Why this matters
Banks, savings and loans companies, mobile money providers, agritech firms, NGOs and research projects in Ghana all collect more data than they can read. An analyst tells them what happened last month. A data scientist goes further: how sure are we that this change is real, what caused it, what is likely to happen next, and what should we do about it? Those questions decide where money, seed, medicine and staff time go, so the people who can answer them carefully are valuable.
Four kinds of question
It helps to sort the work into four kinds of question.
- Describe: how many farmers in the cooperative repaid their input loans on time? This is analyst work, and you already know it.
- Infer: the repayment rate in our sample of 300 farmers is 82 percent; what range is the true rate for all 4,000 members likely to sit in?
- Explain or test: did the new SMS reminder actually raise repayment, or did repayment rise because the harvest was good?
- Predict and decide: which members are likely to struggle next season, and is it worth a field officer visiting them first?
This track spends most of its time on the middle two, because that is where careless work does the most damage. A model that predicts well but answers the wrong question, or a confident claim built on a tiny sample, can send resources to the wrong place.