I trained as a lawyer before I ever worked in marketing, and it shapes how I approach data more than almost anything else I learned in an MBA.
Before that I had a father who was a Judge. So every major request I made as a child, was the equivalent of coming before the supreme court (Ok, slight exaggeration - he was a 'High Court Judge')
A courtroom doesn't accept a claim because it sounds plausible — it wants the chain of evidence, unbroken, from source to conclusion. I bring the same discipline and clarity to my business meetings.
Currently we are innundated with data. It's growing exponentially: roughly 400 million terabytes of data are created every single day at current rates, and IDC's estimate is that global data volume is doubling roughly every three to four years.
"90% of the world's data was created in the last two years"
Ironically, just at the time when we are absolutely over-whelmed by data, scepticism about that very data is at an all-time high. We live in a 'post-truth' world, where almost every statement is questioned. Conspiracy theories abound. And no matter what you do, or where you live, your assumptions, and arguments will be questioned.
I think this is an area in marketing where a typical background data expert (let's say a Maths graduate who then got a masters or phd in statistics) will often run into trouble. Firstly, will their senior leadership audience understand them? Secondly, even if they do, will they believe them?
AI has enormous untapped potential. A key area that has been used in my sector it automation, by GTM 'engineers' for example. But a realm that I have discovered that I think has even more potential, and I have capitalised on is AI's ability to connect and convert disparate platforms data. (Harvard Business School Article)
Despite these advances, research suggests that scepticism is creeping into to discussions on the validity of AI driven data. That's why I always include methodology (for example, Statistical Significance of my findings), and back up my data insights with reasoning, and evidence, particularly when the outputs rely heavily on AI driven (for example in Claude) conclusions.
'The elephant in the meeting room'
The elephant in the room when working with Marketing and Sales data, is the levels of inaccuracy, gaps, and mismatches (especially across different platforms). In my experience, this is commonplace across a range of B2B businesses, and also across sectors; Fintech, Cybersecurity, Marketing software, all experience these challenges. When I hosted an ABM roundtable, heads of marketing at SaaS businesses all discussed trust in and reliability of data.
I always check my insights. And one rule I stick by, which I first heard from Jeff Bezos of Amazon is that 'when the data and stories don't match, I trust the stories'. I am not going to be 'that guy' who stubbornly doubles down on the numbers, when all the other signs are telling me otherwise.
That's where a discpline like Economics, which I studied at varous levels, is useful. Since Economics uses data, and sometimes complex maths to come to conclusions. But the best Economists realise that metrics will only ever measure part of the picture. The real world which Economics captures is complex, and certainly not straightforward, or rational, in the sense of a chemical formula, or a mathematical equation.
Behavioural Economics, which is the closest discipline to Marketing, captures that idea most effectively. I recommend reading 'Predictably Irrational' by Dan Ariely to see my point.
Classical Economics assumes that all actors behave rationally. Anyone who has worked at a company for any length of time will know that this certainly is not the case in the real world.The comic strip Dilbert is the best and most amusing example of this truth.
One of my favorite Economists, and an LSE Professor (Tooke Professorship of Economic Science and Statistics), Friederich von Hayek, said: “The fact that we can statistically ascertain certain magnitudes does not make them causally significant.”, and also,
"Data gives us precision about what we have measured. It doesn't necessarily give us completeness about what is happening."
(Hayek's 1974 Nobel Memorial Lecture, "The Pretence of Knowledge," delivered 11 December 1974 at the Stockholm School of Economics/).
And using calcuations in Sales and Marketing is the same. That philosophy informs my thinking. You should never confuse mathematical accuracy with insights in the real world. In the real world sales and marketing is a messy business with many variables we will never be aware of. The narrative must make sense on multiple levels, not just the mathematical.
Applying the theory
The rule I apply everywhere now: a piece of marketing engagement only counts as having influenced a deal if it can be shown to have happened before that deal existed in the sales system. Not correlated with it. Not "around the same time." Provably prior.
It sounds obvious written down. In practice, most attribution models quietly skip this step, which is how marketing ends up either overclaiming credit it didn't earn or — more often, in my experience — underclaiming credit it did.
Applying that discipline to one reallocation case built the evidence to correctly move a six-figure sum of revenue from "sales-sourced" to "marketing-sourced" in the CRM — not by arguing harder, but by showing the paper trail.
Recovered value like this, across a handful of cases, has added up to significant revenue that existed the whole time; it just wasn't being credited to the work that produced it.
Scoring leads on what actually predicts a sale, not what feels intuitive
The second half of this work is prospective rather than forensic: instead of proving what already happened, building a model of what's about to happen.
The instinct in most marketing teams is to score leads on things that feel like effort — email opens, form fills, general activity. In practice, across the datasets I've built models on, those signals are close to noise.
What actually separates a lead who converts from one who doesn't is a narrower set of behaviours: depth of engagement (not whether someone visited, but how many sessions and how many pages), job function, and — the single strongest predictor — how quickly sales actually makes contact once the signal appears. A high-intent lead left untouched converts below the average rate. The lift isn't in the lead. It's in the handoff.
Once you build a model around that and route the highest-scoring leads to the right rep on the same day rather than into a generic nurture sequence, you're not doing more marketing. You're doing marketing with a better sense of where the ball actually is.
The takeaway
If there's one thing I've learnt in my career so far, it's that "the dashboard says" is not evidence. It's a claim, same as any other, and it deserves the same scrutiny you'd give a witness. Trust the join, not the graph.
The closing argument
None of this — the reconciliation, the scoring model, the evidential discipline — means anything until it becomes a decision. Data and trends are the evidence. They are not the verdict.
That's the step most data people skip. They lay out the exhibits, build the chart, cite the methodology, and expect the room to convict itself. My father never delivered a judgment by reading the evidence aloud and going quiet. He built an argument from it — reasoning applied to fact, aimed deliberately at a conclusion someone else could act on with confidence, in a room full of people whose job was to doubt him.
That's the part of the job nobody teaches a data analyst. Not proving what already happened, or predicting what's about to. Standing in front of a leadership team who have been burned by an over-claimed dashboard, or a plausible-sounding AI summary that fell apart under one follow-up question, and making the case anyway — with the paper trail sitting behind you, ready for cross-examination, because you built it that way from the start.
The data and the trend get you to the leadership room. The argument is what gets you the agreement. And of course, then there is testing. A topic for another day; But you should always test your hypothesis once you make your claim.
