↓Skip to main content

The Project Brief Template

Fill in each field for your own project. Stuck? Click see examples to see how three projects (a churn analysis, an A/B test, and a marathon-training project) filled in that step. Your answers save automatically in this browser. When you’re done, hit copy brief to paste it wherever you want to work on it.

See What is the most important skill for a data scientist? for the full walkthrough of the approach behind this template.

01

What question are we trying to answer, or what problem are we solving?

see examples
Churn Our month-2 retention rate has fallen from 78% to 61% over the last two quarters, and we don't know why.
A/B testing We want to know if the new checkout flow increases conversion.
Marathon training My training log has three years of runs in it, but I have no way to predict my next marathon finishing time from it.
02

What do we think the answer or solution is?

see examples
Churn Our hypothesis is that users who don't complete onboarding within 7 days churn at three times the rate of those who do.
A/B testing Our hypothesis is that removing the second confirmation step will increase completed purchases by 5% or more.
Marathon training My hypothesis is that weekly mileage and average pace over the final 8 weeks of training predict marathon finish time better than total training time alone.
03

How will we know if the hypothesis is confirmed? What number(s) move?

see examples
Churn To test this hypothesis, we will measure the 90-day retention rate, segmented by whether users completed onboarding.
A/B testing To test this hypothesis, we will track the purchase completion rate as our primary metric, with the return rate as a guardrail.
Marathon training To test this hypothesis, I will measure the mean absolute error in minutes between predicted and actual finish time, compared against a naive baseline of simply repeating my most recent marathon time.

Get stakeholder buy-in on problem, hypothesis, and metrics before continuing.

04

How will we collect evidence? (analysis of existing data, experiment, simulation, backtest, other)

see examples
Churn To gather evidence for this hypothesis, we will run a cohort analysis comparing onboarding-completion date against 90-day retention.
A/B testing To gather evidence for this hypothesis, we will run a randomized A/B test for two weeks, comparing the old and new checkout flows.
Marathon training To gather evidence for this hypothesis, I will train a regression model on weekly mileage and pace from the final 8 weeks before each past race, and backtest it against my actual results.
05

What did the analysis show? How confident are we?

see examples
Churn This analysis confirmed the hypothesis: users who completed onboarding within 7 days retained at 68%, compared to 22% for those who didn't.
A/B testing This test confirmed the hypothesis: completions rose by 7%, but returns also rose, so the win was smaller than it first looked.
Marathon training This analysis confirmed the hypothesis: the model's predictions were off by 6 minutes on average, versus 19 minutes for the naive baseline.
06

What does this mean for the original question? What's the next problem?

see examples
Churn Based on these results, we triggered a re-engagement email at day 5 for anyone who hadn't finished onboarding, and the next question is whether that email actually moves completion rates.
A/B testing Based on these results, we shipped the simplified flow, and the next question is whether we can recover the lift without the extra returns.
Marathon training Based on these results, I now use the model to set a realistic goal pace before each race, and the next question is whether adding weather and course-elevation data narrows the error further.