Data-Driven Decision-Making
Data-Driven Decision Making
What is data-driven decision-making?
- Decide with data, not gut feeling
- Overcoming HiPPO (highest-paid person’s opinion)
- A core DX mindset
Once data-driven decision-making takes hold, even the way people talk in meetings changes. 'In my experience, this banner will work better' becomes 'Over the past four weeks the click-through rates were 1.2% and 2.7%, and once you count checkout the gap shrinks to 0.4 points.' The bigger change is that decisions leave a record. If you note even one line on which numbers you looked at and why you chose that option, then when results disappoint three months later, you can ask which assumption was wrong instead of whose fault it was. That is usually what separates organizations that learn quickly from failure from those that don’t.
In practice, start by sorting decisions by their nature. For reversible decisions (email subject lines, shipping notice wording, coupon amounts), testing is faster than meetings. For hard-to-reverse ones (exiting a business, multi-billion-won equipment investments), use data to narrow the options to three or four, then leave the final step to executive judgment. A/B tests have conditions too. With only a few thousand weekly visitors, a 1–2 point difference is not statistically distinguishable, so it is better to wait 3–4 weeks for the sample to build, or to replace the test with customer interviews. Metrics should also be read in pairs: lagging indicators like revenue and churn alongside leading indicators like 'usage in the first two weeks.' And separating the areas where data decides (data-driven) from those where it only informs (data-informed) in advance cuts unnecessary debate.
There are four places where things most often go wrong in practice. First, a company has 30 dashboards but only five monthly users. The team produced reports; it did not change any decisions. Second, reading correlation as causation. If you conclude that 'customers who use coupons repurchase more' and send coupons to everyone, you add discounts for people who would have bought anyway and cut your margin. Third, the moment a metric becomes the target, it breaks. Make the number of support cases a KPI, and three-minute calls start getting split into two cases. Fourth, 'pull more data' becomes a shield for avoiding decisions. Data-driven is not a way to postpone decisions; it is a way to be wrong in small ways and fix it fast.
Jaryong KooMarketing PhD and authority in AI, data analytics, and personal branding
John KimAuthor of B2B Marketing Design and CEO of Nanum ConsultingA management approach that emerged alongside big data and analytics, and is widely seen as a core part of DX culture.
Let’s walk through a hypothetical case. Company A sells health supplements online, with 90 employees and annual revenue of KRW 42 billion (≈ US$31,423,000). Its repurchase rate fell from 38% to 26% in two years, yet the weekly meetings kept repeating the same lines: 'the economy is bad' and 'competitors are discounting aggressively.' The CEO told the team to find the cause in the data within three months.
- When the CRM lead split purchase history into monthly cohorts, customers who reordered within 45 days of their first purchase had a one-year retention rate of 61%, versus 14% for those who passed that point. "So the problem wasn’t the economy. It was day 45."
- A marketer randomly split 12,000 first-time buyers in half and ran a four-week A/B test sending a dosage-reminder text on day 30. It started with plain informational copy, no coupon.
- The result: a 19.4% reorder rate for the text group versus 15.1% for the control, a 4.3-point gap. A separate group that also got a KRW 3,000 (≈ US$2) coupon reordered 0.6 points more often, but its average order value was KRW 8,200 (≈ US$6) lower, so the team dropped the coupon idea.
- The CS lead objected, saying "More texts will cause a flood of opt-outs." Over four weeks, though, the opt-out rate rose just 0.2 points, from 0.7% to 0.9%. Once the numbers were in, the CS lead was the first to back a company-wide rollout.
- Six months later, the repurchase rate rose from 26% to 33% and monthly revenue grew by about KRW 180 million (≈ US$134,700). The bigger change, the CEO says, is that the first question in meetings became "Which cohort’s number is that?"