Ola retweeted
Dear Data Analyst,
If 100 customers are in your table and 20 of them have a blank where their spend should be, then total spend is $4,000 and SQL's AVG will tell you the average customer spends $50, because it quietly ignores the blanks and divides by 80, but if those 20 people actually spent nothing, the real average is $40.
That's why you should always check how many blanks are sitting in a column before you average it, because SQL won't tell you it skipped them, and whether a blank means "zero" or "unknown" can change your answer completely.
#DataScience #DataAnalytics
Ola retweeted
Business Intelligence is five stages, and Power BI is involved in all of them.
Collect, clean, model, measure, show and share.
Most people learning Power BI are learning the fifth one, because it is the one with the pretty pictures. The reports that break are broken at the third.
Save this one if you are starting out, it is the map.
#PowerBI #PowerBITraining #dataanalytics
Ola retweeted
Dear Hiring Manager,
If you ask an analyst why sales dropped 30% last week and they send you a chart straight away, you might start planning a meeting about it, but if they first check and come back saying "it's actually down 4%, the 30% was because one store's data didn't load on Tuesday", they've just saved you a very awkward conversation with the leadership team.
That's how I try to work, I'd rather take an extra ten minutes checking the data is right than hand someone a confident number that sends them in the wrong direction.
#DataAnalytics #DataAnalyst #Hiring
Ola retweeted
3 things decide whether a recruiter ever sees you, and only one of them is about keywords
1️⃣ Parsing - Your CV becomes a record with fields. Put your name in the page header and it can vanish, because employers' own guidance warns that header and footer text sometimes gets dropped entirely. Use two columns and some systems read straight across, so your job title comes out as "Analyst | 2019 - Present | Birmingham". Put your skills in a table and they may not arrive at all.
2️⃣ The employer's own filters - Not a keyword score, an actual configured rule, a degree requirement or an employment gap flag. Harvard Business School and Accenture found 94% of employers using a recruiting system let it make the first cut on mid skill roles. You cannot write your way past that one.
3️⃣ Search - A recruiter types something like "Power BI" AND SQL AND ("data analyst" OR "BI analyst") and reads what comes back. Nobody scored you. If your record is empty you simply are not in the results.
So the fixes, in order of how much they matter:
• One column, no text boxes, no tables.
• Contact details in the body, never in the header or footer.
• Standard headings, Experience and Education, not My Journey.
• Mirror the advert's exact nouns, since Power BI and PowerBI are different strings.
• Put each keyword in a bullet that evidences it, never in a stuffed block.
• Answer screening questions straight, and find a referral instead.
Readability first, keywords second, myths never.
#ATS #Hiring #PowerBITraining #dataanalytics
Ola retweeted
Dear Hiring Manager,
If you ask an analyst for last month's sales and they come back with a number in five minutes, that's great, but if they come back in fifteen and say "it's £84k, but £6k of that is one bulk order, so the normal month is closer to £78k", that's the person you actually want, because they've answered the question you meant to ask, not just the one you typed.
That's how I try to work, the number is only half the job, and the context around it is what helps people make the right decision.
#DataAnalytics #DataAnalyst #Hiring
Ola retweeted
One measure, written once, returning a different number in every visual on the page.
Total Sales = SUM(Sales[Amount])
With North selected the card says 400, the matrix rows say 40, 300 and 60, and the CALCULATE with ALL card says 1,200. The measure never changed. What changed is the set of filters each visual handed it, and that set is called the filter context.
The one that catches people is the Avg Order Value matrix. Rows of 5.00, 30.00 and 10.00, total of 16.67, which is neither 45 nor 15. North made 400 across 24 orders and 400 divided by 24 is 16.67, so the total row is simply the same measure evaluated with no category filter applied.
Eight things worth holding onto:
• A measure is a recipe, not a number.
• The visual's grouping is a filter.
• A slicer is a filter.
• CALCULATE is what changes the filters it was handed.
• ALL removes them.
• The total row is its own calculation.
• SUM is not SUMX.
• And a calculated column is worked out once at refresh while a measure runs every time a visual renders.
Save this one, it answers most of the "why is my number wrong" questions in one page.
#PowerBI #PowerBITraining #powerbitutorial
Ola retweeted
Dear Data Analyst,
If one system saves a date as 04/10/2026 and another one is set up the American way, the same order can show up as the 4th of October in one report and the 10th of April in the other, and suddenly your monthly sales look like they've dropped off a cliff when nothing has changed at all.
That's why checking how dates are stored is one of the first things you should do with any new data, because a wrong date doesn't throw an error, it just quietly puts your numbers in the wrong month.
#DataScience #DataAnalytics
Ola retweeted
7 of these 36 formulas do not return a value.
They return a range and that is the single biggest shift between old Excel and the Excel you should be writing now.
FILTER, UNIQUE, SORT, SORTBY, SEQUENCE, TRANSPOSE and TEXTSPLIT spill into the cells below and resize themselves when the data grows, which is exactly what you want in a report that gets rebuilt every month.
Dragging a formula down does not do that.
Save this one for later.
#excel #msexcel #exceltips #dataanalysis #datascience #dataanalytics
Ola retweeted
42 SQL commands, grouped the way you actually use them, and the six that change your row count without telling you
Reading data, filtering and logic, aggregating, joining and stacking, window functions, changing data.
The six marked ones change your row count without telling you.
LEFT JOIN fans out when the right-hand table has duplicates.
CROSS JOIN multiplies every row by every row.
UNION quietly drops duplicates, UNION ALL quietly keeps them.
GROUP BY collapses your grain to one row per group.
DISTINCT hides a fan-out rather than fixing it, which is why reaching for it is usually a sign something upstream is wrong.
Notice that the window functions card has no marks at all. That is the whole difference between GROUP BY and a window, one gives you a row back per group and the other gives you every row plus the answer beside it.
AND, OR and NOT are not on here on purpose, because nobody has ever needed a cheat sheet to find AND, and the space went to CASE and COALESCE which people genuinely do look up.
Save this one, you will reach for it.
#sql #dataanalyst #dataengineering #datascience
