sequences defense first, you can't build a useful model on a number two teams dispute, then offense.

flowchart TD
  A(["Name the decision
to improve"]) --> B{"Defense or
offense?"}
  B -->|Trust & consistency| C("Declare an SSOT:
one definition,
one owner")
  B -->|Growth & insight| D("Activate existing data
against the decision")
  C --> E(["Then build offense
on the trusted core"])
  D --> E
Mara's sequence: a decision first, a defense/offense diagnosis next, trust before models. Leaders Loop

Six weeks later Mara hasn't bought a warehouse. She has one agreed churn number, an owner for it, and a small experiment turning idle usage logs into a cancellation-risk flag. The "untapped asset" got tapped not by collecting more, but by pointing what already existed at two decisions that mattered.

Frequently asked questions

Isn't "treat data as an asset" just obvious?

The slogan is obvious; the bill is not. An asset gets valued, governed, quality-controlled and owned. Most organisations skip every one of those and keep the slogan. The useful version of the idea is the demand it makes, name an owner, a definition and a quality bar, not the comforting phrase.

Do we need a data warehouse and data scientists before we have a strategy?

No, and starting there is a common, costly mistake. Strategy is the decision about what your data is for and how you'll tilt between offense and defense. Tools and hires implement that decision; they can't substitute for it. Plenty of valuable first moves, agreeing one definition, naming one owner, need no new technology at all.

If data is non-rival, should we just collect everything?

No. Non-rivalry means the same data can serve many uses, which is an argument for sharing and activating what you hold, not for hoarding more. Every dataset you collect carries governance, security and privacy cost. Collect against a decision or a credible future use, not "just in case."

How is this different from data governance?

Strategy decides what your data is for and where you place your bets; governance is the machinery, quality, lineage, stewardship, that keeps the data trustworthy enough to bet on. A single source of truth is the point where the two meet. You need both, but the strategy comes first or the governance has nothing to aim at.

What's the one number to measure?

Resist a single vanity metric. The honest measure is decision-linked: for each priority decision, is the underlying data trusted, and did the decision get better or faster because of it? If you can't trace a data investment to a decision it improved, treat that as a flag, not a footnote.

Related in the Toolkit

Where to go next