90% of Data Projects Fail: Human-Centered Data Matters

This article discovers why nearly 90% of data projects never deliver real value and explores the overlooked human-centered data analysis.

Overview

We live in an era where being data-driven has become a badge of credibility. Organizations proudly talk about the dashboards, AI strategies, predictive models, and automation they have invested and reaped benefits from. As the internet would inform you, nearly every Fortune 1000 company is increasing its investment in data and AI to stay agile and competitive. And yet, despite the unprecedented access to the quality and quantity of data, a vast majority of analytics and AI initiatives do not make it to production or can’t make a lasting impact. 

Data models are created, insights are shared, decks are applauded and then quietly forgotten only to become (what I like to call) trashboards.

In this day and age of machines taking over our decision-making capabilities, the problem isn’t a lack of data, talent, or tooling but it’s the human that we are starting to forget to talk to.

This is where Human-Centered Data Analytics becomes not just relevant, but essential.

What is a Human-Centered Approach?

Data is nothing but the digital traces of human interactions. A human-centered approach can enhance the choices data scientists make every day, by making the process more transparent, asking questions, and considering the social context of the data. 

A human-centered approach asks a very simple question:

Who is this for and how will it actually be used?

Now think about it this way by asking “What can we predict from this data?”, the human-centered approach makes us want to ask “What should we help people understand or decide with this data?”

Human-Centered Data Analytics is the concept of understanding how people interact and make sense of social situations, enabling humans to explore and gain insights, and design data models with the end-user in mind (not just the business).

At its core, human-centered Data Analytics means designing models and metrics with the end-user in mind, not just the business KPI. It asks us to improve the everyday decisions data professionals make: how we frame problems, what features we engineer, which metrics we optimize, and how we communicate the solutions to those problems.

Why Human-Centered Data Analytics Is the Future

As the world becomes more technically sound and business-driven, we as a society have a declining social and behavioral relevance. Organizations, regardless of their line of business, have reduced people to profits and probabilities. We forget that every dataset comes from someone deciding to buy, click, move, vote, or opt out and end up treating these behaviors as a signal instead of a story. 

Ignoring that human context can lead to optimizing the wrong outcome entirely. The human-centered approach introduces a new dimension and forces us to ask:

• Who benefits from this model?
• Who might be harmed?
• What assumptions are baked into the data?

How Can You Practice Human-Centered Data Analytics In Your Work

Inclination toward a human-centered approach is not a newfound love.

Early in my career, I was deeply interested in Human–Computer Interaction (HCI). A field that studies how people design, use, and interact with technology. Working with HCI, without a huge realization, I developed an attitude to prioritize understanding the human cognition, behavior, and social context when solving a problem.

So even though we are in the field of data and AI now, the human-centered attitude should become our second nature, and here’s how we can practice Human-Centered Data Analytics at the workplace.

1. Start With People, Not Metrics

Our mindset should not be fixated on designing pretty dashboards because that is a tangible outcome and that would get you visibility. But , as a data professional, we should realize that dashboards don’t create value on their own. Decisions do.

You need to design your analysis around the decisions people can make from an analysis, not mere dashboards. Before defining any steps or KPIs for your analysis or dashboard, you should ask:

• Who would use and act upon these insights?
• What decision are they trying to make?
• What constraints do they face?

Asking these questions to the impacted people upfront usually defines the next steps, removing guesswork and ensuring that the metrics they share actually serve the problem, instead of hoping that the metrics which we have are true for the problem, sometimes they might be but asking them and finding solution around those metrics and your insights are the true problem solver.

2. Interrogate the Problem’s Origin

Every problem has a history. 

Human-Centered Data Analytics asks us to think of questions relevant to the problem and take a small pause before gathering, scraping, and manipulating the necessary data. You should document assumptions and known biases, not just as footnotes, but as part of the analysis. Ask questions like:

• Where did the problem originate? Under what conditions?
• What behaviors are missing or underrepresented?
• What data can answer this problem in the asked context?

This creates transparency and sets realistic expectations for how insights should be interpreted.

3. Design for Understanding, Not Just Accuracy

A data model with some 94% accuracy that no one understands rarely delivers impact. 

But, if you pair the output from that same data model with a short narrative that explains why the result exists, not just what it is, test for yourself how that delivers impact. Human-centered analytics pushes you to translate technical language into simple human understanding.

Once your data model is ready, ask:

• Can a non-technical stakeholder explain your insights after hearing it once?
• Can you replace feature-importance charts with decision-oriented visuals (e.g., “If X increases, here’s what changes”)?
• Can you trade marginal accuracy gains for clarity?

The human-centered approach lets you design models that have an improved adoption along with precision.

4. Account for What the Data Cannot See

You cannot emphasize enough how much this will allow you to grow in this career! Being able to see the short-comings of a dataset, anticipating questions on those gaps and preparing to answer that gap will be a key driver for your future, to go up the ladder.

But hey, no points for guessing where that comes from, the human-centered approach of working with data! 

A human-centered approach allows you to explicitly acknowledge blind spots. As you familiarize yourself with a dataset, start documenting the known data gaps, behavioral patterns of the dataset, and call out assumptions during presentations instead of letting them remain implicit. You could ask:

• What does this data not show?
• What group or behavior is underrepresented?
• Can the judgment made by decision-makers from these data insights stand itself when gaps are significant.

5. Design for Ethical Impact, Not Just Performance

Working with sensitive data makes ethics unavoidable. But thanks to the human-centered approach, it allows us to treat ethics as a design constraint, not a compliance checkbox. Ask ethical questions early and plan for it, and not as an after-deployment thought, like:

• What happens if this data model is not the best fit?
• Who will bear the cost of errors?
• How will feedback be incorporated?

By planning for these scenarios upfront, we can build solutions that are not only effective, but responsible and more sustainable.

6. Build Feedback Loops Into the System

As a part of the workforce, we all know the importance of feedback and integrating that into our work and not just from a data perspective, but holistically, the human-centered approach pushes me to treat solutions as evolving systems rather than one-time deliverables.

According to the human-centered approach, your structure for adding feedback loops into your systems is a 3-step process: 

1. Define success metrics beyond launch (such as adoption, overrides, and stakeholder confidence)
2. Schedule recurring check-ins with users and stakeholders to understand how insights are being used or ignored
3. Incorporate qualitative feedback into future iterations, not just quantitative performance metrics.

The results from step 2 above on how insights are being used or ignored might not always be what you wished for. we hear a lot of “oh we don’t use that tool anymore” for tools that we had built in the past. So to avoid that, keeping the human-centered approach in mind, ask questions before and after the tools are used

• How will this analysis be evaluated and used once it’s in use?
• Should this be a one-time deliverable or a robust tool?
• How many users stopped using the tool only after a couple of uses? What changed?

Closing Thoughts

Data Is Powerful Because People Are.

The future of analytics isn’t about more data, bigger models, or faster pipelines. It’s about wisdom!

Human-Centered Data Analytics reminds us that data is powerful not because it is objective, but because it reflects human life in all its complexity. When we design analytics with empathy, context, and responsibility, we don’t just build better models but better systems!

And that matters more than ever.

Ready to Master Human-Centered Data Analytics? 

Join WhiteScholars Data Science Program

Want to turn these principles into career superpowers? WhiteScholars Data Science Program equips you with people-first skills: Ethical AI, stakeholder storytelling, feedback-driven projects, and production-ready analytics.

Why WhiteScholars Aligns Perfectly:

• Human-Centered Curriculum: HCI-inspired modules on user-focused modeling, bias auditing, decision dashboards (e.g., “Predict churn? Design for sales teams’ actions”).
• Real-World Practice: 5+ projects like ethical fraud detection, inclusive recommendation systems and deployed with feedback loops.
• Mentorship: 1:1 with pros teaching “Who benefits?” framing, gap analysis, narrative viz.
• Outcomes: 90% placement, portfolios proving impact (not just accuracy).
• Community: Alumni network for ethical AI discussions, North-East scholarships.

FAQ’s

1. What is Human‑Centered Data Analytics, and how is it different from traditional data analytics?

Human‑Centered Data Analytics puts people—users, stakeholders, and impacted communities—at the core of every analytical decision, rather than treating them as passive consumers of dashboards or KPIs. It differs from traditional analytics by focusing on why the data exists, who will use it, and how it shapes decisions, not just on model accuracy or technical elegance.

2. Why do so many data and AI projects fail to deliver real impact?

A large share of data projects fail because they optimize for metrics and dashboards instead of human decisions and context. Teams often build complex models without deeply understanding stakeholders’ constraints, workflows, or the social and ethical consequences of their outputs, leading to “trashboards” that are praised in meetings but ignored in practice.

3. How can I start applying a human‑centered approach in my own data projects?

You can start by asking human‑centered questions early: Who will act on these insights? What decisions are they trying to make? What constraints do they face? Then, design your problem framing, metrics, models, and visualizations around those answers, not around available data or model performance alone. Iterate based on qualitative feedback, not just technical metrics.

4. Isn’t “human‑centered” just slowing down data projects by adding extra questions and ethics checks?

Adding human‑centered thinking may shift where you spend time, but it usually speeds up overall impact by reducing wasted effort on models and dashboards that never get adopted. Treating ethics, context, and feedback loops as core design constraints—not add‑ons—helps you build systems that are more trustworthy, usable, and sustainable, which often pay back quickly in adoption and value.

5. How does Human‑Centered Data Analytics relate to ethical AI and responsible data practices?

Human‑Centered Data Analytics is a natural bridge to ethical AI because it forces you to ask who benefits, who might be harmed, and what assumptions are baked into the data before you even build a model. By designing for transparency, fairness, feedback, and accountability from the start, it turns ethics from a compliance exercise into an integral part of how data solutions are conceived and evolved.