Portfolio case study · Education analytics

From student data
to informed action.

Connecting marketing, CRM, admissions and learner support through clear reporting, data modelling and early risk analysis.

Download Power BI + ebook + Excel Read the ebook PDF

One ZIP · approximately 175 MiB · Power BI Desktop for Windows required. Public downloads — no GitHub sign-in required.

11.5Msynthetic rows modelled

87/87dbt models and tests passing on Microsoft Fabric

26/26metrics identical on DuckDB and Fabric

1,091records loaded to a live Salesforce org and extracted back

Pipeline: Salesforce CRM, NiFi and REST extract, OneLake bronze, dbt warehouse, Power BI and Excel — End to end: Salesforce CRM through NiFi and OneLake into a tested dbt warehouse, reported in Power BI and Excel.

How the platform is built

Medallion layers: 13 bronze, 13 silver, 8 gold and 2 quality models — The medallion warehouse: 36 dbt models across bronze, silver, gold and quality, with 51 data tests.

Canonical data model covering catalogue, CRM acquisition and the learner lifecycle — The canonical data model: catalogue, CRM acquisition and the learner lifecycle.

What the data shows

Funnel from 1.5 million leads to 304,168 enrolments — From 1.5 million leads to 304,168 enrolments across the synthetic population.

Response rate and return on ad spend by marketing channel.

Enrolments and average discount by programme.

Attendance and assessment decline while the at-risk share rises over the study weeks — Attendance and assessment decline across the study weeks while the at-risk share rises.

Early risk, and what to do about it

Power BI page showing predictive models and recommended actions — The Power BI report: two predictive models and the actions they point to.

Feature importance for the enrolment propensity and withdrawal risk models — What drives each model: lead source for enrolment propensity, early attendance for withdrawal risk.

Understand the funnel

Explore campaign performance, CRM activity, applications and enrolments. Compare programme demand and delivery modes.

Spot early risk

Explore illustrative propensity and withdrawal models, course rankings and learner support priorities.

Guide decisions

Review reconciliation, data quality, SEO findings and business and systems analysis recommendations.

What is included?

The Power BI project includes report and semantic-model definitions, 23 Parquet data files, branding assets and a path-configuration script. The same download includes the ebook PDF and Excel workbook.

Tools: SQL, Python, Power BI, Salesforce, Apache NiFi, dbt, DuckDB and Microsoft Fabric. Documentation also explores a Google Cloud / BigQuery architecture option.

Excel workbook Release details

Open the report on your PC

Download the ZIP and choose Extract All. Keep the extracted folders together. Do not open the project inside the ZIP.
Close any previously opened copy of the report. In the extracted folder, review and run Setup-DataPaths.ps1 with PowerShell. It points the model to that folder's included data files.
Open RedAndYellow.pbip in a current Power BI Desktop version with project support, then select Refresh.
If you move the extracted folder, run the setup script again before reopening the report.

The source repository excludes the large data files; use the release ZIP for the complete portable package. The original project filenames are retained so report references continue to work.

How to interpret the evidence

This is a portfolio case study. Analytical people, spend and student outcomes are synthetic or illustrative, not an institution's actual results. The ebook distinguishes the live Salesforce demo extract from staged custom-object data and the larger analytical population. Predictive scores support human review; they are not validated production outcomes. Fabric and Google Cloud implementation status and remaining limitations are documented in the ebook.