Jack Lawrence

Having to explain your data is fine if you have complexity that needs to be understood. However if your data is simple and you aren't there to explain it you need to avoid ambiguity. The BBC have gone stats mad, which if you like data is great. However you then start to think about these stats. (taken at the 71st minute of Atletico Madrid vs Tottenham when the score was 5-1) Atletico have had 7 shots on target and there were 2 goalkeeper saves, giving them 5 goals. So far so good. Spurs have had 3 shots on target and there were 3 goalkeeper saves. So how does that give them 1 goal? Now you end up with more questions. Are the 3 goalkeeper saves for Spurs, saves by their goalkeeper or saves from their shots? With no explanation you lose clarity.

8

5 Comments

This has to be a must-read for any leaders out there wanting to maximise the impact of their investment in data and AI for the future...

3

📢 If you are (in anyway) a Data Analyst you need to focus on being a 'Decision Shaper' not an 'Order Taker'. 🎯 The same very much applies to the wider Data, Analytics & Insight function. 🎯 If Data & Insight are not shaping & informing decisions, if Analysts are not seen as (and behave as) Business Partners, you will never get real value from the function.

69

9 Comments

There’s a lot of debate about why data teams don’t always deliver the value they should. I think this is part of the reason. We were discussing process mapping as part of our transformation programme, thinking through how and when data would be captured within redesigned journeys. It became clear we needed to spend more time agreeing what the data is, what it means, and how it will be defined and used. In fast-paced delivery, it’s easy to focus on user flows, screens and forms, and assume the data will take care of itself. But without a clear, shared understanding of the data model, we risk creating processes that look good but fall short when it comes to reporting, automation or insight. Somewhere along the way, we’ve lost the habit, and the skill, of designing data properly. Not heavyweight diagrams and long documents, but simple, deliberate alignment across business, data and tech. This isn’t about slowing things down. It’s about making space for practical, agile data design at the point it matters most. 🧩 Define what data is needed alongside the process 📋 Align business analysts, product owners and data teams early ✔️ Build structure and validation into workflows from the start 👉 Because data should always be aligned to the business, and earn its place on the pitch. When that happens, everything else flows better. Curious how others are approaching this; are you bringing data into the room early enough?

97

28 Comments

I’ve been pondering the biggest issue in Data this week. Amongst all of the excitement (and let’s be honest, hype) around AI, as a profession we are still not getting the basics right. The most important of which is Data Quality. I’ve been working with Data for nearly 40 years and I don’t think it has ever been this bad. We have gone from programming languages like COBOL (yes I’ve been there) which get you to define the structure of the values in variables, and constraints on fields in databases that enforce the same, to just tons of rubbish in excel spreadsheets. How did we get here?

24

5 Comments

Metadata changed my view of data engineering. I was introduced to its real importance during my graduate programme at PBT Group. Back then, I thought building pipelines was about writing code and delivering fast. But I learned something fundamental: 👉 without metadata, scale breaks 👉 without metadata, trust disappears 👉 without metadata, people become the documentation My time at Old Mutual Insure and later Nestlé reinforced this at enterprise level. Across industries, the most successful platforms always had strong control across seven areas: source configuration, schema evolution, transformations, data quality, security, orchestration, and lineage. When these are right, organisations move faster with less risk. This is when I shifted from asking: “How do I build this pipeline?” to “Will this system still make sense in five years?” That’s metadata thinking. And it’s what turns data platforms into decision platforms. #DataEngineering #Metadata #DataGovernance #DataArchitecture #AnalyticsEngineering #DigitalTransformation #DataLeadership #AI #InstitutionalIntelligence

8

A question that comes up a lot in data is: “HOW DO WE IMPROVE DATA QUALITY?” In my experience, I've come to realise over the years, that’s usually not the best place to start. Trying to improve everything isn’t realistic (trust me, I’ve tried!). And most of the time, it’s not needed. The more useful question is: “Where does data quality actually matter?” Which data is used to make decisions? Which data affects clients or customer? Which data has a real commercial impact? That’s where the effort should go. Everything else still needs to be managed. But it doesn’t need the same level of attention. The challenge isn’t just improving data; it’s being clear about what’s worth improving in the first place. #Data #DataQuality #DataStrategy

10

2 Comments

It seems to me that the really clever data analysts and engineers of the world spend an awful lot of time cleansing and tidying data before they can do the real work with it. This is less than super efficient and I'm willing to bet a lot of them privately think it's a waste of their time. The answer is right in front of us - ensure you have good quality data at source. This is not usually a system problem but rather a people problem and, guess what, it's not really their fault. Millions and millions of people in the world do jobs where they are shown how to use a system and what data to put into it but no-one is telling them why. Why is it so important that we have quality data? To most of you fine folks reading this it's obvious but not everyone gets it. So it's on us to try and ensure that they do get it. Talk to them, educate them in their own language, encourage them to think about what happens to that piece of data further down the line. Here at the NFCC we have released a methodology for measuring and improving data quality, and we're working on ways of improving our data literacy so that the quality goes up in tandem. Put people at the heart of data quality, and data literacy and the results will absolutely follow.

63

12 Comments

Things learned from 100 Data Assessments - There are a lot of misclassified products, more than you would think. On average I have seen about 5-10% of items being misclassified and have seen it get frighteningly high. If items are misclassified they are hard to find and have the wrong attributes associated with them, make that findability even harder. I recommend cyclical data audits to ensure everything is in its right spot and that any new categories are added as needed.

11

1 Comment

Your Data Team Doesn't Need A Hero. It Needs A Leader! Leadership in the world of data isn’t about knowing every tool or writing perfect code. It’s about creating space for ideas, experimentation and collaboration. I’ve learned that the best data teams don’t just analyse data, they inspire trust through transparency and teamwork. Every innovative idea starts with a safe space to share one. What are you doing to create that space for your team? How do you build trust within your team? #Leadership #DataTeams #InnovationCulture #Collaboration #GrowthMindset #PeopleFirst #DataCulture #AnalyticsLeadership #TeamDynamics #PsychologicalSafety #HighPerformingTeams #DataDrivenCulture #FutureOfWork

11

3 Comments

The latest buzz? 🐝 AI isn't just a fancy buzzword, it's becoming the BA's secret weapon! From rapidly analysing complex datasets to identifying hidden trends, AI tools are supercharging our ability to deliver insights faster than ever. Imagine predictive analytics flagging potential issues before they even become problems. It's not about replacing the human touch, but augmenting it, freeing BAs to focus on strategic thinking and stakeholder engagement. Practical tip: Start exploring AI-powered data visualisation tools today and your insights will thank you! #AIinBA #BusinessAnalysis #FutureofWork #DataDriven

7

What is "data governance"? It's the weekend, and I found myself thinking about the time my sister spent working with Damien Hirst on his famous Spot Paintings. Damien Hirst's Spot Paintings work on a deceptively simple principle: no single dot means anything alone. Its meaning emerges from its position within the grid, its colour relative to every other colour, and the rules governing the whole — no colour repeated, every dot the same size. But a single Spot Painting can fetch $500,000 at auction. Not because any individual dot is valuable — the paint is cheap, the canvas is standard — but because the system that organises them creates something whose value vastly exceeds the sum of its parts. Data is the same. We have become comfortable talking about data as an asset — but a raw data point, on its own, is worth very little. What creates value is the rules, the connections, and the context: who can access it, under what conditions, for whose benefit, and with what accountability. That is what data governance actually is. Not just a compliance framework. Not just a legal obligation. It is the political architecture through which societies decide who controls the conditions under which data becomes information — therefore determining economic and political power. This week, my team and I had a planning workshop and we decided to discuss the basic question: what is data governance, really? I wrote up my thinking in a short blog post. I'd love to hear your views — especially from those working at the intersection of digital policy, AI, and development. https://lnkd.in/dzPbC-8g

10

1 Comment

From Readiness to Winning with Data (1)...Completing a maturity model isn’t a winners’ medal. Most data transformations feel like a success when the maturity model is complete. The platform is live. Dashboards are being used. Adoption metrics are climbing. Governance is in place. Literacy programmes have run. The roadmap has been delivered! We’ve enjoyed the end of transformation party. But here’s the uncomfortable truth: Completing a data maturity model is a taking-part award, not a winners’ medal. It proves you’ve built capability. It does not prove you can use it. And we treat those signals: platform launches, adoption metrics, governance completeness as evidence of success, rather than evidence that success is now possible. So, organisations stop. They pause investment. They move on to the next initiative. They assume value will now show up “over time”. What they’ve really done is stop just as things get hard. Because maturity models measure readiness. They do not measure performance. They don’t tell you: - which decisions should now be taken differently - where trade-offs must become explicit - what changes under pressure - or how you’ll know you’re genuinely outperforming Readiness qualifies you for the race. Winning requires something else entirely. In this arc, From Readiness to Winning, I want to explore what has to change once the data capability is built: - how data capability must be designed for performance, not just readiness - who owns winning when evidence becomes competitive - and what separates organisations that qualify from those that compete for the podium Can you clearly articulate what “winning” with data looks like for your organisation — and how you’ll know you’re competing, not just taking part? #DataLeadership #DecisionMaking #FromReadinessToWinningWithData

17

9 Comments

Inspired by the excellent posts of Warren Powell and Adam DeJans Jr., I’d like to share my perspective on modelling optimisation problems. My aim is to contribute to the ongoing conversations on LinkedIn that promote operational research in practice. 1. Start with problem scoping and plain language An imperfect model or heuristic that addresses the right business pain points will outperform a perfect solution to the wrong problem every time. Clear, accessible problem definitions are the foundation of successful optimisation projects. Defining the entities and their attributes is the first step before writing any mathematical notation. 2. Invest time upfront I dedicate a significant amount of time to scoping the problem before writing a single line of code. This upfront effort not only avoids costly rework later, but also ensures tangible business benefits. Jumping into solutions too quickly is like stepping into a dark room without a flashlight. 3. Collaborate widely Engage stakeholders early, discuss their objectives, constraints, and priorities, and hold workshops to build a shared understanding. This process not only clarifies the decisions and data involved, but also creates the consensus needed to implement changes. Business leaders who feel part of the modelling journey are more likely to champion the outcomes. 4. Start simple, evolve fast Don’t aim to build the entire solution from day one. Deliver a simple version that captures most of the value, then gather feedback from stakeholders. Use this input to build a prioritised backlog, focusing on features that offer the highest value for the least effort. Keep iterating and keep the conversation open. I’d be keen to hear how others in the OR and data science community approach this balance between technical rigour and business adoption. #OperationsResearch #Optimisation #DataScience #Analytics #OR #DecisionScience #MathematicalModelling #OperationsManagement #AI #MachineLearning

81

11 Comments

I get asked all the time about buying cold data - company names, numbers - probably because I’ve spent a big chunk of my career in the data information business. Should you buy it? What do you do with it? Intent data? Variables? You name it! Here’s the truth: unless you’ve invested in the right technology and the right people, it’s a pointless purchase. I’ve never seen a winning GTM plan that looks like this: Data vendor → Send email → Hope for the best or Data vendor → Send DM → Hope for the best Without a robust plan for how you’ll turn data into pipeline, you’re just burning money. My advice: don’t bother until you can make the data work for you. I have a framework for this - happy to share. Just ask.

10

The UK public sector is stuck in a data mess. Not because of a tech shortage — but because it still runs on scripts. The future? Autonomous data agents. They don’t just automate tasks. They achieve outcomes. → Clean data → Faster insights → Less manual work Engineers stop coding. They start commanding. This is the agentic shift. #AgenticAI #AIagents #PublicSectorInnovation #GovTech #DataOps #DigitalTransformation #AutonomousAI #AIinGovernment #AIforGood #SmartData #RplusAnalytics

15

2 Comments

Data people, I'm begging you to stop asking other data people what their data stack looks like. It's as annoying as asking people what their 5k running time is. It's dull and actually gives very little by ways of useful information. Talk to me about problems we're solving or challenges you're facing, not what ETL tool we're using. PLEASE.

87

26 Comments

🎧 A NEW podcast episode is now available! This week, I’m doing a deep dive into Data Governance road maps and how essential they are when doing Data Governance. What’s your experience of a Data Governance road map - Essential or optional? I'd love to hear your thoughts... 👇 🎧 Listen to the full episode: https://lnkd.in/eHVxpVwk https://lnkd.in/eHkKJ5Za Or stream it on your favourite podcast platform. #datagovernance #datagovernanceprocesses #datapodcasts

16

4 Comments

This week’s blog post on my training at The Data School focused on Level of Detail calculations in Tableau Prep. If that all sounded like a load of jargon, don’t fear (I literally write these blog posts to make topics less confusing!). So, Tableau Prep as I may have mentioned before is a software that we use to clean data, making it easier to analyse and use in visualisations. Level of Detail calculations, in this case specifically FIXED LODs, allow you to choose what fields you include in a calculation to impact the end result. For example, fixing the level of detail for a population calculation at town rather than county would produce different values. There are lots of different adjustments that you can add to this calculation such as including multiple fields and changing the way you calculate your measure value. If you’d like to read more about the topic, here’s my blog post with a step-by-step guide on implementing and understanding FIXED LODs in Tableau Prep: https://lnkd.in/emBb86pk

3