This blog post was written by Jaskirat Mann (Senior User Researcher) and Sophie Strong (Senior Data Analyst).

Think of data like bread.

Artisan bread like a homemade sourdough is unique and tailored, but hard to slice, store, or use in different recipes. Great for personal use, but not easily shared.

Supermarket bread, on the other hand, is uniform and consistent. It may be less exciting, but it’s easy to package, scale, and use across many contexts.

Data collection works the same way.

Designing it to be structured and consistent like supermarket bread, means it can be easily analysed, shared, and used across the business. It may feel less bespoke, but it’s what makes scalable insights possible.

Often, we can fall into the trap of designing services, launching surveys, and capturing information. The result? Data that doesn’t quite meet user or business needs, becoming incredibly difficult to analyse. This isn’t just an inconvenience; it’s a major roadblock that can lead to skewed insights, wasted time, and ultimately, poor decision-making.

So, why does this happen, and more importantly, how can we fix it?

Examples of where we might go wrong

The core problem is that the way data is often captured doesn’t serve both user and business needs, primarily because the data analyst wasn’t involved early enough in the process.

Let’s look at some common pitfalls that cause data to “go stale”:

Example 1:

  • Free-text fields for categorical data: Imagine asking users, “Which training provider did you use for your teacher training?” and providing a free-text box: “SCITT,” “School Centered Initial Teacher Training,” “School Centered Training,” “PGCE,” “Post Graduate Certification,” “University training”.
    In this example, there were an array of responses meaning the same thing.
  • Why it’s a problem: An analyst then faces thousands of entries that need manual cleaning and standardisation. This painstaking process makes it incredibly difficult, and prone to errors, to perform trend analysis and segment your data accurately.
  • How to avoid it: Involve an analyst to suggest predefined options or dropdowns from the outset, balancing user freedom with data structure.

Example 2:

  • Missing forced choices: You might design a satisfaction survey with radio buttons from 1 to 5, but inadvertently allow submissions with blank responses.
  • Why it’s a problem: Suddenly, over 20% of your responses might have missing satisfaction scores. This directly skews your key metrics, making it hard to confidently determine true user sentiment.
  • How to avoid it: Analysts can advise on mandatory fields or validation rules to ensure complete data capture for critical metrics.

Example 3:

  • Date fields without validation: Asking “When did you do your teacher training?” with a free-text field can yield responses like “last year,” “4 years ago,” “2016,” or even date ranges like “03/09/2014-20/07/2015”.
  • Why it’s a problem: Analysts are left to interpret ambiguous entries or discard them entirely, leading to incomplete or unreliable datasets.
  • How to avoid it: Collaboration with an analyst can lead to implementing structured date pickers or specific format requirements to ensure consistency.

Because of these pitfalls, data analysis can take four times longer than expected. The final dataset requires heavy cleaning, with a significant amount of data either discarded or heavily transformed. Most critically, you can’t make confident decisions based on these insights. Without an analyst’s early input, even well-intentioned forms can lead to inconsistent, messy, or unusable data.

The Solution: Thinking in Data from the Get-Go

The solution is simple yet powerful: involve a data analyst when you’re first creating that ‘sandwich mould’ – the framework for your data collection. Interview them!

The core principle here is to understand that your data analyst is also a user. Just as you design for external users, you must design for your internal “users” who will interact with and interpret the data.

While data always needs some transformation to be “ready to analyse,” when we “penser en Data” (think in data) earlier, we significantly reduce the amount of transformation required and the time it takes.

Early involvement ensures that the data collected will be actionable, easier for policy and the wider business to understand, and will meet both user and business needs. It even allows data to be used to predict and influence future decisions.

The Balancing Act: User Needs vs. Data Needs

However, it’s not always straightforward. Structured data is undeniably easier to analyse and share across the business. However, sometimes users genuinely need to input data in a free-text format to express nuanced information. This is where the art of balancing user needs with business and data needs comes into play.

A resettlement worker at the Ministry of Justice once said, “You’re limiting the checklist in my head”. And a probation services officer shared “You won’t be able to think of everything in your drop down […] I once had a man arrested with a cat in his pocket. I had to make a referral to RSPCA […] I need to be able to input my rich notes to inform safeguarding measures”. These examples perfectly illustrate why some qualitative, free-form data is crucial. You can’t always anticipate every unique scenario with predefined categories.

How to achieve balance:

For some questions, a simple “yes/no” or a forced choice might provide a quick win. For instance, an easy-to-understand percentage of users self-identified as being in a gang. This provides immediate, structured insight.

For the unique, flavourful details, allow for more open input where necessary. Then, plan for further analytics like word clouds or text analysis to extract insights from this qualitative data.

Why it works: You are still able to analyse the data and share it widely, like “bread from the mould,” but you also retain the “flavour”. The extra, rich data needed to meet user needs.

Dashboards and the Full Data Sandwich

Often, dashboards come into play far too late in the process. Many teams find themselves having to go back and reshape a service just to collect the right data for a dashboard. This is like forgetting to “prove” the dough; the data becomes rushed, flat, and underbaked.

How to improve: As early as possible, talk to the users who will need the data, especially those who rely on dashboards.

By thinking about dashboard users and business needs early on, you can avoid that awkward scramble later and bake something that actually rises

Think of your data as a delicious sandwich, with each ingredient playing a vital role.

Data Analysts: Designers in Disguise

It’s crucial to remember that your data analyst isn’t just a number-cruncher; they are also, in part, designers. While a UX designer might design the product interface, a data analyst is equally capable of designing the dashboard that brings the data to life.

Actionable Insights and Collaborative Success
When you actively involve your data analysts early, and work in collaboration as a multidisciplinary team, the benefits are immense:

  • The insights your data analyst provides will be actionable
  • It becomes significantly easier for policy makers and the wider business to understand the data
  • Data will consistently meet both user needs and business needs

Most powerfully, your data can be used to predict and influence future outcomes.

Going forward, remember to ask everyone involved in the data journey: “How would you like your sandwich?”. By fostering cross-profession collaboration and embracing a data-informed, user-centred mindset in projects, we can all contribute to creating data that meets both user and business needs.