Brett LindenbergData-driven decision making means basing a decision on analyzed evidence, such as sales figures, survey results, test scores, or operational records, instead of on intuition, habit, or the opinion of the most senior person in the room. A team that works this way names the decision, agrees on the evidence that would change its mind, checks the data, compares the options against a baseline, decides, and later checks whether the result matched what it expected. Judgment still matters: data narrows the options, and people weigh what the data cannot show. In the short demo below, you review completion, quiz, and on-the-job data from a training module and choose what the team should change next.
What data-driven decision making means
The term names a method, and a team can follow it with nothing more than a spreadsheet. A decision counts as data-driven when four things are true:
- You wrote down the question before you pulled the data.
- The evidence relates to that specific decision.
- You compared the numbers with a baseline period or a similar group.
- You recorded the decision and the result you expected, so someone can check them later.
Some teams say "data-informed" instead, to stress that people still make the call. The two terms describe the same practice, with "data-informed" putting more weight on judgment.
In US manufacturing, the practice spread quickly in the late 2000s. Using Census Bureau survey data from a large, representative sample of US manufacturing plants, Erik Brynjolfsson and Kristina McElheran found that the rate of data-driven decision making nearly tripled, from 11% of plants to 30% of plants, between 2005 and 2010 (American Economic Review, 2016). In a companion Census Bureau working paper, they found that the average value added of plants that adopted it later was 3% greater than that of non-adopters, after controlling for production inputs. Adoption was more common in larger plants and in plants that also had information technology and educated workers, so the tools and the skills to use them tended to arrive together.
What's the difference between data-driven decision-making and traditional decision-making?
Data-driven decision-making is based on data analysis. Traditional decision-making is based on intuition and personal experience. If you are working through a quiz that offers four options, this is answer B. The other options reverse the two definitions, treat them as the same thing, or claim that traditional decision-making is gaining ground.
| Aspect | Traditional (intuition and experience) | Data-driven |
|---|---|---|
| Basis | Experience, gut feel, precedent, seniority | Analysis of data relevant to this decision |
| Speed | Fast, because no data work is needed | Slower at first, faster once the data and reports exist |
| How others can check it | Hard to audit, because the reasons often go unstated | The reasoning and the expected result are written down |
| Main risk | Bias and overconfidence, especially in unfamiliar situations | Poor data, or reading a correlation as a cause |
| Works best when | The situation repeats and the decision maker has had years of fast feedback | The situation is new or the stakes are high |
Intuition and evidence can work together. Daniel Kahneman and Gary Klein reviewed when expert intuition can be trusted (American Psychologist, 2009). They concluded that to judge an intuition, you check how predictable the environment is and whether the person has had the chance to learn its patterns. They also found that feeling confident is not a reliable sign of being right. A veteran nurse's quick call on a familiar ward has years of feedback behind it, while a manager's hunch about a market the company has never sold in has no such track record. Good decisions usually use both: data to frame the choice, and experience to weigh what the data leaves out.
How to make data-driven decisions in six steps
This process works for a budget call as well as a training revision. The running example is illustrative: a customer support team deciding whether to add weekend coverage.
1. Name the decision, the owner, and the deadline
Write the choice as options: "Add Saturday and Sunday coverage, add Saturday only, or keep the current schedule." Name the person who decides and the date of the decision. A deadline keeps the analysis from running on indefinitely.
2. Agree on the evidence that would change your mind
Before anyone looks at the numbers, write down the measure and the threshold. For example: "If weekend tickets are more than 15% of weekly volume, and customers who wait more than 24 hours give lower satisfaction scores, we pilot weekend coverage." Setting the rule first stops the team from picking the numbers that support what it already wanted.
3. Gather and check the data
Pull the data from the system that owns it, here the help desk and the satisfaction survey. Then check it:
- Does "ticket" include spam and duplicates?
- Is the period long enough to show seasonal swings?
- Are any days or channels missing?
- Do two reports define the same metric differently?
4. Compare against something
A number on its own means little. Compare it with a baseline period, a similar group, or the threshold from step 2. When you can, run a small test: one region gets weekend coverage for eight weeks and a similar region does not. A simple before-and-after comparison cannot separate your change from a holiday or a product release in the same weeks.
5. Decide and record why
Write the decision, the evidence, the result you expect, and how confident you are. One paragraph is enough. The record shows what you knew at the time, which is the fair standard for judging a decision later.
6. Check the result and adjust
Set a review date and compare what happened with what you expected. If satisfaction rose in the pilot region and stayed flat in the comparison region, extend the coverage. If nothing moved, stop, and record what you learned for the next decision.
Examples of data-driven decisions at work
Most teams make decisions that data can improve. The table shows common ones, with the check teams most often skip.
| Team | Decision | Data that informs it | Check before you act |
|---|---|---|---|
| Sales | Which leads reps call first | Win rate and deal size by lead source | Enough closed deals per source to compare |
| Marketing | Where to move next quarter's budget | Cost per qualified lead and pipeline by channel | How the attribution model credits each channel |
| Operations | How to staff each shift | Volume by hour, service level, overtime hours | Seasonal and holiday patterns |
| Human resources | Which manager skills to train | Engagement survey items, exit interview themes, team turnover | Survey response rate for each team |
| Customer support | Which help articles to write next | Ticket topics and repeat contacts | Whether agents tag tickets the same way |
| Learning and development | Which part of a course to fix | Missed questions, drop-off points, errors on the job | Whether the same people show the problem at work |
A documented example: Google's Project Oxygen
Google has described how it used data to decide what to teach its managers. In a multi-year study called Project Oxygen, the company analyzed more than 10,000 data points about its managers from performance reviews and feedback surveys, along with hundreds of pages of interview notes. The study identified 10 key behaviors of highly effective Google managers. According to an August 2025 article on Google's re:Work site, those behaviors now guide feedback and performance reviews and underpin leadership development programs at The Google School for Leaders. Google chose what to teach its managers from evidence about what its most effective managers already did.
Common mistakes in data-driven decision making
- Starting with a conclusion and then searching for numbers that support it. Write the decision rule first, as in step 2.
- Measuring what is easy to count. Course completions and page views are simple to collect and rarely answer the question.
- Reading a correlation as a cause. Employees who attended a workshop may perform better because their managers chose to send their strongest people.
- Trusting a small sample. A rise from 4 complaints to 6 complaints is a 50% increase in complaints and is probably noise.
- Turning a measure into a target that people can game. If agents are judged only on call length, some will end calls before the problem is solved.
- Waiting for perfect data. Keep the deadline from step 1, decide with what you have, and state the uncertainty.
Data-driven decision making training and courses
People look for data-driven decision making training for two reasons: to build the skill themselves, or to get a team using data in everyday decisions. Both need the same core skills.
How to upskill for data-driven decision making
Learn these skills in roughly this order:
- Turn a vague goal into a question with a measurable outcome.
- Read charts and summary statistics: averages, medians, spread, and trends over time.
- Use a spreadsheet to filter data, build a pivot table, and write simple formulas.
- Compare groups fairly with baselines, comparison groups, and A/B tests.
- Judge data quality and whether a sample is big enough.
- Present a recommendation, with its uncertainty, to the person who decides.
People who will build reports also need SQL or a business intelligence tool. Managers who mainly read reports can skip those and focus on asking good questions of the analysts who build them.
Data-driven decision making courses
These courses cover the skills above at different depths. We checked each one on the provider's site in October 2026.
| Course | Provider | Length and level | Cost | Good for |
|---|---|---|---|---|
| Data-driven Decision Making | PwC, on Coursera | About 9 hours in 4 modules, beginner | Free enrollment option; the certificate requires a purchase | Business staff new to analytics. Ends with a project that asks you to make a decision in a simulated business case. |
| Ask Questions to Make Data-Driven Decisions | Google, on Coursera (course 2 of the Google Data Analytics Professional Certificate) | About 12 hours in 4 modules, beginner | Free enrollment option; the certificate requires a purchase | Framing questions, spreadsheet work, and communicating with stakeholders |
| Business Analytics | Harvard Business School Online | 20 to 24 hours in 4 modules, self-paced | $1,949 | Managers who want statistics, prediction, and experiment design for business decisions |
| Google Analytics courses 101 to 301 and the Google Analytics Certification | Google Skillshop | 4 courses, beginner to advanced | Free. Google's Skillshop FAQ says Skillshop is available at no cost. | Marketing teams that measure campaign results in Google Analytics |
A course teaches the vocabulary and the methods. The habit forms when people apply the six steps to their own decisions, which is what an internal program adds.
Training for marketing teams
Marketing decisions are mostly about budget and channels. Pair a measurement course with practice on your own campaign data, and agree on what counts as a qualified lead and as pipeline before anyone compares channels. Our guide to data-driven content marketing covers which content metrics to track, and our guide to content marketing ROI shows how to connect them to revenue.
How to build an internal data-driven decision making program
- Collect a handful of real decisions your team made in the last year, with the data that existed at the time.
- Teach the six steps, using one of those decisions as the worked example.
- Give learners a realistic data set and a decision. Ask them to choose and explain why, then show what happened.
- Give feedback on the reasoning as well as the choice. A right answer reached with a weak method can fail on the next decision.
- A few months later, review the team's decision records. Look for thresholds written before the analysis and for review dates that someone kept.
Interactive video fits step 3. In Mindstamp, you can show a dashboard or a short situation on video, pause for a decision, and use branching to play the consequence of each choice. The demo at the top of this article works that way. A question can ask which data the learner would check next. The report records each learner's answers and path with timecodes, so you can see which wrong choice is most common and fix the lesson that leads to it. For design advice, see our guide to scenario-based training, or draft a first version with the branching scenario generator.
Data-driven training: using data to make L&D decisions
Data-driven training applies the same method to learning and development. L&D teams make four recurring decisions, and each one needs different evidence.
| Decision | Evidence | Where it comes from |
|---|---|---|
| What to build | Performance gaps tied to a business metric | A training needs assessment, manager interviews, operational data |
| What to fix in a course | Questions most learners miss and sections where they stop watching | Quiz and video reports |
| Whether it worked | Behavior on the job and the business result, compared with a baseline | Observation, CRM or help desk data, HR systems |
| What to retire | Low use, no change in behavior, outdated content | LMS and video reports, manager feedback |
Few L&D teams reach the third row. In The Future of Evaluating Learning and Measuring Impact, a 2025 report from ATD and Zensai, only 4% of the talent development professionals surveyed rated their organization as excellent at using learning data for business decisions related to evaluating learning outcomes.
Take the case in the demo at the top of this page. Completion is high, so celebrating the completion rate ignores the evidence that matters. Replacing the whole program throws away the parts that work. You can trace the problem to one section: 48% of learners miss the handoff question, and the matching error persists at work. The data-driven choice is to revise that practice, retest it, and watch the error rate. Even then, the data supports a hypothesis. A comparison group, or the error trend before and after the change, tells you whether the revision worked.
What interactive video adds to training data
A standard video report shows how many people pressed play and how far they got. When questions sit inside the video at the point each concept is taught, you get evidence you can act on. Mindstamp's reporting shows:
- Watch progress for each learner
- Every answer and click, with its timecode
- Scores for graded questions and an answer breakdown for each question
- Drop-off by section, on Core and higher plans
- The branch path each learner took
- CSV exports of views, viewers, and interactions, including a completion report with one row per learner (see exporting training records)
One Mindstamp customer used this kind of data to decide what to change. Diamond Scheduler, which has made sports scheduling software since 1999, added short quizzes after the main sections of its tutorial videos, buttons that link to its help documentation, and a feedback prompt at the end. Its reports showed where viewers stopped watching and where quiz answers showed confusion. The team used that evidence, with the viewers' written feedback, to find the topics that needed clearer tutorials and to rework parts of the software. The Diamond Scheduler case study describes the setup.
If you have no baseline yet, our platform data gives a reference point. Across 8.07 million views of 18,315 Mindstamp videos (2018 to October 2026), the average view played 46.5% of the video. These are mostly training, marketing, and sales videos made by Mindstamp customers, so treat the figure as context and compare each video with its own history first.
For the full measurement method, read how to measure training effectiveness and our comparison of training evaluation methods. To put a money value on the result, see training ROI. To study what learners do inside the content, read what behavioral analytics is and our guide to interactive video analytics.
Common questions
What is an example of data-driven decision making?
A support team finds that weekend tickets make up 18% of weekly volume and that customers who wait more than a day give lower satisfaction scores. It pilots weekend coverage in one region for eight weeks, compares satisfaction with a similar region, and extends coverage only if the pilot region improves. This example is illustrative.
True or false: AI is a technology that enables machines to learn and perform tasks that would normally require human intelligence
True. NASA describes artificial intelligence as computer systems that can perform complex tasks normally done by human reasoning, decision making, and creating. The federal definition NASA cites includes systems that can learn from experience and improve their performance when exposed to data sets. In data-driven decision making, AI tools can find patterns and draft an analysis, but a person still owns the decision and its trade-offs.
Is data-driven decision making always better than intuition?
No. When the situation is familiar and the decision must be made in seconds, experienced judgment works well and the data work may not pay for itself. Use the full process when a decision is costly, hard to reverse, new to the team, or disputed.
How long does it take to learn data-driven decision making?
The introductory courses in the table above take about 9 to 24 hours. The habit takes longer than the course. It forms when people use the six steps on real decisions and someone reviews their decision records.
Start with one decision
Pick one decision your team will make in the next month. Before anyone opens a spreadsheet, fill in this record with the person who will decide.
| Field | Example |
|---|---|
| Decision and options | Add weekend coverage, add Saturday only, or keep the current schedule |
| Owner and deadline | Support director, by the end of the quarter |
| Evidence that would change our mind | Weekend tickets above 15% of weekly volume, and lower satisfaction scores after waits of more than 24 hours |
| Data sources | Help desk tickets and satisfaction surveys from the last 12 months |
| Comparison | Eight-week pilot in one region, with a similar region as the comparison |
| Expected result and confidence | Satisfaction after weekend contacts rises by 5 points on a 100-point scale; medium confidence |
| Review date | Two weeks after the pilot ends |
If the decision is about a training course, the training revision example shows the same pattern as a short interactive video your team can practice with.
