“Data-driven” is one of the most overused phrases in financial services. Every firm claims to be moving towards it. Most are not. This article is a practical playbook — five concrete moves that distinguish a data-driven organisation from a merely data-rich one.
The diagnosis
A data-rich organisation has lots of data. It has dashboards, data lakes, BI licences, an analytics team, possibly a Chief Data Officer. It produces decks. It runs initiatives. It buys platforms. None of that, on its own, makes it data-driven.
A data-driven organisation does something specific: when it makes a decision, the people in the room reach for data first, the data is fast enough to use, and the decision rests on the data rather than on the most senior opinion present. That is the test. Most organisations fail it not because they lack data but because the data takes too long, comes from a different team, or arrives in a form that the decision-maker cannot work with.
Move 1 — close the time-to-data loop
The first investment is not better dashboards or fancier ML. It is reducing the time between “I have a question” and “I have an answer good enough to act on”. If that loop is days, no amount of dashboard polish will help — the decision will be made on intuition before the data arrives. If it is minutes, the organisation behaves differently almost immediately.
Practical first step: name the three decisions that get made most often by your executive team, time how long it takes to get the underlying numbers, and put a target on bringing each under fifteen minutes. That is a tractable engineering problem. It is also the single highest-impact move most organisations can make.
Move 2 — agree the numbers once
The second move is harder. It is the agreement, written down and signed, that there is one definition of each business-critical metric, owned by one team. Capital ratio. Loss ratio. ECL coverage. Whatever the most-quoted three numbers are at your monthly executive meeting — those need to be governed metrics, not extracts.
If finance and risk and operations each compute their own version of the capital ratio and the numbers disagree, the executive meeting will spend its time arguing about which is correct. That argument is not data-driven decision-making. That argument is a tax on the operating model. The cure is editorial rather than technical: a metrics catalogue, owned by someone, that says “this is the definition, this is the owner, this is the source pipeline”.
Move 3 — make the data product, not the dashboard, the unit of work
Most BI work fails because the unit of delivery is a dashboard. A dashboard is a presentation layer; it has no audit trail, no version, no contract. The next time finance changes how they bucket loss exposures, every dashboard built on the old logic silently produces wrong numbers.
The unit of delivery should be the data product underneath: a versioned, owned, contract-tested dataset that the dashboards consume. Five governed data products that all your dashboards consume from is a much stronger position than fifty bespoke dashboards each with their own extraction logic.
Move 4 — measure decision quality, not data volume
The CDO function in most organisations measures success by data quantity (terabytes ingested), data quality (rows passing tests) and platform usage (dashboard views). None of those are decision quality. None of them tell you whether the data is changing the way the business decides things.
The data-driven version is harder to measure but more honest. Pick three classes of decision the business makes monthly — pricing reviews, capital allocation, hiring approval, whatever. For each, log whether data was consulted, whether the data was good enough to be useful, and whether the decision tracked the data or overrode it. Six months of that data will tell you more about your data programme than any dashboard view count.
Move 5 — pick a small number of high-stakes use cases and own them
Data-driven transformations almost always fail when they try to fix everything. A data team that is running 40 simultaneous workstreams is doing none of them well. A data team that is running three high-stakes use cases — IFRS 17 close, ECL re-pricing, executive MI — and owning the end-to-end pipeline for each, is building muscle that compounds.
The picking is the hardest part. Most organisations will not say no to anything. The question to ask: which three decisions, if we made them ten times faster and with twice the confidence, would change the business this year? Build the data products underneath those three. Refuse the others.
The investment mindset
Becoming data-driven is not a transformation programme. It is a steady, multi-year reprioritisation of where engineering and operating effort goes. The teams that succeed treat it like a capital allocation question: invest in the few decisions that move the business, get them working end-to-end, then move on. The teams that fail keep buying platforms.
If you want to scope which three decisions are worth getting right first, our Diagnostic Assessments are designed to surface exactly that.