(CASE STUDY · 10)

POWERBI ANALYSIS

Data Analysis
Report Design
Power BI
Data Storytelling

PowerBI Analysis is a Power BI report on 5,000 viral social-media posts across 4 platforms, 8 regions, 6 content formats and 10 hashtags. The first version charted a count of posts as if it were shares, likes and views. The rebuild summed and averaged instead, and showed what this dataset says: each platform's total follows how much it posts, and reach per post is flat.

ROLE: Data Analysis & Report Design
TIMELINE: Mar 2025 - Jun 2026
YEAR: 2025-2026
TEAM: Solo

(MY ROLE)

Built the first report on a 5,000-row dataset: 6 charts on one page and a written analysis.
Audited it chart by chart and traced the error to the aggregation.
Redesigned and rebuilt the report, with a finding strip on every page.
Designed and shipped the landing site with a .pbix download.

(WHY IT EXISTS): It was built to help brands see where reach and engagement come from. The question was which platform, format, market or hashtag earns the views.
(1,324 SHARES. IT WAS A POST COUNT.): The March 2025 report set every chart to Count. YouTube's 1,324 shares were its 1,324 posts, and the same 4 numbers filled 4 charts. The write-up also called 13.54% Germany's share of views. It was the USA's share of posts.
(WHAT CHANGED): The June 2026 rebuild summed and averaged instead. On summed views the USA leads with 14.09%, and Germany is 7th of 8. A written analyst's note now sets out the method and the limits of the data.

Process

01
Audit & Trace

Read the aggregation behind each visual in the first report's layout file, then checked every chart against row counts from the CSV.
02
Re-aggregate

Sum for totals, Average for per-post numbers, Count only where a count is meant. Views came to 12.47B, and likes, shares and comments to 13.1% of that.
03
Normalise by Volume

Divided platform and format totals by their post counts. Views per post sat between 2.40M and 2.55M on every platform. No two metrics correlated above 0.02, which the report reads as a sign of synthetic data.
04
Rebuild & Ship

1 page became 3: Executive Overview, Platform & Content, Findings. It ships as a .pbix file behind a static landing site on Vercel.

How It Works

One CSV becomes one table with no relationships. Each chart aggregates that table on its own, with Sum, Average or Count, and the finished file ships from a static site.

The Report

Every page has a navy header, a finding strip and white cards, and every chart is sorted by value. The stills are rebuilt from the report file's layout, with values from the CSV.

(SCOPE)

posts analysed: 5,000
charts across 3 pages: 13
KPI tiles on page 1: 6

Page 2, Platform & Content: average views per post by platform, from 2.55M on YouTube to 2.40M on Instagram, average likes per post, the post volume mix, views per post by platform and content format, and average engagement per post by format.

Page 3, Findings: region share of total views led by the USA at 14.09%, top hashtags by reach led by #Fitness at 1.39bn, average views by engagement level, and the Analyst's Note with its first point, Corrected aggregation, circled.

In The Detail

Power BI said it first. Filtered to Japan, the first report's tooltip read Count of Shares 153: YouTube's 153 posts there. In the rebuild, posts labelled High average fewer views than Medium or Low, so the report says not to rank content by that label.

The first report in Power BI, filtered to Japan by a click on its region donut: hovering 1.Shares by Platform shows the tooltip YouTube, Count of Shares 153, YouTube's post count in Japan, with that line circled.

The rebuild's Avg Views by Engagement Level chart: High 2.45M, Low 2.51M, Medium 2.53M, with High circled, above the Analyst's Note point on the engagement-label paradox.

WHAT EACH CHART AGGREGATES

Sum: Totals: views, likes, shares and comments. 5 charts.
Average: Per-post reach, likes and engagement. 6 charts.
Count: Only where a count is meant: posts per label and per platform. 2 charts.
No measures: Every chart aggregates the one table directly. None uses a DAX measure.

The Landing Site

One static page on Vercel, rebuilt with the report. It leads with the headline numbers, explains what the report shows, lists 6 insights and offers the .pbix as a download. It also walks through the report's 3 pages and names the file every figure comes from.