For the question on "second highest order amount in the fashion department", should we be using DENSE_RANK() instead of RANK() - since if 2 orders have the highest amount then the third order (which is the second highest amount order) will get a rank of 3 using RANK() and a rank of 2 using DENSE_RANK()?
at 27:27 do we really need to group by c.first_name and c.last_name? I mean we already grouping by c.customer_id which has a kind of one-to-one relation with the first and last name
Thanks a million for this work, highly appreciated as i'll be applying for data positions at the end of the year. Quick observation; The cte in the third query was not used, cte = orders_per_year . And the rank was not also used either
I did not understand the code of the last question. He had to calculate the increase or decrease in month-over-month growth for the year 2022. What was the basis for creating the CTE for November and December (hardcoded)? In the second CTE, he filtered out only the December amount. Does it calculate and compare all the data back to January?
I feel like interviewer's focus on the parts that aren't really important i.e 18:20. Focus should be on testing the logic and the ability to get to the solution.
Almost every written query had very serious issues and either won't run at all or would give wrong answers. I'm not talking about typos, I'm talking about using an aggregate function when defining an order inside a partition clause, for example, which is not possible. Or using a window function AND a group by in a same query, which won't give the result he was hoping for. Or, as was done in his last query, filtering the data to only include one month and then using the lag function. Filtering with where is executed before the lag, so by the time lag is executed, the dataset has only records for december, meaning he will get nulls for every department. This is not a comprehensive list of issues, mind you, there are more. And that's just serious issues. In addition to that most of his filtering conditions were non-sargable, and he never thought about edge cases. For example, in the second problem it is possible that there were no users who bought something from one of the departments, and in that case the inner join he used would have lost that department, it wouldn't show in the result at all. Is it a plausible situation when working with real data? No. But it is definitely possible, and it should have been at least mentioned. Overall I am very disappointed with both the interviewee and the interviewer, who missed all of the mentioned mistakes.
With nov_dec_sum as ( Select department_name, sum(case when year(order_date) = 2022 and month(order_date) = 11 then order_amount else 0 end as nov_sum, sum(case when year(order_date) = 2022 and month(order_date) = 12 then order_amount else 0 end as dec_sum From orders o join department d on o.orders_id = d.orders_id Group by department ) Select department_name, From nov_dec_sum Order by nov_sum - dec_sum desc Limit 1
list of cust_id w most orders in last 5 years 5 years,5 last_name, 5 first_name, total amount of orders per customer SELECT c.cust_id, c.last_name, c.first_name, year(o.order_date) as years, COUNT(o.order_id) AS total_orders, SUM(o.order_amount) AS total_order_amount FROM customers c JOIN orders o ON c.cust_id = o.cust_id WHERE o.order_date >= DATE_SUB(CURDATE(), INTERVAL 5 YEAR) GROUP BY c.cust_id, c.last_name, c.first_name ORDER BY total_orders DESC LIMIT 5; how about this answer ?
SELECT customer_id, COUNT(order_id) AS total_orders FROM orders WHERE order_date >= CURRENT_DATE - INTERVAL '5' YEAR GROUP BY customer_id ORDER BY total_orders DESC LIMIT 1;
Don't leave your SQL interviews to chance. Sign up for Exponent's SQL interview course today: bit.ly/3TYla1W
Thanks for all you do. Always a joy to follow along. Wish there was more videos like this.
For the question on "second highest order amount in the fashion department", should we be using DENSE_RANK() instead of RANK() - since if 2 orders have the highest amount then the third order (which is the second highest amount order) will get a rank of 3 using RANK() and a rank of 2 using DENSE_RANK()?
Landed my data analyst job thanks
Congrats!! 🎉
at 27:27
do we really need to group by c.first_name and c.last_name?
I mean we already grouping by c.customer_id which has a kind of one-to-one relation with the first and last name
where o.order_date in (2019 etc) should not give expected result. Need to use year(o.order_date) in(2019,2020 etc
27:51 Mistake
CTE order_rankings have mention but in query orders_ranked
hello! i would like get to know such information like: this interview is for what level, for instance, junior or middle, maybe senior?
For the second to last question, can we do ORDER BY o.order_amount DESC, LIMIT 1, OFF SET 1, so that we don't have to use rank() ??
You can’t group by within rank function actually
@27:46 orders_ranked =1 should be order_rankings = 1
Thanks a million for this work, highly appreciated as i'll be applying for data positions at the end of the year. Quick observation; The cte in the third query was not used, cte = orders_per_year . And the rank was not also used either
I did not understand the code of the last question. He had to calculate the increase or decrease in month-over-month growth for the year 2022. What was the basis for creating the CTE for November and December (hardcoded)? In the second CTE, he filtered out only the December amount. Does it calculate and compare all the data back to January?
I feel like interviewer's focus on the parts that aren't really important i.e 18:20. Focus should be on testing the logic and the ability to get to the solution.
How would the Lag function pick up month 11 records when there is a where filter on month = 12?
Great job dude. Well Donee
w7.w3 wrong to group by customer and year,.. it should be only group by customer.
Is this mock interview for an entry-level data analyst position or a senior level?
Generally, you can expect SQL questions for both entry-level and senior data analyst positions
You also used orders_ranked=1 in your main query but your cte is orders_ranking. I don't think that'll work
thanks for the video, it's really useful
Almost every written query had very serious issues and either won't run at all or would give wrong answers. I'm not talking about typos, I'm talking about using an aggregate function when defining an order inside a partition clause, for example, which is not possible. Or using a window function AND a group by in a same query, which won't give the result he was hoping for. Or, as was done in his last query, filtering the data to only include one month and then using the lag function. Filtering with where is executed before the lag, so by the time lag is executed, the dataset has only records for december, meaning he will get nulls for every department. This is not a comprehensive list of issues, mind you, there are more.
And that's just serious issues. In addition to that most of his filtering conditions were non-sargable, and he never thought about edge cases. For example, in the second problem it is possible that there were no users who bought something from one of the departments, and in that case the inner join he used would have lost that department, it wouldn't show in the result at all. Is it a plausible situation when working with real data? No. But it is definitely possible, and it should have been at least mentioned.
Overall I am very disappointed with both the interviewee and the interviewer, who missed all of the mentioned mistakes.
order_date>= YEAR(CURRENT_DATE - INTERVAL 5 YEAR);
The queries for sure have a lot of errors. Cant group by in rank
thanks for all
He could have used the extract() to get the year.
Can AI answer those questions?
for the last question...
With nov_dec_sum as (
Select
department_name,
sum(case when year(order_date) = 2022 and month(order_date) = 11 then order_amount else 0 end as nov_sum,
sum(case when year(order_date) = 2022 and month(order_date) = 12 then order_amount else 0 end as dec_sum
From orders o join department d on o.orders_id = d.orders_id
Group by department
)
Select
department_name,
From nov_dec_sum
Order by nov_sum - dec_sum desc
Limit 1
👍
list of cust_id w most orders in last 5 years
5 years,5 last_name, 5 first_name, total amount of orders per customer
SELECT
c.cust_id,
c.last_name,
c.first_name,
year(o.order_date) as years,
COUNT(o.order_id) AS total_orders,
SUM(o.order_amount) AS total_order_amount
FROM
customers c
JOIN
orders o ON c.cust_id = o.cust_id
WHERE
o.order_date >= DATE_SUB(CURDATE(), INTERVAL 5 YEAR)
GROUP BY
c.cust_id, c.last_name, c.first_name
ORDER BY
total_orders DESC
LIMIT 5;
how about this answer ?
SELECT customer_id, COUNT(order_id) AS total_orders
FROM orders
WHERE order_date >= CURRENT_DATE - INTERVAL '5' YEAR
GROUP BY customer_id
ORDER BY total_orders DESC
LIMIT 1;