Data Projects – securing the win

One of the hottest debates in the data space right now is the “success factor” discussion. 

Many data leaders argue that no matter how sophisticated the approach or technologically adept the solution if a clearly demonstrable business outcome has not been realized, the data project cannot be classed as a success. 

It’s a controversial topic, all right. And it’s one that will require the assistance of some serious heavy weight data heroes to tackle.

Good job we know a few of them, right?

So, in this short blog, we’re going to look at 5 Top Tips from those who know best – the data leaders who specialize in successful deployment – and ask them to help us build the ultimate short “how to” guide to delivering a successful data project.

The Importance (or not!) of Algorithms

Our first pearl of wisdom comes from Martin Thorn – Head of Data Science & Data Platform at abrdn.  

Martin writes:

Algorithms aren't important. Or to be clear, they aren't as important as data scientists think they are. In fact, no-one in business cares what algorithm you used.

However, many data scientists not only think it's important to discuss their methodology but lead with it. 

I don't care which algorithm you used. Your stakeholders don't care either. We care about outcomes. 

Wow. Strong words to start our journey with. 

Martin’s take is that “seeing your working” isn’t nearly as important to stakeholders as some data practitioners might think it is. As a member of the first community, I don’t know the difference between a linear regression and a decision tree. I don’t care. Both sound scary.

Whatever technique works is the right one, and I don’t need to understand what it is. I just need an answer.

Regarding outcomes, our second tip comes from another hugely experienced data science leader – Marta Portugal – Director of Data Science at Forth Point. 

Marta suggests:

Identify your need first

For a data project to be successful it needs to solve a need. 

If it’s just a passion project and not rooted in business needs/improvement and with the end user in mind it will never be adopted and will therefore fail. 

So, I guess in summary my tip would be “Don’t do data science for the sake of data science, always put value and outcomes first”

Marta focusses us very firmly on working within the business, engaging with stakeholders, and gaining a thorough understanding of the domain first – before choosing a project or challenge to tackle. 

These needs will be varied, and complex, and some more difficult to solve than others. It is crucially important that, if success is the desired end goal, you choose a challenge that will be both solvable AND deliver an outcome to the business. Without both, confidence erodes and key stakeholders within the business may not be convinced, long-term, of the viability of the data team’s outputs.   

Get the scale right to start with

Our third piece of advice comes from Dr Adam Sroka, Head of Machine Learning Engineering at Origami.

With regards to looking at what outcomes might be the best place to start within an organisation, 

Adam suggests:

Think small - try to think of the smallest possible improvement you can make and iterate towards big. 

Going big up front carries a lot of risk. 

This game requires a lot of mixed experience and skills. 

You need all the feedback you can get so that you don't spend months moving in completely the wrong direction

So, following on from Marta’s guidance, Adam stresses the importance of not biting off more than can be chewed. 

Looking for the small victories first, not overstretching or over committing, and building incrementally from a position of existing strength before tackling more sizable internal data projects will give a successful platform to build from.  

Our next contributor is Dr. Jasmina Lazic, Head of Research & Development at bigtincan. 

Jasmina writes:

Quality (of data) rather than Quantity

What really moves the needle in building effective AI systems is having high-quality data, rather than huge amount of noisy data or tinkering with the models. 

So, in every data science project, it is crucial to think through and define what data is used to train the algorithms, what data collection and processing is going to look like, and what kind of data governance will be in place. 

Building smart AI systems is all about leveraging the least amount of high-quality data possible, as opposed to looking for a needle in a haystack, which is trying to process and derive useful insights from terabytes of noise.

Jasmina highlights that a considerable amount of work needs to be done to ensure that the data itself is in the best possible shape and that the conditions for its usage are optimized before undertaking a data project. In other words, do not neglect the (data) foundations of the project. Build upon a solid base to give yourself the best chance of success.  

Our fifth and final tip comes from Dr. Christopher Foley, Chief Data Scientist and Executive Board Member at Optima Partners. 

Chris says:

It’s (very much) a Team Game

Look to statistics to help demystify many of the mysteries of data science, but fundamentally, pulling ideas and experience from your team and colleagues leads the way.

Chris’ tip showcases an opinion held by many data leaders that a successful data project is a successful group exercise. With a wide variety of skills and experience to draw opinion (cognitive diversity, anyone?) from team members and the wider organizational gene pool, success can often hinge upon how quickly or readily this experience and inspiration is drawn upon. Communication and collaboration can often help to uncover potential roadblocks, particularly in the early stages of a data project. Without this group support, any individual could potentially seriously jeopardise their chances of success. 

So, there we have it. The data heroes have spoken.

Focusing on outcomes rather than inputs, don’t bamboozle your stakeholders with unnecessary technical jargon, ensure that your data project solves a business problem at a scale that you can manage, get the foundations right in terms of the data you are utilizing and how it is treated and, finally, ensure that you draw upon collective assistance and inspiration as and when you require it will ensure that you have increased your chances of success in any data project that you undertake.

A massive thank you to the 5 legends who have contributed to this blog – Martin, Marta, Adam, Jasmina and Chris. Your expertise is very much appreciated.

Author Bio


As Director of Client Services for MBN Solutions, Rob has spent over two decades at the sharp end of Talent Acquisition practice for the Data sector. During this time, he has partnered with some of the UK’s leading data-driven businesses to deliver best-in-class talent solutions. In addition, working in an advisory capacity, Rob designed, built, and delivered the Data Lab’s MSc Placement Programme, has contributed to forums including Scotland’s AI Strategy and DMA Council and sits on University of Glasgow’s School of Maths & Stats Industrial Advisory Board. A regular data industry blogger and event host, Rob also now hosts a data leadership focussed podcast called Boss’n’Data and has been recognised by Data IQ as one of their 100 most influential Data and Analytics practitioners in UK organisations for two years running