Does NPS Really Predict Revenue in Healthcare Clinics? I Spent 12 Weeks Finding Out.

1. Why I Questioned NPS in Healthcare (The Origin Story)

For 6+ years grinding in Brazilian aesthetic healthcare clinics, I saw NPS treated like gospel:

Everyone worshipped NPS.

But no one could prove it actually linked to revenue.

Every month:

Dashboards. Meetings. Bonuses. Marketing decisions.

All based on one metric: the Net Promoter Score.

No clinician, manager, or exec could point to data linking patient "loyalty" to the bottom line. That gap haunted me – and sparked my Master's deep dive at Torrens University Australia, under Prof. Dr. Bushra Naeem (ICT R&D expert in digital transformation).

Research Methodologies (REM502)'s structure turned this frustration into rigorous research:

• 
  Assessment 1 (Critical Literature Review): We had to dissect 12+ studies and pinpoint gaps. I zeroed in on NPS's unproven link to financials – literature screamed "patient experience matters" (e.g., Godovykh & Pizam, 2023), but nothing validated revenue correlation in healthcare. This nailed the "knowledge gap" (emotion detection, sentiment-to-outcomes, NPS weaknesses) and set my direction: Test it empirically.


• 
  Assessment 2 (Research Tools & Methodologies): Building on A1's gap, this formalized my hunch into science. I crafted RQs (e.g., "Does NPS correlate with revenue growth?"), hypotheses (H1: Positive link), and a quantitative toolkit (Python pipeline for correlation/regression). It bridged intuition to method – like designing an ML experiment before coding.


• 
  Assessment 3 (Full Research Proposal): The capstone synthesized everything into a publication-ready plan. A1's gap + A2's methods became a pragmatic-positivist design, ethics framework, and ICT software flow. It's not just academic – it's a blueprint for production BI systems.

What NO ONE HAS DONE

❌ validate whether NPS statistically predicts revenue in healthcare

❌ test whether higher NPS tracks with monthly business performance

❌ evaluate NPS using real clinic financial data

❌ build evidence-based decision frameworks for healthcare executives

So that’s the gap. And that became my research question.

Code Example: Correlation Analysis

import pandas as pd

from scipy.stats import pearsonr, spearmanr



df = pd.read_csv("clinic_nps_revenue_clean.csv")



# Calculate revenue growth

df['revenue_growth'] = df.groupby('clinic')['revenue'].pct_change() * 100



# Pearson Correlation

pearson_r, pearson_p = pearsonr(

    df['nps_score'].dropna(),

    df['revenue_growth'].dropna()

)



print(f"Pearson r = {pearson_r:.3f}, p = {pearson_p:.4f}")

Planned Models

• Pearson + Spearman correlation


• Linear regression (with lagged variables)


• K-means clustering (clinic behavior patterns)


• Visual dashboards (Streamlit)

5. So… Does NPS Predict Revenue? (3 Possible Scenarios)

This is the part everyone asks.

I built a decision-impact framework:

6. Code + Architecture for Engineers

This project is structured like a mature ML pipeline:

research_project/

│

├── data/

│   ├── raw/

│   └── processed/

│

├── notebooks/

│   ├── 01_preprocessing.ipynb

│   ├── 02_eda.ipynb

│   ├── 03_correlation.ipynb

│   └── 04_regression.ipynb

│

├── src/

│   ├── preprocessing.py

│   ├── analysis.py

│   ├── models.py

│   └── visualization.py

│

└── streamlit_app/

    └── dashboard.py

It mirrors real-world data engineering pipelines used in BI, ML, and healthcare analytics.

7. What This Means for Healthcare & AI

Short version:

You can’t build AI systems on shaky assumptions.

If NPS ≠ revenue, then:

• AI sentiment systems need rethinking


• Dashboards should emphasize different KPIs


• Marketing & retention strategies must pivot


• Budget allocation may change dramatically

If NPS = revenue, then:

• NPS becomes a validated financial lever


• Clinics can forecast demand using survey data


• AI models can be trained on validated correlations

In both cases, ICT R&D gains a validated foundation.

8. Explore the Research

9. What’s Next (Phase 2 & 3)

Phase 2 – Extended Analysis

• More complex regression (ARIMA, lag models)


• Business segmentation via clustering


• Qualitative coding of text comments


• Publish as interactive Streamlit app

Phase 3 – Publication Pathway

• Submit to International Journal of Market Research
  


• Present at ACIS Conference


• Develop predictive model for revenue forecasting


• Integrate into ClinicTrends AI v3
  

10. Key Takeaways

• NPS is widely used but poorly validated.


• Healthcare lacks evidence-based KPI frameworks.


• 27K+ real-world data points allow statistical validation.


• ICT R&D + ML pipelines can settle long-standing assumptions.


• This study builds the foundation for smarter healthcare analytics.

11. Let’s Connect

If you're working with healthcare data, KPIs, BI dashboards, or ML pipelines — I’d love to connect.

• 
  LinkedIn:
  


• 
  Portfolio: luisfaria.dev
  

Final Note

This isn’t just a research assignment — it’s a problem the healthcare industry has ignored for years.

If the link between NPS and revenue exists, we’ll finally quantify it.

If it doesn’t, we’ll stop wasting money on metrics that don’t matter.

Either way, data wins.