The Approval Queue That Costs You Months
Data-driven decision making is the practice of using real-time data to inform business choices at every level of an organisation, from campaign adjustments to product pivots, without routing every decision through a centralised committee.
Most companies get this wrong. They hear “data-driven” and build governance structures that centralise every data decision at the board level. Committees get formed, approval workflows get created, and six months later, teams are still waiting for permission to run a basic experiment.
MIT’s Centre for Information Systems Research surveyed 342 respondents across 333 organisations and found that teams in large decentralised organisations took an average of 244 days to sense and seize business opportunities, compared to 566 days in centralised ones (MIT CISR, 2023). That is not a marginal difference. Decentralised organisations moved more than twice as fast.
TL;DR: Only 48% of organisations report being data-driven, and 92% say the barrier is people, not technology (Wavestone, 2024). This article covers a practical framework:
- Use the one-way door test to separate irreversible decisions (board-level) from the 80% that teams should own.
- Track four health metrics (quality, adoption, time-to-insight, incidents) instead of approving individual actions.
- Start with one team for a quarter, then compare speed and output to the old model.
Why Boards Default to Centralisation
The instinct makes sense on paper. Data carries real risk: privacy regulations, security exposure, reputational fallout. A 2025 IBM study of 1,700 data leaders found that over a quarter of organisations estimate they lose more than $5 million per year to poor data quality, with 7% reporting losses exceeding $25 million (IBM IBV, 2025). When boards see that risk profile, they pull decision-making authority upward.
But 92% of Fortune 1000 C-suite executives say the primary barrier to becoming data-driven is people and organisational change, not technology (Wavestone, 2025). Centralising decisions does not solve a people problem. It just adds a queue to it.
Boards meet quarterly, sometimes monthly. They are the furthest people from the day-to-day reality of how data gets collected, processed, and used. Asking them to govern data decisions is like asking your CFO to approve every ad creative.
I have watched this play out at startups and mid-size companies alike. A marketing team wants to integrate a new analytics tool. An ops team needs to restructure how customer data flows between systems. These are not bet-the-company moves. But they sit in a backlog because the governance framework routes everything to the same committee that also handles vendor contracts and compliance reviews.
A survey of 232 data practitioners across 48 countries found that over 70% wait more than a full business day just to access the data they need, with one in five waiting more than five days (Modern Data Company, 2024). Those delays compound. A week of waiting becomes a month of delayed experiments, which becomes a quarter of missed insights.
What Happened When I Stopped Approving Everything
I learned this the hard way at 2Stallions. When we were five people, every decision ran through me. Campaign strategy, tool selection, reporting changes, data access requests. That worked when we were small enough that I had context on everything.
Then we grew to twelve, and the model broke. I was the bottleneck. Campaign teams in Singapore wanted to test a new attribution tool but needed my sign-off. The SEO team wanted to restructure how we tracked client reporting data across markets. Both were reasonable requests sitting in my inbox for days while I dealt with client work and hiring.
The shift happened when I started asking one question about each request: “If this goes wrong, can we undo it within a week?” If yes, the team made the call. If no, we talked. That single filter cleared out most of the queue overnight. Did mistakes happen? Yes. Someone picked an analytics tool that turned out to be a poor fit and we switched three weeks later. Another team restructured a client dashboard in a way that confused the account managers, and we reverted it the next day. Small costs, easily absorbed.
What we gained was speed. Campaign tests launched in days instead of weeks. Teams started proposing experiments I never would have thought of because they had the authority to act on what they saw in the data. By the time we scaled to 40+ people across four countries, decentralised decision-making was not a philosophy. It was an operational requirement. You cannot run teams in Singapore, India, Australia, and the Philippines through a single approval chain.
The One-Way Door Test
Not all data decisions carry the same weight. The mistake most governance frameworks make is treating them identically.
Jeff Bezos introduced this concept as Type 1 and Type 2 decisions at Amazon. I’ve adapted it into something more practical for the companies I advise.
One-way doors are irreversible or extremely costly to reverse. Choosing your core data platform. Signing a three-year vendor contract. Deciding what customer data to delete permanently. Migrating your entire analytics stack from one provider to another. These deserve board-level scrutiny and careful deliberation.
Two-way doors are reversible. Testing a new dashboard tool. Adjusting how you segment customer data for a campaign. Adding a tracking parameter to your website. Switching from one A/B testing tool to another. If it doesn’t work, you roll it back. No lasting damage.
The overwhelming majority of data decisions in any company are two-way doors. I estimate that 80% or more of the data requests I used to review at 2Stallions were reversible, low-risk decisions that my involvement actually slowed down rather than improved.
How many of your pending data requests are genuinely irreversible? If you ran through the last month’s approval queue, you would likely find the same ratio.
| Centralised governance | Decentralised governance | |
|---|---|---|
| Who decides | Board or committee | Team closest to the work |
| Speed | 566 days avg. to act (MIT CISR) | 244 days avg. to act (MIT CISR) |
| Best for | One-way door decisions (platform changes, vendor contracts, data deletion) | Two-way door decisions (tool tests, segmentation, dashboards, tracking) |
| Risk model | Pre-approval for all decisions | Guardrails + post-hoc audit |
| Failure mode | Paralysis, queue backlogs | Occasional reversible mistakes |
How to Build a Decentralised Data Governance Framework
Seventy-one percent of organisations now have a formal data governance program, up from 60% a year earlier (Precisely & Drexel University, 2024). But the structure remains unsettled: organisations are evenly split between centralised (36%), federated (36%), and hybrid (29%) models (Board.org, 2025). The question is not whether to have governance. It’s where the decision authority sits.
Your governance framework should exist to make teams faster, not to protect the board from having to think about data risk.
Provide tools, not approval queues
Instead of requiring teams to submit requests for data access, give them self-service platforms with built-in guardrails. If the data has privacy constraints, bake those constraints into the platform itself so teams can move freely within safe boundaries. Row-level security, automated PII masking, and role-based access controls do more to protect sensitive data than any committee meeting.
Set standards, not permissions
Define how data should be labelled, stored, and documented. Then trust teams to follow those standards. Audit after the fact rather than gatekeeping before. Post-hoc review with clear consequences is faster and more scalable than pre-approval for every action.
Invest in literacy over bureaucracy
Most data governance problems are not malicious. They come from people not understanding what they should and should not do. Training your teams costs far less than building an approval bureaucracy, and it scales to new hires automatically.
This approach works particularly well in companies operating across multiple markets. When teams in different countries deal with different regulatory environments and customer expectations, centralised governance becomes not just slow but genuinely impractical. The team on the ground needs the authority to act within a clear framework, not to wait for a committee that lacks their context.
What Should Boards Track in a Data-Driven Organisation?
If boards should not be making individual data decisions, what should they focus on? The role shifts from gatekeeper to oversight. Four metrics tell you whether the decentralised system is healthy.
| Metric | What it measures | Warning signal |
|---|---|---|
| Data quality scores | Are teams maintaining clean, well-documented data? | Standards exist on paper but quality is declining |
| Adoption rates | Are teams actually using available data tools? | Low engagement with dashboards, self-service platforms |
| Time-to-insight | How long from question to answer? | Growing delays suggest governance is adding friction |
| Incident frequency | Breaches, compliance issues, quality failures | Rising incidents mean the guardrails need tightening |
Wavestone’s 2024 survey found that only 42.6% of large organisations have established a data and analytics culture, up from 21% the year before (Wavestone via Datanami, 2024). Low adoption in particular is a signal that something is broken, whether it’s the tools, the training, or the culture. The MIT CISR data reinforces this: the best decentralised organisations cut their response time by more than half compared to centralised peers.
These four metrics give boards genuine visibility into data health without requiring them to approve every dashboard change or tool evaluation.
The Financial Case for Pushing Decisions Down
This is not just about speed. The MIT CISR research found that large decentralised organisations reported net profit margins 6.2 percentage points higher and revenue growth rates 9.8 percentage points higher than their centralised counterparts (MIT CISR, 2023).
A separate HBR and Google Cloud study found that organisations leading on data and AI outperformed peers on operational efficiency (81% vs 58%), revenue growth (77% vs 61%), and customer retention (77% vs 45%) (HBR/Google Cloud, 2024).
These numbers matter when you present the case to a board that is reluctant to let go of centralised control. Framing it as “we want less oversight” gets resistance. Framing it as “decentralised governance correlates with 6-10 percentage points better financial performance” gets a conversation.
Start With One Team
If your organisation currently runs centralised data governance and this resonates, do not try to overhaul everything at once.
Pick one team, ideally one that is already mature in how they handle data. Give them explicit authority to make their own two-way-door data decisions within a clear set of standards. Track the four metrics above for a quarter. Compare their speed and output to teams still operating under the old model.
At every company where I’ve seen this tested, the results follow the same pattern. Teams that own their data decisions move faster, experiment more, and build genuine data literacy because the stakes of learning are real, not theoretical.
The board’s job is not to make data decisions. It is to build an organisation where good data decisions happen naturally, at speed, by the people who have the most context. That is the difference between governance that controls and governance that enables.
Frequently Asked Questions
What is data-driven decision making?
Data-driven decision making means teams at every level use real-time data to make calls without waiting for committee approval. It is not a quarterly board report or a dashboard nobody checks. It is the marketing team adjusting spend based on last week's conversion data because they have the authority and the access to act on what the numbers show.
What is a data governance framework?
A data governance framework is the set of policies, standards, and decision rights that determine how an organisation collects, stores, and uses data. The best frameworks define boundaries (privacy rules, labelling standards, access controls) and then trust teams to operate freely within them, rather than requiring approval for every action.
Why does centralised data governance slow teams down?
Centralised governance routes every data decision through the same overloaded committee: tool adoption, segmentation changes, access requests, reporting changes. Boards meet quarterly or monthly. Teams using data daily cannot wait that long. MIT research found centralised organisations took 566 days on average to act on opportunities, compared to 244 days for decentralised ones.
How do you decide which data decisions to decentralise?
Use the one-way door test. If a decision is reversible and the downside is limited (testing a new tool, adjusting campaign segments), the team makes the call and documents what they did. If it is irreversible or has significant financial or legal exposure (choosing a core data platform, permanent data deletion), it escalates. In most companies, 80% or more of data decisions are reversible.
What should boards track instead of approving individual data decisions?
Four metrics: data quality scores (are teams maintaining clean data?), adoption rates (are teams using available tools?), time-to-insight (how long from question to answer?), and incident frequency (breaches, compliance problems, quality failures). These give boards oversight without creating bottlenecks.
How does decentralised data governance affect financial performance?
MIT CISR research across 333 organisations found decentralised companies reported net profit margins 6.2 percentage points higher and revenue growth 9.8 percentage points higher than centralised peers. A separate HBR study found data leaders outperformed on operational efficiency (81% vs 58%) and revenue (77% vs 61%).
I advise founders and leadership teams on building organisations that scale without losing control. If your data governance is creating more friction than insight, let’s talk.
More on leadership and governance: Leadership & Governance | Related: Building a Data-Driven Culture at Startups
