Data analysis and insights
Someone has a question. Why did repeat purchases drop in March? Which customers are worth keeping? What is behind the rise in returns? We dig into the data, tell you what we found and how sure we are, and say what we would do next.
Signs you could use this
- You have a specific question and nobody on the team has the time to chase it properly.
- An earlier analysis gave a result nobody quite trusted.
- You'd like to predict something (churn, demand, conversion) but aren't sure the data supports it.
- Your analysts spend most of their week on requests and have no room for deeper work.
What we do
- Pin down the question
- A vague question gets a vague answer. We agree what decision the analysis will inform before touching any data.
- Clean and join the data
- This is most of the work in any real project. We document what we changed and why, so the analysis can be repeated.
- Choose the simplest method that works
- Often a careful comparison does the job. When it doesn't, we use regression, classification, clustering, time series or text analytics, and we explain the choice in plain words.
- Explain it so people act on it
- You get findings written for the person who has to make the decision, with the caveats stated clearly rather than buried.
What you receive
A focused analysis takes one to three weeks. Predictive models take longer because they need testing against fresh data before anyone relies on them.
- A written report of findings with the key charts
- The queries or notebooks behind it, so your team can re-run it
- A call to walk through results and take questions
- Recommended next steps, ranked
Tools we use: SQL, Python, R, Excel, BigQuery, AWS Redshift, PostgreSQL.
Common questions
What methods do you use?
Whatever fits the question: linear and logistic regression, decision trees and random forests, k-means and k-prototypes clustering, time series analysis, market basket analysis, text analytics and support vector machines. Often a good comparison table beats all of them.
Is our data good enough?
We tell you early. A short look at the data before we commit usually shows whether the question can be answered, and if not, what would need to change.
Can you work with our analysts?
Yes, and we prefer to. We hand over code and documentation so your team keeps what we build.