Тестовое задание
на аналитика данных
Wise

Task instructions

Attached is a dataset representing daily transfer volumes from GBP to ZAR for the latter three quarters of 2023.

Please analyse it and answer the following questions.

Answers can be presented and working done in any tool you wish, but please include your working for reference.

Please don't put a lot of effort into the presentation. We're not assessing presentation ability with this task, just your answers and reasoning. You can present in any format you like. Notebooks are fine, as are slide based files.

1 a) Describe the distribution that our daily transfer volumes follow.
1 b) What real world cause do you think is behind this shape of distribution?
1 c) What are some of the implications that this distribution would commonly have on analysis that you might do?

2 Have transfer volumes along this route changed significantly from quarter to quarter in 2023?. How would you determine that the observed differences from quarter to quarter are 'real' as opposed to being the result of background fluctuations?

3 Estimate the total transfer volume for October 2023 (bonus, include a measures of range & certainty).

Checking assumptions for applying the t-test

You need to answer the following question and provide a method for testing the assumption:
Is it necessary for the original data to be normally distributed in order to apply the t-test?

Different sources provide different interpretations. Some emphasize that the sampling
distribution of the test statistic should be approximately normal, while others state that the
original data itself must follow a normal distribution.

The result is expected in the form of a clear answer to the question with a brief explanation, as
well as working code that implements a check for whether the assumption holds for a given
dataset.

Analysis of the user behaviour

You need to analyze the logs attached to the task and determine how user behavior differs
between the two samples (each log file represents one sample).
You should also interpret the identified difference.

The result is expected in the form of a working code (script or notebook) performing the
analysis, along with the conclusion based on the results.

Log format (users_log_raw_a(b)10000.txt):

The logs include Session metadata, Query, and Click actions.

● TypeOfRecord — type of the log entry: query (Q), click (C), session metadata (M).
Session metadata (TypeOfRecord = M):
Fields: SessionID \t TypeOfRecord \t Day \t USERID
● SessionID: unique session identifier
● Day: day number when the session occurred
● USERID: unique user identifier

Query actions (TypeOfRecord = Q):
Fields: SessionID \t TimePassed \t TypeOfRecord \t SERPID \t QueryID \t ListOfTerms \t
ListOfURLsAndDomains

● SessionID: unique session identifier
● TimePassed: time units passed since the beginning of the session (the unit duration in
milliseconds is unspecified)
● SERPID: unique search results page identifier
● QueryID: unique query identifier
● ListOfTerms: list of search terms in the query
● ListOfURLsAndDomains: ordered list of URLs with domains shown on the search results
page
○ Format: (URLID, DomainID) — unique URL identifier, unique domain identifier

Clicks (TypeOfRecord = C):
Fields: SessionID \t TimePassed \t TypeOfRecord \t SERPID \t URLID

● SessionID: unique session identifier
● TimePassed: time units passed since the beginning of the session (the unit duration in
milliseconds is unspecified)
● SERPID: unique search results page identifier
● URLID: unique URL identifier that was clicked

Clicks are attributed to the QueryID that precedes the clicks in the log.

Тестовое задание на аналитика данных в Wise. Ознакомьтесь с примерами реальных тестовых заданий, которые предлагаются кандидатам. Узнайте, какие задачи могут встретиться и как они связаны с будущей работой. Это поможет лучше подготовиться к собеседованию в Wise и понять ожидания работодателя.