From Business Question to Data Question

Is social working?

In September 2025 Alpenkorb roughly doubled its monthly spend on social ads. Lena's predecessor called the push a success. The CEO asks her a simple question: is social working? The question is fair, but no table can answer it, because nobody has said what working means, for whom, and compared with what.

A data question is a business question with every word pinned down. A useful checklist has four parts.

• Measure: which number? Sessions, orders, new customers, revenue?
• Population: which rows count? All sessions, or only those from social ads? Delivered orders only?
• Period: exact start and end dates, including whether the end date is inside or outside.
• Comparison: against what? The period before, another channel, a target?

Pinning it down

Lena writes: Since the push began on 1 September 2025, how many sessions per month came from social ads, how many of them ended in an order, and how does that compare with January to August 2025? Now every word maps to a column. The sessions table has one row per visit to the website or app, with session_date, source and ordered, which is true when the visit ended in an order.

SELECT COUNT(*) AS sessions,
       SUM(CASE WHEN ordered THEN 1 ELSE 0 END) AS orders
FROM sessions
WHERE source = 'social'
  AND session_date < '2025-09-01';

CASE WHEN ordered THEN 1 ELSE 0 END turns true into 1 and false into 0, so the SUM counts the visits that ended in an order. For January to August 2025 the query returns 1,360 social sessions and 132 orders.

What the answer looks like

Over those eight months that is 170 social sessions and 16.5 orders a month. In the seven months from September 2025 to March 2026, social brought 370.6 sessions and 19.9 orders a month. Sessions more than doubled; orders rose by about a fifth. The share of social visits that ended in an order fell from 9.71 % to 5.36 %.

So is social working? It depends on the data question you chose. Measured by sessions, the push is a clear success. Measured by orders, things improved a little. Measured by what each new customer cost, which lesson 4 calculates, the picture changes again. None of these answers is false. Lena has to choose the measure that matches the decision, which is about money, and say in the memo which one she chose.

Four ways a data question goes wrong

• A vague period: recently or since the push without a date lets two analysts count different months.
• A silent change of population: counting all sessions when the question is about social sessions.
• A measure that cannot answer: sessions for a question about money.
• An unfair comparison: September to March contains December, the strongest month of the year, while January to August contains the summer holidays. Monthly averages help, but the seasons still differ, and the memo should say so.