Stop building an output factory. Start moving the needle.
Peter Everill on why the biggest unlock in data and AI isn't the technology, it's knowing which decisions actually matter to your P&L.
There is a version of a data organisation that most of us would recognise. Requests flood in from across the business. The team is stretched across dashboards, ad hoc reports, and platform maintenance. Output is plentiful. Impact is harder to point to. And somewhere in the background, a leadership team is quietly wondering whether the investment is paying off.
Peter Everill has spent the better part of two decades trying to solve that problem. First in consultancy, then at Sainsbury's where he led data products for seven years, and now at IAG Loyalty as Head of Data Product. In a recent conversation on the Driven by Data podcast, he laid out the thinking that has shaped his approach: a five-decision framework that connects data and AI directly to P&L performance.
The hat you're wearing matters more than you think
Peter is quick to point out that the data world contains multitudes. Technologists, mathematicians, governance leads, infrastructure specialists; all valuable, all necessary. But he wears a different hat.
"What drives me is making commercial impact and being able to transform the performance of a business. It just so happens that I think the most effective way to do that is using data. The business performance is my what and why. Data, analytics, and AI is just the how."
— Peter Everill, Head of Data Product, IAG LoyaltyThat reframing matters enormously for Chief Data Officers. When the mandate is defined primarily around technology delivery, building platforms, managing infrastructure, producing outputs. It becomes almost structurally impossible to demonstrate P&L impact. The goal posts are in the wrong place from the start.
Peter's early career crystallised this for him. As an analyst he could describe what had happened, but struggled to change what would happen next. When he moved into consultancy and eventually took on a transformation programme at the Financial Ombudsman Service during the PPI mis-selling crisis, something clicked. He was using AI to route cases to the right adjudicators and automate recommendations, embedding decisions and workflows with data, not just decorating them with it.
Top-down, not bottom-up
One of the more practical tensions Peter addresses is the pull towards bottom-up demand management. Left to their own devices, operational stakeholders will generate a near-infinite list of requests; reasonable ones, individually, but collectively a recipe for sprawl with no coherent commercial logic.
The alternative is to start from the P&L and work down. It sounds obvious. It is surprisingly rare.
"At P&L level, there's only a handful of decisions that really turn the needle. Building a roadmap against a handful of things rather than a myriad sprawl of things becomes a lot easier."
— Peter EverillThe discipline this requires, pushing back on requests that cannot be tied to a commercial outcome, asking stakeholders what action they will actually take from a given piece of analysis — is uncomfortable. But it is precisely this discomfort that separates data organisations with commercial credibility from those that remain perpetual request-fulfilment services.
The five decisions that move a P&L
Over fifteen years, Peter has distilled his approach into five types of decisions that any data and AI strategy should be organised around. Each builds on the last, and each delivers increasing levels of P&L impact.
A single version of the truth. Without it, leadership wastes time debating what is happening rather than what to do about it.
Understanding where performance problems actually originate, often across multiple teams, so effort goes to the right place.
Optimising how constrained budgets are allocated, with clear visibility of the trade-off between cost and lost sales.
Real-time, at-scale decision making embedded directly in operational workflows, acting before performance deteriorates.
Resolving the conflicts between departmental goals that look fine locally but damage overall P&L performance.
To ground each type, Peter uses a supermarket availability example throughout. Take root cause: if products are out of stock, the problem could sit with supply chain forecasting, logistics, in-store replenishment, or the online picking team. Without structured root cause analysis, the business will likely focus in the wrong place, burning budget on a problem it doesn't actually have, while the real one goes unaddressed.
The framework also exposes a subtler organisational failure at the enterprise optimisation stage. A ranging team might rationally increase the number of products stocked, more choice, better customer satisfaction scores. What they may not see is that increasing the number of SKUs reduces shelf space per product, which increases replenishment frequency, which drives up operational cost and lost sales in a different part of the business. The incentives are locally coherent and globally destructive. Only the P&L view reveals the conflict.
Where AI fits, and where it doesn't
Peter is measured on the topic of AI, in a way that CDOs navigating board-level enthusiasm might find useful to borrow.
He is entirely supportive of AI for workflow productivity — automating repetitive tasks, drafting content, accelerating operational processes. But he is clear that this is not where the significant P&L value sits. That value is in the decisions.
"If you just automate the workflow without the decisions being smarter, you're making it really cheap to send the wrong product to the wrong customer at the wrong time. You're not going to materially improve your conversions."
— Peter EverillThe CRM example he uses is pointed. Generative AI can produce personalised content at scale and near-zero marginal cost. But if the underlying decision, which customers to contact, when, with what offer — remains unsophisticated, the organisation has simply industrialised mediocrity. Speed is not the same as intelligence.
Two questions worth taking into your next leadership conversation
When Peter assesses the maturity of a data organisation, including, he notes candidly, during interview processes, he asks two things of business leaders. What are the most important decisions affecting your P&L? And how well do you understand the performance and root cause of those decisions right now?
The first question, he finds, most commercial leaders can answer confidently. The second is where things get interesting. Most organisations can report on performance. Far fewer can reliably and quickly identify why performance has moved. Particularly when the root cause spans multiple functions.
For CDOs, those two questions also serve as a useful diagnostic for their own teams. If your roadmap cannot be mapped back to a clear answer on both, it is worth asking what it is actually optimised for.
Equally, for data leaders, Peter offers pointed advice on focus: resist the sprawl, identify one or two stakeholders genuinely willing to engage, and demonstrate impact in a narrow place before expanding. The organisational change that comes with making performance problems visible is not always comfortable. Finding a sponsor who welcomes that scrutiny, rather than fearing it, is half the battle.
This article is based on an episode of the Driven by Data podcast hosted by Kyle Winterbottom, CEO & Founder of Orbition Group. Peter Everill is Head of Data Product at IAG Loyalty. Listen to the full episode for the complete conversation.
