Here's a guide we prepared before our discussion that might be helpful as you explore these topics yourself and with your team.

Episode Overview
This guide contains 50 questions and sample answers across 10 categories, designed to structure a rich, exploratory conversation between the host and Lawrence Deju-Wiseman on the theme of Playing With Data.
■ Host answer
■ Lawrence Deju-Wiseman answer
Category 1: What Is 'Playing With Data'?
Q1. What does it mean to you when you say someone is 'playing with data'?
Host: For me it's that exploratory phase — no fixed agenda, just curiosity. You load up a spreadsheet and start asking 'what if' questions without knowing where you'll end up.
Lawrence Deju-Wiseman: I'd say it's giving yourself permission to be wrong. In compliance and surveillance we're trained to be precise, but play means you hypothesise, you test, you discard. The insight comes from the failures as much as the wins.

Q2. Is 'play' the right word, or does it risk making data work sound trivial?
Host: I use it deliberately. The word 'play' lowers the stakes psychologically. People who are afraid of data will engage if you call it play rather than 'analysis' or 'modelling.'
Lawrence Deju-Wiseman: In financial regulation the word play would raise eyebrows — we're dealing with market integrity. But behind the scenes, the best surveillance models I've seen were built by people who played with the data before they formalised anything.

Q3. What separates play from random clicking around in a dataset?
Host: Intent and reflection. Even play should have a loose question behind it. Random clicking gives you nothing to build on; play gives you a direction you can pivot from.
Lawrence Deju-Wiseman: Structure. I approach a new dataset like a legal brief — I want to understand the parties, the timeline, the relationships. That framework is what makes exploration productive rather than chaotic.

Q4. Can you play with data if you don't have coding skills?
Host: Absolutely — that's the whole premise of my podcast. Pivot tables, Power Query, even well-designed Excel formulas are enough to start playing meaningfully.
Lawrence Deju-Wiseman: I'm living proof. My background is law and accounting, not computer science. I learned to ask the right questions first and then find tools to answer them. SQL came later; the curiosity came first.

Q5. What's the most surprising insight you've ever stumbled on while playing with data?
Host: I was cleaning a client's invoicing data and accidentally sorted by time-of-day. Every large invoice was submitted between 4 and 5 pm on Fridays. That pattern turned into a whole fraud-risk conversation.
Lawrence Deju-Wiseman: In FX surveillance I noticed a pattern of small trades clustering just before large ones from the same desk, in the same currency pair, on the same day. That wasn't in any rulebook — pure exploratory data play flagged a layering behaviour we hadn't defined yet.

Category 2: Data Quality & Preparation
Q6. You've said data should be 'at least a little bit organised.' What's the minimum viable level of organisation?
Host: I mean consistent column headers, one concept per column, and no merged cells. If I have those three things I can work with almost anything.
Lawrence Deju-Wiseman: Consistency of reference data is my baseline. In trading surveillance, if a single counterparty has three different identifiers across three systems, your aggregation is broken before you start. Unique, consistent entity IDs are non-negotiable.

Q7. How much time should you spend cleaning data before you start exploring?
Host: I apply the 80/20 rule. Get the data clean enough that the 80% of rows that matter are reliable, then play. You'll discover the remaining 20% of cleaning issues during exploration anyway.
Lawrence Deju-Wiseman: Clean as you go, but document every transformation. In a regulatory context, you cannot retrospectively clean data without audit trails. The cleaning is part of the evidence.

Q8. What are the biggest data quality red flags you've encountered in financial services?
Host: Timestamps in the wrong timezone, undocumented nulls, and currency fields stored as text. Those three will derail any analysis.
Lawrence Deju-Wiseman: In derivatives surveillance — missing LEIs, trade IDs that aren't truly unique, and lifecycle event data where novations and amendments aren't linked to the original trade. The data model is often an afterthought.

Q9. Is dirty data ever actually useful?
Host: The patterns in errors are often data themselves. If five percent of records have a field blank, that's random noise. If it's always the same client, that's signal.
Lawrence Deju-Wiseman: Absolutely. In market abuse monitoring, incomplete or late-reported transactions are themselves a behavioural indicator. The absence of data, or data that arrives out of sequence, tells its own story.

Q10. How do you convince a non-technical colleague that data quality is worth investing in?
Host: I ask them to show me the last report they sent to a senior stakeholder and then I find one obvious error in it. That conversation usually lands faster than any theory.
Lawrence Deju-Wiseman: I frame it as legal liability. If your surveillance system misses a manipulation case because the underlying data was wrong, that's a regulatory failure. Risk language works in financial services.

Category 3: Data Literacy & Who Should Be Playing
Q11. What does data literacy actually mean to you — is it about technical skills or something else?
Host: It's the ability to ask a question of data and critically evaluate the answer. Knowing Excel formulas is secondary; knowing whether an answer makes sense is primary.
Lawrence Deju-Wiseman: It's numeracy plus scepticism. In accounting I was trained to always ask 'does this number make intuitive sense before I present it?' That instinct is the core of data literacy.

Q12. Can data literacy be taught, or is it something you either have or don't?
Host: It can definitely be taught, but motivation matters enormously. People who are naturally curious about 'why' tend to pick it up fast. You can't teach curiosity — but you can create environments that reward it.
Lawrence Deju-Wiseman: At the University of Connecticut I teach graduate students who arrive with very different quantitative backgrounds. The ones who succeed aren't always the most technical — they're the ones who ask the most uncomfortable questions about the data.

Q13. Who should have data literacy in an organisation — just analysts, or everyone?
Host: Everyone who makes decisions. If a manager can't interrogate the dashboard they're using to make hiring or budget decisions, they're flying blind.
Lawrence Deju-Wiseman: In financial services, every first and second line of defence function needs it. Compliance officers who can't read a surveillance alert report are dependent on whoever wrote it — which is a control risk.

Q14. What's the difference between someone who is data-literate and someone who is data-dependent?
Host: Data-literate people use data to challenge their intuitions. Data-dependent people use data to justify decisions they've already made. One is discovery, the other is confirmation bias dressed up in spreadsheets.
Lawrence Deju-Wiseman: I call it the difference between reading the trading book and being told what it says. You need to be able to read the book yourself so you know when someone is summarising selectively.

Q15. How do you assess someone's data literacy in an interview?
Host: I give them a messy spreadsheet for ten minutes and ask them to tell me the three most interesting things they notice. I'm not testing skill — I'm testing curiosity and scepticism.
Lawrence Deju-Wiseman: I ask candidates to walk me through a time when data told them something unexpected and what they did about it. The answer reveals whether they trust data blindly or engage with it critically.

Category 4: Organisational Permission to Play
Q16. You mentioned that without your boss's permission to play, you're done. How common is that barrier?
Host: Extremely common. Most office environments reward certainty and punish exploratory work that doesn't immediately produce a report someone can put in front of a board.
Lawrence Deju-Wiseman: In heavily regulated industries it's structural. You can't just 'play' with client data — there are data governance frameworks, access controls, and privacy regulations that constrain what you can touch and how.

Q17. How do you create permission to play within a culture that doesn't naturally support it?
Host: Start small and show results. I tell clients to carve out two hours a week labelled as 'data review' — innocuous enough to survive any manager — and do their playing in that slot.
Lawrence Deju-Wiseman: I've framed it as scenario testing. 'I want to test whether our surveillance model would have caught this historic case.' That's defensible because it sounds like risk management rather than play.

Q18. What does a 'permission-to-play' culture actually look like day to day?
Host: It's a team meeting where someone says 'I was messing around with last month's data and found something weird' and the manager says 'tell me more' instead of 'was that in your workplan?'
Lawrence Deju-Wiseman: It looks like a chief compliance officer who reviews not just findings but false negatives — asking 'what did we miss and how might we find it?' That question creates space for exploration.

Q19. Can you play with data without organisational permission if you're careful about it?
Host: To a degree. You can play with anonymised, aggregated, or your own sandbox data. But the best insights usually come from real operational data, and for that you need legitimate access.
Lawrence Deju-Wiseman: In financial services you should never try to circumvent data governance — the consequences are severe. But you can influence what gets put into the sandbox. Advocacy within the system is more durable than workarounds.

Q20. How do you convince leadership that play has ROI?
Host: I keep a log of insights that came from unstructured exploration versus structured analysis. In my experience, the ratio of 'unexpected high-value finds' strongly favours the unstructured time.
Lawrence Deju-Wiseman: Regulatory fines are a compelling data point. If I can show that exploratory surveillance work would have flagged a behaviour that later became an enforcement action at a peer firm, the ROI conversation changes quickly.

Category 5: Domain Expertise & Context
Q21. How important is it to know the field you're analysing data about?
Host: Critical up to a point. You need enough context to know what's normal so you can spot what isn't. But too much context can make you blind to anomalies you've been trained to explain away.
Lawrence Deju-Wiseman: In trading surveillance, someone without derivatives experience will flag everything or nothing. You need to understand how a legitimate hedge looks so you can differentiate it from a manipulative position.

Q22. Can domain expertise actually get in the way of good data analysis?
Host: Yes — I call it 'that's just how we do it' syndrome. The most dangerous phrase in any data project is 'that number looks about right.' About right is not analysis.
Lawrence Deju-Wiseman: Absolutely. The most sophisticated market manipulation schemes exploit exactly this. They structure activity to look like normal trading to the trained eye. An outsider might ask 'why does everyone in this market do it this way?' and expose the whole game.

Q23. How do you balance insider knowledge with outsider scepticism?
Host: I bring in someone unfamiliar with the client's industry for at least one review session. Their naive questions are worth more than ten hours of my analysis.
Lawrence Deju-Wiseman: I deliberately hire from adjacent disciplines. Lawyers who understand market structure, accountants who understand coding, technologists who understand financial regulation. The intersections are where the insight lives.

Q24. Lawrence, how has your dual qualification as a lawyer and accountant shaped how you approach data?
Host: I'll let Lawrence speak to this one, but I imagine two very different epistemological traditions collide in interesting ways.
Lawrence Deju-Wiseman: Enormously. Law teaches you to interrogate every assumption and demand evidence for every assertion. Accounting teaches you that numbers must balance and that if they don't, something is wrong. Together, you approach data with a healthy prosecutorial scepticism.

Q25. What's an example where cross-domain thinking led to a data breakthrough in your work?
Host: I worked with a logistics company and applied customer churn analysis borrowed from telecoms. They'd never thought about fleet operators as 'customers who might leave.' The model was a direct transplant.
Lawrence Deju-Wiseman: Bringing legal concepts of intent and knowledge into algorithm design for surveillance. Most systems flag behaviour; the best ones try to infer purpose. That conceptual leap came from legal training, not data science.

Category 6: Cross-Domain Thinking & Fresh Eyes
Q26. You argue for people with experience in OTHER fields. Why is that specifically valuable?
Host: Because every field has solved problems that other fields haven't discovered yet. The person who has worked in both healthcare analytics and retail knows that patient flow models look a lot like customer journey models.
Lawrence Deju-Wiseman: In RegTech, the best surveillance models I've seen borrowed from epidemiology, network theory, and even criminology. None of those people came from a trading background — they came in with different pattern libraries.

Q27. How do you avoid the 'too close to the paradigm' problem in your own work?
Host: I read outside my field deliberately — a lot of behavioural economics, some organisational psychology, the occasional paper on animal cognition. Weird inputs produce non-obvious outputs.
Lawrence Deju-Wiseman: I teach. Standing in front of students who have no preconceptions about how financial markets 'should' work forces me to question my own assumptions regularly. It's the best calibration tool I have.

Q28. How do you find people who have the right cross-domain background for data work?
Host: I look for unusual career paths. Someone who moved from nursing into project management into data. The zigzag usually signals someone who is genuinely curious rather than someone who followed a predetermined track.
Lawrence Deju-Wiseman: I hire for intellectual range. If a candidate's reading list is entirely within one discipline, I'm less interested. The people who can move between frameworks tend to build more robust models.

Q29. Is there such a thing as too much cross-domain exposure — too many frameworks, too little depth?
Host: Yes — I call it 'framework tourism.' Someone who knows the name of every model but has never run one to completion. Depth in at least one area is what gives you the credibility to borrow from others.
Lawrence Deju-Wiseman: In professional qualification terms, absolutely. You need a home discipline. For me it's law and accountancy. That gives me a rigorous foundation from which I can borrow analytical frameworks without losing the thread.

Q30. What's the best cross-domain analogy you've ever applied to a data problem?
Host: Treating a company's internal email network like a biological ecosystem — mapping predator-prey dynamics to understand which departments had outsized influence over others. Came directly from reading ecology papers.
Lawrence Deju-Wiseman: Applying the legal concept of 'reasonable person' as a calibration benchmark for surveillance models. Instead of asking 'is this unusual?' we ask 'would a reasonable experienced trader have acted this way?' It changed how we set thresholds.

Category 7: Tools, Technology & RegTech
Q31. What tools do you consider essential for playing with data at the office worker level?
Host: Excel with Power Query for most people, Python or R once they're ready to scale, and a decent visualisation tool — I'm a Tableau fan, though Power BI has caught up considerably.
Lawrence Deju-Wiseman: For surveillance specifically — a proper case management system, a robust data pipeline, and something that can do network analysis. But I always tell teams: understand your data before you deploy any tool on top of it.

Q32. What is RegTech, and how does it change how financial firms interact with their data?
Host: I'll let Lawrence define it, but from the outside it looks like compliance process automation — using technology to do what lawyers and accountants used to do manually.
Lawrence Deju-Wiseman: RegTech is the application of technology to regulatory compliance, monitoring, and reporting. It changes the relationship with data fundamentally — instead of sampling, you can monitor everything in near real-time. But that creates a new problem: you generate more alerts than humans can review.

Q33. How does AI fit into data play — does it accelerate it or replace it?
Host: Accelerates, without question. AI gives me a first draft of analysis in minutes that would have taken hours. But the question-asking, the 'does this make sense,' the insight — that's still human.
Lawrence Deju-Wiseman: In surveillance, AI can surface patterns no human would find by looking. But AI also inherits the biases and gaps in the training data. You still need a human who understands the market to validate what the model surfaces.

Q34. What's the biggest mistake organisations make when they invest in data tools?
Host: Buying the tool before defining the question. I've seen companies spend six figures on a BI platform with no agreed definition of what success looks like. Tools don't create insight — questions do.
Lawrence Deju-Wiseman: Assuming the tool will compensate for bad data. A sophisticated surveillance platform sitting on top of inconsistent, incomplete reference data will produce sophisticated wrong answers. Technology amplifies whatever quality you start with.

Q35. Where do you see the most interesting intersection of data and regulation in the next five years?
Host: ESG data reporting. The pressure to report on sustainability metrics is going to force companies to build data infrastructure they don't currently have, and that infrastructure will generate insights they can't yet imagine.
Lawrence Deju-Wiseman: Digital assets and crypto. The on-chain data is extraordinarily rich — every transaction is permanently recorded. The challenge is building the analytical frameworks and regulatory guidance to interpret it meaningfully. That's a genuinely open frontier.

Category 8: Data Ethics, Privacy & Governance
Q36. Where does ethical responsibility sit when you're playing with data that involves real people?
Host: With the person doing the playing. Intent matters — are you exploring to improve something, or are you looking for ammunition? That distinction should govern every decision.
Lawrence Deju-Wiseman: In regulated financial services, it sits simultaneously with the individual, their firm, and the regulatory framework. Those three layers sometimes align and sometimes create genuine tension, particularly around personal data in surveillance.

Q37. How do you balance the surveillance role — monitoring people's behaviour — with their privacy?
Host: This is a profound tension. Most employees don't realise the extent to which their trading and communication activity is monitored. How do you think about that, Lawrence?
Lawrence Deju-Wiseman: It's a known, contractual relationship in financial services — employees consent as a condition of employment. But the ethical obligation is proportionality. Surveillance should be targeted, purposeful, and not drift into general monitoring of behaviour that has nothing to do with market integrity.

Q38. What does good data governance look like in practice, not just in a policy document?
Host: It looks like someone being able to answer 'where did this number come from?' in under two minutes. If you can't trace a figure back to its source, your governance is theoretical.
Lawrence Deju-Wiseman: It looks like a data lineage map that anyone on the team can read, a change log that captures every transformation, and regular data quality reviews that are reported to a senior committee — not just the team that owns the data.

Q39. How do you handle a situation where playing with data reveals something uncomfortable about your own organisation?
Host: You document it, escalate it, and don't sit on it. The temptation to explain it away is strong, but the pattern is usually telling you something true.
Lawrence Deju-Wiseman: In compliance you're legally obligated to report certain findings. Beyond that, the professional obligation is to present the data accurately and let the governance process run its course. The lawyer in me says: paper every step of that process.

Q40. Is there data you simply shouldn't play with, regardless of the potential insights?
Host: Personal health data combined with employment data, for starters. The potential for misuse is so high and the harm so significant that it deserves much stronger justification than most 'playing' allows.
Lawrence Deju-Wiseman: Legally privileged communications, for one. And in my world, anything that could constitute a breach of client confidentiality. The value of a potential insight never outweighs those protections — and frankly, any insight derived from improperly accessed data is tainted anyway.

Category 9: Real-World Applications & Case Studies
Q41. Can you give an example of a data play session that turned into a formal process or product?
Host: I was playing with email metadata for a client — not content, just frequency and timing. The exploration became a team health dashboard that the HR director now runs every quarter.
Lawrence Deju-Wiseman: An exploratory analysis I did on the timing of FX trades relative to news releases became the basis for a formal alert type in our surveillance platform. It started as a 'what if we looked at this?' and ended as a production model.

Q42. What's a data project that looked promising but ultimately led nowhere — and what did you learn?
Host: I tried to build a model predicting employee attrition from calendar data. The correlation just wasn't there — or rather, the confounding variables were too many to isolate. I learned that HR data is messier than financial data in ways that are easy to underestimate.
Lawrence Deju-Wiseman: We tried to use sentiment analysis on broker commentary to predict trading intent. The language was too specialised and too deliberately ambiguous — traders are trained to be opaque. The model produced noise. The lesson: domain language needs domain training data.

Q43. How do challenger financial firms approach data differently from incumbents?
Host: Challengers have the advantage of building their data infrastructure from scratch — no legacy systems, no thirty-year-old databases. The disadvantage is they often lack the historical data that makes models robust.
Lawrence Deju-Wiseman: Challengers are often data-native but compliance-immature. They built products on the assumption that data would flow freely and then discovered that financial regulation creates significant constraints on how you can use it. The tension is fascinating and often underestimated.

Q44. What's the most valuable dataset in financial services that most firms underuse?
Host: Customer complaint data. Companies treat it as a problem to be managed rather than a signal to be mined. The patterns in complaints tell you more about product failure than most internal metrics.
Lawrence Deju-Wiseman: Transaction time-stamp data at millisecond resolution. Most firms aggregate it into daily summaries before they analyse it. The intra-day sequencing is where market abuse lives, and most firms aren't looking at it at the right granularity.

Q45. If you could give every office worker one data habit to develop, what would it be?
Host: Question the denominator. Whenever someone shows you a percentage, ask them what it's a percentage of. Half of misleading statistics disappear when you establish the base.
Lawrence Deju-Wiseman: Document your assumptions. Every piece of analysis rests on assumptions about what the data means, what's been excluded, and what 'normal' looks like. Write those down before you start. You'll be surprised how often revisiting them changes your conclusion.

Category 10: The Future of Data Work & Closing Thoughts
Q46. How is the role of the 'data person' in an organisation evolving?
Host: It's fragmenting and specialising. Five years ago everyone wanted a 'data analyst.' Now organisations are realising they need data engineers, data storytellers, and domain-fluent data translators — very different roles.
Lawrence Deju-Wiseman: In financial services the role is converging with compliance. The demand for people who understand data infrastructure AND regulatory requirements AND the markets themselves is growing faster than the supply. It's the most interesting skills gap in the industry.

Q47. Will AI eventually replace the need for human data literacy?
Host: No — it will raise the floor of what passes for data literacy and create a ceiling that requires much higher-order skills. The people who ask bad questions will be replaced; the people who ask great questions will be amplified.
Lawrence Deju-Wiseman: The legal system hasn't replaced lawyers; it's changed what lawyers do. AI in data analysis will follow the same pattern. The human value will migrate toward judgement, challenge, and contextual interpretation.

Q48. What advice would you give someone who wants to become a better data player starting tomorrow?
Host: Download a dataset about something you love — sports, music, food, anything — and spend one hour asking it questions with no agenda. The habit of curiosity is the skill; the tools come later.
Lawrence Deju-Wiseman: Find the data that describes your own professional world and learn to read it critically. You already have the domain knowledge. Add the habit of interrogating the numbers and you become exponentially more effective.

Q49. What's the most important thing you've learned from playing with data that you couldn't have learned any other way?
Host: That the stories we tell about our organisations are almost always simpler than the data that underlies them. The data is messier, more contradictory, and more interesting than any narrative we impose on it.
Lawrence Deju-Wiseman: That behaviour leaves traces. Every decision, every transaction, every communication — in data terms it exists. The question is whether you have the framework to interpret it. Playing with data is the process of building that framework.

Q50. What question should I have asked you today that I didn't?
Host: Maybe: 'What do you do when the data contradicts something your gut is telling you?' Because that tension — between instinct and evidence — is at the heart of everything we've discussed.
Lawrence Deju-Wiseman: I'd have liked you to ask: 'What does good look like?' We talk a lot about problems and failures. Good surveillance, good data governance, good data literacy — those exist, and studying them is as valuable as studying the failures.