What you want from your data is impact, efficiency, or income. What you're getting are dashboards

If you want to get value from your data, don't just focus on the technology

Your organisation has data, maybe some dashboards, maybe even some machine learning models. You know that the dashboards and machine learning models are supposed to help you get something that you want, like impact, or efficiency, or money. But is that actually happening?

I’ve worked at places that had a little bit of data and a few dashboards, and places that had lots of data, many dashboards, and some machine learning models. In neither case was it clear if any of the data or dashboards or models were helping the organisation to progress toward its goals. The responses to Data Orchard’s 2025 data maturity survey indicate that this isn’t unusual: only five percent of organisations reported being at the ‘mastering’ stage of data maturity where they are able to use data purposefully to boost their internal operations and service delivery. 

Why is so much time and money spent on data, but so little value coming back? One problem is the misleading hype around technology and data, which leads you to believe that data will solve many of your problems if you just buy some expensive technology and, of course, make dashboards. The message is that technology will get you the outcomes you want. 

However, the reality is that it takes an interdisciplinary process, not just a set of shiny tools, to get value out of data. (When I say ‘get value from data’, ‘value’ refers to anything that is important to you, like impact. ‘Value’ can mean money but is not limited to money). I’ve sketched out the process in the diagram below. Successfully getting all the way through the process takes more time and effort than is often anticipated. Although there are a few occasions when technology can get you all the way to the end of the process, most of the time it can’t because human involvement is necessary. In those cases, the most difficult part of getting value from data usually isn’t anything to do with technology; it’s changing people’s behaviour and an organisation’s culture. Yikes. So in the rest of this article, I’m not going to focus on technology. I’m instead going to focus on the orange boxes, which have nothing to do with technology, but which are critical to get right if you want to get to an outcome that you care about.

To get value from data, you first need to know what value you want - what problem do you want to solve?

If you want to get to a destination, you need to know what your destination is before you start travelling. Setting the destination for data means defining what problems you want to address and the questions that data will be able to answer. For example, ‘How many people are we reaching with our programs?’ or ‘Which patients are the most likely to drop out of their treatment regime?’ or ‘How many staff do we have employed right now?’. However, vague statements like ‘We need to use data for reporting’ and ‘We want to use data for decision-making’ are common. Without any further details, those broad statements don’t provide the necessary direction for what data you need to produce or collect, and how that data will need to be processed and analysed. It’s like saying to a travel agent ‘I want to go to Africa’ when you want to visit your friend who lives in the outskirts of Arusha in Tanzania. Maybe you’ll get to her house, or maybe you’ll end up in a market in Accra, Ghana, or maybe somewhere in the Congo jungle. Who knows? 

If you find that you can’t define or agree what value you want to get from your data, it may be a sign that your business objectives are unclear, or that your objectives are clear but no one knows how to reach those objectives. A clearer business strategy and roadmap may be what you need, before you focus on data.

Don’t use data just because you can, if there’s an easier way to solve the problem

Around a decade ago I was working in data science, which was the world’s latest obsession at that time. Every tech company wanted to show off how much data it had and what it could do with the data. I remember one night there was a minor earthquake in the Bay Area in California. A few days later, an up-and-coming fitness wearables company proudly announced to the world, ‘Our data shows that people wake up at night when there’s an earthquake!’. I could sort of see why some people might get excited about using data to find out that fact. However, being a common sense enthusiast far more than a tech enthusiast, my own reaction was ‘Um, isn’t that obvious?’. The point is that just because you can use data to answer a question, that doesn’t mean that you should use data to answer that question if there’s an easier way to get to the answer. If data will give you useful information that goes beyond common sense, then use the data. But otherwise, just use common sense. It’ll be cheaper and quicker, I promise.

Make sure that someone is able and willing to act on the insights that you get from data

Sometimes, data can be incorporated into automated workflows that will automatically make routine decisions and carry out the appropriate action. An example is sending a special thank you email to a supporter once their cumulative donation amount passes a threshold. There’s no need for a human to firstly look at each supporter’s donation history, and then send an email, because it’s a straightforward decision and task. However, for decisions that are strategic, complex, or high-risk, a human needs to be involved. Specifically, a human will need to look at the data, and then take some action based on the data.

Sometimes, actions need to be enabled

Sometimes, people may be willing to act but aren’t able to. I wrote about some of these examples in a blog called ‘AI doesn’t save lives, people do’, based on my experience of working in the field of global digital health: machine learning models might be able to diagnose or predict which health problems someone will suffer from, but the health systems in many countries can’t do anything in response because there aren’t enough doctors, nurses, medication, or equipment to provide the necessary treatments. Doctors and nurses would love to save more lives and relieve the suffering of more of their patients, but if all you give them is a machine learning model to diagnose a disease but not the means to treat that disease, they cannot.

It’s perfectly reasonable to use data as evidence for advocacy, for example to advocate for more nurses to be trained and properly paid. But if you’re the one that is trying to improve outcomes directly, then knowing what outcomes you have the capability to change, and focusing your data on informing decisions about how best to use that capability, will get you further towards your mission than finding out interesting insights about outcomes that you cannot change. 

If you’re expecting someone else to act on the insights from data, make sure they are motivated and therefore willing to do so

It would be nice if everyone you worked with shared a common goal, and agreed on how to reach that goal. Sometimes that’s the case, but sometimes it’s not. Politics and personal interests play a huge role in how people make decisions. There are a lot of people out there who perfectly understand the data that’s in front of them, but who don’t make decisions based on that data because that won’t get them to their goal. 

Humans are human, not machines, which means that we are all motivated by different things. (No wonder AI might get irritated with us and kill us all). If you want someone to make their decisions based on data, but they repeatedly don’t use data, then finding out what really motivates them, and how to make your goal of data-informed decision-making overlap with whatever motivates them, will be far more effective than simply shoving another dashboard at them. This applies even if the new dashboard is interactive and updates in real time. A good way to test the water before embarking on any major data initiatives, is to find out whether there is any existing data around that is trustable and accessible. If there is trustable, accessible data currently available, is anyone using it? If people are using data, then that’s a sign that the right motivation, behaviours, and culture already exist. Given more data, people will make data-informed decisions even more of the time. Amazing. But if no one’s currently using data even though they could, then simply providing them with more data isn’t going to change that; you’ll almost certainly need to do some advocacy, training, and navigation of political roadblocks, to get where you want.

And the tech? Technology is necessary but not sufficient if you want to get value from data

Technology tools are necessary. Unless you’ve only got a tiny amount of data that you can manage on sheets of paper, you are going to need tools to help you produce, collect, clean, analyse, and present your data. I haven’t written about the technology here though, because it gets enough coverage. What isn’t highlighted enough is what I have written about, which is what needs to happen before you start producing data, and what needs to happen after all the dashboards and pretty outputs are made. Not thinking through or investing enough in those steps will prevent you from getting value out of your data, even if all the technology has been exquisitely set up. Once you understand that, you can set realistic data ambitions, and plan and resource your data initiatives effectively. Good luck!

Wait, isn't there going to be a section about AI?

I know it’s mandatory these days to make generative AI the focus of every discussion, but I’m not going to because I don’t want to distract from the main message about the importance of human behaviour if you want to get value out of data. So there. 

This blog was written by a human (me) who likes sharing their thoughts and experience with other humans (you, hopefully). If you are a human: Hello 👋! Thanks for reading. I hope you found something interesting in this article. 

Contact me if you'd like to discuss anything I’ve written, or if you’d like to hire me to help fix your data.