Meet Freedom,

The question

Inflation peaked at 34.8% in December 2024 and had fallen to about 14.6% by September 2026. Families, traders and businesses want to know whether it will keep falling, level off or turn back up.

What drives it

• Inflation comes in waves, and each of the last three peaked higher: 2016, 2020 to 2021, and 2023 to 2024.
• Almost every wave followed a sharp fall in the naira. The link is strongest one month later and still strong after three.
• Crude oil showed almost no link while petrol was subsidised. That may change now the subsidy has ended.

Handling a change in measurement

In January 2025 the statistics bureau rebased the price index, and inflation dropped from 34.8% to 27.6% overnight, mostly because of the new basket. Every chart marks the break, and the models were tested with different ways of handling it.

Testing the forecasts fairly

• The last twelve months were hidden, and every method was scored against the simple rule that next month equals this month.
• A naira-based regression looked excellent on that hidden year, but across 56 rolling start points from 2010 to 2023 it usually did worse than the simple rule.
• ARIMA held up best over the long run and gives a range as well as a single number.
• No model foresaw the June 2023 naira float. Forecasts made around it were mostly 4 to 8 points too low.

The forecast

• Headline inflation: about 14.4% by September 2027, with a likely range of 6.5% to 22.4%. If the naira stays strong, about 11.8%.
• Food inflation: about 19.6%, with a likely range of 7.8% to 31.4%. If the naira stays strong, about 15.1%.
• Main risks: the naira, unsubsidised fuel prices, harvests, interest rate cuts and the 2027 elections.

The problem

Every Monday at Saint Freedom Health, two people read the same hospital in different ways. Kunle in Finance saw a ₦13M outstanding balance and a 56.2% collection rate, and wanted to cut costs. Folashade, the Medical Director, saw 27.3% of patients with hypertension and rising clinical risk, and worried that cuts would hurt care. Neither could see the other's numbers.

What I built

A two-page Excel dashboard on one set of data: 8,000 claims and 2,000 patients for fiscal year 2023, with slicers for year, gender and ethnicity on both pages.

• The patients: population size and age, clinical risk factors (smoking, diabetes, hypertension), blood pressure and BMI groups, average vitals, and patients by state and age group.
• The business: total charges and payments, claims paid, average pay per claim, collection rate and AR days, each with year-on-year and month-on-month change, plus claim status, revenue by month and revenue by payer.

What the data showed

• ₦29.7M was charged and ₦16.7M collected, a 56.2% collection rate, leaving ₦13.0M outstanding with an average of 32 days to collect.
• 6,801 claims (85%) were paid. 1,199 were denied or pending, about 8% and 7%.
• Payments rose 73% year on year, but the collection rate slipped by 0.2 points, so revenue grew without the hospital collecting a larger share of what it billed.
• Revenue is spread evenly across payers, from ₦2.89M (Employer Plan A) down to ₦2.66M (Blue Shield Health), so no single insurer is a concentration risk.
• On the clinical side, 41% of blood pressure readings were high and 39% elevated. 27.3% of patients are hypertensive, 18.7% diabetic and 18.0% smokers, and 58% are overweight or obese.

What changed

With both pages on the table, the Monday meeting started from the same facts. Finance could see that patients whose conditions are managed well produce cleaner, faster claims. The Medical Director could see that a steady collection rate is what pays for equipment and staff. You can't have clinical excellence without financial sustainability, and you can't have either without a clear view of the numbers.

The questions

• When did Christmas sales actually peak?
• Which campaigns drove holiday revenue?
• Did discounts hurt profit, or was volume the real driver?
• What happened to margins and returns around Christmas?

Changing course midway

The data looked like four months at first. Partway through, it turned out to be almost entirely November and December. Month-on-month trends across two months would have said very little, so I reframed the analysis: December became the focus, November became the baseline for every headline KPI, and each insight was framed around the effect of Christmas.

What I built

A two-page Excel dashboard with slicers for season (pre-holiday, holiday, post-holiday) and region.

• Sales overview: revenue, orders, average order value and gross margin against November with weekly bars, plus sales by channel, peak sales days and revenue by campaign.
• Profit and operations: gross profit and return rate, the value of gift features, returns by channel, profit by product category, and a scatter plot with a correlation test of discounts against margin.

What the data showed

• December revenue was $116.7K, 36% above November, from 1,449 orders (up 26%). Gross margin dropped to 23%, so sales grew faster than profit.
• Online brought in 59% of revenue. Cyber Monday was the biggest single day at $7.5K, ahead of Christmas Eve at $4.6K.
• 12 Days of Savings was the strongest campaign at $46.8K, followed by Gift Card Bonus at $28K. The Christmas Eve Flash Sale brought in only $3K.
• Orders with gift features averaged $92.84 against $87.03 without, about 7% more per order.
• Returns were 3.3% overall, higher online (3.8%) than in store (2.5%).
• Food and treats was the most profitable category ($4.8K), then home and fragrance ($4.6K). Decorations earned the least ($2.8K).
• Discounts showed only a weak negative link with margin (correlation −0.14). With so few data points that is a signal rather than proof, but it suggests volume, not discounting, explained most of the margin change.

The takeaway

Good questions and a willingness to change the plan mattered more than extra tools. Excel was enough to turn two months of sales into a clear Christmas story for decision makers.

The problem

It's 8:30 in the morning and Daniel, a loan portfolio manager, has a management meeting in a few hours. His director's question is simple: how is the loan portfolio doing? The answer is spread across separate files for disbursements, customers and loan performance, and pulling them into one more spreadsheet takes most of the morning.

What the screen needed to answer

• How much has been lent, and how much interest and repayment to expect.
• How many customers are in the portfolio.
• Which departments and products contribute most.
• How disbursements have moved over the year.

Building the foundation first

• Power Query: four Excel tables, 297 loans across 16 columns. I trimmed and standardised text, set data types by locale, capitalised names, replaced bad values, and derived age, years on book and age group.
• Data model: a star schema with the loans table at the centre, linked to officer, product and calendar tables.
• DAX: a dedicated measures table organised into folders, with a current value, prior year and year-on-year change for each KPI, plus colour measures that turn red or blue with the trend.

Designing beyond the usual Power BI look

Only after the model and measures were in place did I work on the design. I laid out the page in PowerPoint first, then built the visuals with HTML content visuals. Each KPI card is written in DAX that outputs HTML and CSS. It shows the value and year-on-year change, and flips when clicked to reveal the peak month and a 12-month trend line. I used Claude as a coding partner to turn the design ideas into HTML.

What the dashboard shows

• A loan portfolio of ₦2.16B, up 35.2% (₦561M) on the previous year, across 282 customers, 71 more than last year.
• Estimated interest of ₦1.68B, up 37.3%, and expected annual repayment of ₦766M.
• Home mortgages account for ₦1.06B of estimated interest, about 63% of the total and more than the next five products combined.
• Processing fees of ₦275M are spread fairly evenly: retail loans 30%, mortgage 25%, consumer finance and SME loans 23% each.

Open the live dashboard

The question

A two-month record of 4,207 outpatient visits at a fictional clinic, covering 35 doctors in 7 departments. Management wanted to know who carries the load, who the patients are, how long they stay, who needs a follow-up, when the clinic is busiest and where referrals come from.

How I worked

• Cleaned and joined the patient and doctor tables in pandas, with sanity checks on dates and IDs.
• Answered each question in its own step, showing group rates with 95% ranges.
• Tested the largest gaps and flagged differences that could be chance.

What the data showed

• Neurology and General Physician handle over a third of all visits.
• Visits per doctor range from 23 to 234 against an average of 120, and the gap appears in every department.
• The clinic averages 68 visits a day. Overnight beds peaked at 45 on 29 December against 26 on a typical night.
• One in four visits needs a follow-up, and 85% of patients walk in.

Recommendations

• Share workload more evenly within departments, starting with Neurology and General Physician.
• Plan beds for the holiday peak, not the average, and keep full staffing from 24 December to 2 January.
• Shape follow-ups and waiting areas around patients aged 45 and over, who make up 59% of visits.
• Build referral links once capacity is freed, aiming to lift referrals from 15% toward 20%.