Data analysis · Manchester, UK
I'm Ibomeno. I work in airline customer service in Manchester, and I do data analysis on the side. Everything on this page I built in my own time, because when I see some interesting data I can't help digging into it. It isn't all analysis, either: if something piques my interest, or I can build something that makes my day-to-day easier, I'll build it.
About
It's not as thrilling as Batman's, but Football Manager, if I'm honest.
Stay with me here. Imagine you're a lower table team with some bite and ambition for silverware, but you can't compete with the financial muscle at the top of the table. So what do you do? You look at the data for bargain players who overperform their attributes. The striker with poor finishing who still beats his xG. The slow winger completing a high number of take-ons and dribbles. The midfielder who doesn't look like much but has a high xA and leads in key passes and progressive passes. And that older centre-back who doesn't have the physique of his youth but is so mentally there that he has a 95% tackle success rate.
Did these sorts of players end up winning me silverware? Sometimes, but the other times I'd rather not talk about.
What kept me at it once I'd started was the logic of it, and the storytelling. Writing a query is a puzzle with a shape to it, and there's a very particular satisfaction in getting a join or a window function to do exactly what you meant instead of nearly what you meant. Then once I've found something, I like working out how to tell it: what comes first, what someone needs to see, and whether it wants a bar, a line or a scatter. I didn't expect to care about picking charts this much, but the wrong one costs whoever's reading it, and I've not got bored of that question yet.
Skills
Grouped by what I use them for. No percentage bars, because I've never worked out what 80% at Excel is supposed to mean.
I have been writing SQL for about seven years, since university, and it is where most of the projects on this page start. Joins across a lot of tables, CTEs, window functions, and getting normalised source data into a shape you can actually answer a question with.
The part nobody sees. Getting the source data into a state you can trust, then building a model that answers questions instead of one that just holds the rows. Most of the mistakes I have caught in my own work were caught here.
Dashboards built around the decision somebody has to make, not around every field that happens to exist in the table. The test I hold my own to is whether somebody could read one without me there to explain it.
Still where most business analysis really happens, whatever anyone says. I build workbooks assuming somebody else will inherit them, so the structure stays consistent and nothing breaks the first time a row gets added.
Still learning this one, mostly pandas. It is the reason DataBites exists, because the lessons I wanted did not fit the way I learn and building the thing turned out to teach me more than the lessons would have. I am not claiming more than that yet.
Every project on this page has a write-up saying what I did and what I found, and most of them also say what I would not claim from it. Saying which bits the data cannot answer is part of the answer, and it is the habit I would most want a team to check me on.
I use generative AI tools as part of the workflow: drafting SQL and DAX, explaining code I have not seen before, debugging queries that will not run, getting cleaning steps written faster, and drafting and proofreading write-ups and reports. Prompting is the easy half. The half that matters is validating what comes back, because AI-generated code can run without an error and still answer the wrong question, and a well-written summary can still get a number wrong. DataBites was built this way, with every DAX answer checked against the same numbers worked out separately in pandas.
Pick a query. It runs against a small fake dataset sitting in the page.
Projects
All of it is public and the code sits on GitHub, so you can go and poke at it. Open a card for the write-up or skip straight to the source.
Eighteen months of short-haul departures for a made-up UK carrier, built to see where punctuality actually goes. OTP15 came out at 78.26% against an 80% target, and the delay type that happens least often turned out to cost the most each time it does.
Two driver bonus schemes for a busy Saturday. Option 1 costs $1,050 and Option 2 costs $2,696, and only 2 drivers would get a bonus under Option 1 and miss out under Option 2. The more interesting find was a group of drivers neither scheme reaches.
Vehicle registrations for every US state, about 287 million vehicles. California has 35.3% of the country's EVs with about 13% of its cars, and Texas shows why count and rate have to be reported together: almost as many EVs as Florida, but the lowest rate of the four biggest fleets.
Three things everyone says about the NBA, checked against 12,000-odd player seasons. Players with a usage rate over 30% had the best true shooting, not the worst, and part of the drop in average height since 2015 was the league changing how it measured players in 2019.
The follow-up to the punctuality report, same carrier and same 18 months. I expected complaints to track delay. They don't. The correlation is 0.11 and three-quarters of complaints came from flights that left on time.
Deaths and vaccinations by country, joined on country and date. The main lesson was that no single death rate told you much, because the odds of dying if you caught it kept falling as treatment got better. Running vaccination totals showed how uneven the rollout was.
Eighteen months of in-app purchases from a made-up football management game. Revenue grew about twelvefold, mostly because more players started paying, and a tenth of the paying players bring in 64.5% of it.
8,360 matches from 2000/01 to 2021/22, rebuilt into a league table for any season you pick. My favourite find: 2020/21, played without fans, is the only season where away teams won more games than home teams.
Short lessons that run in the browser with nothing to install. It started as pandas on my phone and is now 370 lessons across Python, SQL and DAX, 139 practice problems and PL-300 exam prep. I built it with Claude as my coding partner, because most tutorials assume you can hold an hour of context in your head at once, and some days I can't.
No projects in that category yet.
Experience
Where I am now and where I studied. The full history is on the CV.
April 2023 — Present
British Airways Gold Guest List and Premier team · Manchester, UK
Previously in the Gold and Change Booking teams
Graduated 2021
Middlesex University · First Class Honours
Final-year project: researched YouTube addiction through user workshops, then designed and tested an alternative in Figma.
Contact
Hiring for any kind of role, or just wanting to complain about dashboards nobody reads. All welcome.
Manchester, United Kingdom
Open to all work options: on-site, hybrid or remote. Full right to work in the UK.
keybindings
Mouse works fine too.