Intuit Data Analyst Interview Questions

“First I would reduce the events to one row per user with the furthest step reached, using MAX on the step order. Then I count users who reached each step and work out the share who went on to the next. The biggest fall shows where to look. Two cautions. Not every user sees every step, because the flow branches, so I compare only users for whom the step applied. And someone who stopped yesterday may come back tomorrow, so I would give people a fair time window before counting them as lost. I would then split by device, since phone users may stop at different points.”

“The first measure is how often the owner accepts the suggested category without changing it. But acceptance can also mean the owner did not look, so I would track later corrections too, especially those made by the accountant at the end of the period. That shows the silent errors. I would measure the time owners spend on categorising before and after the feature. I would split results by type of business, because a category that is right for a cafe may be wrong for a builder. I would also take a sample and have a trained person check it by hand.”

“Because traffic changes so much through the season, I would never compare one week against another. Users are split at random and both versions run over the same days, so each sees the same mix of early and late filers. The main measure is completion of the whole flow, not just that one step, since clearer wording could move the problem further down. Before starting, I would work out how many users we need to detect a useful difference. I would add guard measures, such as more corrections later or more support contacts. And I would agree the end date in advance.”

“I would first agree the definition with the team, because a sign-in is not the same as doing business. My suggestion is that a business is active in a month if it created an invoice or recorded a payment. In SQL I would select business and month from each of those tables, combine them with UNION, and count distinct businesses by month. I would report sign-in only businesses as a separate line, since those may be about to leave. Then I would follow each sign-up cohort month by month to see how long businesses stay active. Seasonal businesses need care, as a quiet month is normal for them.”

“At your company the team believed a new reminder email was bringing users back. My analysis showed that users who got the email returned at about the same rate as a similar group who did not. Before saying anything, I checked my groups and asked a colleague to review the query. Then I showed the product lead in private first. I presented the method, the result and what I could not rule out. I suggested a proper test with a holdout group to settle it. She agreed. The test gave your result.”

“At your company I was asked how income level related to use of a feature. I did not need names or exact amounts, so I asked for a data set with identities removed and income grouped into bands. I did the work inside the approved environment and did not download files to my laptop. In the report I left out any group with very few people, since a small group can point to one person. I also asked our privacy contact to look at the plan before I started. The result was your result.”

“I would first check the data itself. Is the pipeline late, or did an event name change in a release? Many falls are faults in tracking. If the data is sound, I would compare with the same weekday in earlier weeks, not just yesterday, because filing has a strong weekly pattern. Then I would split the number by device, region and step. A fall that sits in one browser or one step points to a product fault. A fall everywhere may have an outside cause, such as a public holiday. I would tell the team early what I know so far, and what I am still checking.”

“I enjoy analysis most when it shows where real people get stuck. Intuit's products guide people through tax filing and help small businesses keep their books, and the data shows exactly where that gets hard for them. Finding the confusing step and seeing it fixed is work I find worthwhile. I also take seriously that this is private financial data, and I like working where that is respected. At your company my favourite project was your project, a funnel study that changed a form. I want to do more of that and learn better experiment methods.”