Get Started with SKARI: Your First Clustering Project in 5 Minutes

You don't need to write code — or even know which algorithm to pick — to run a real clustering analysis. This walkthrough takes you from a raw spreadsheet to interpreted customer segments in five steps.

Note

You'll need one thing: a table where each row is a record (a customer, a product, a survey response) and each column is a numeric or categorical attribute.

Step 1 · Upload Your Data

Drag in a CSV or Excel file. SKARI reads the columns, infers each type, and shows a quick preview so you can confirm it parsed correctly.

Supported formats: CSV, Excel (.xlsx), and common database exports
Each row = one record; each column = one attribute
No need to clean headers by hand — types are detected automatically

Step 2 · Automatic Preprocessing

This is the step most tutorials skip and most projects get wrong. SKARI handles it for you: scaling, missing values, and encoding — fit on your data, applied consistently.

Numeric features are standardized so no single column dominates
Missing values are imputed with a sensible strategy
Categorical columns are encoded automatically

Tip

Scaling matters more than the algorithm choice for distance-based clustering. If you want the why, read the Data Normalization guide linked below.

Step 3 · Choose a Clustering Method

Pick an algorithm, or let SKARI recommend one based on your data's shape and size. For a first project, the default K-Means with an auto-suggested k is a solid start.

Not sure how many clusters? SKARI suggests k using the elbow and silhouette
Expecting noise or odd shapes? Switch to HDBSCAN in one click
Want the robust answer? Turn on the ensemble to combine several methods

Step 4 · Run

Click run. Results appear in seconds — cluster sizes, centroids, and a scatter you can explore. No waiting on a training script, no environment to set up.

Step 5 · Interpret and Export

The last step is the point of the whole exercise: understanding what the clusters mean. SKARI profiles each group so you can name it and act on it.

Cluster Analysis

K-Means · k=3

K-Means partitions points into k groups, each summarised by its centroid.

Cluster 1Cluster 2Cluster 3

Read each cluster's profile — which features define it
Rename groups to something meaningful (VIP, at-risk, new)
Export a report, or publish the model as an API for reuse

Takeaway

That's a complete clustering project — upload to insight — without a line of code.

Frequently Asked Questions

Do I need to normalize my data first?

No — SKARI does it automatically in Step 2, fitting the scaler correctly so there's no data leakage.

What if I don't know how many clusters to use?

Let SKARI suggest k. It evaluates the elbow and silhouette across a range and recommends a starting point you can adjust.

Can I reuse the result later?

Yes. Export the trained model as an API and score new records with the same clustering.

Conclusion

The hard parts of clustering — scaling, choosing k, comparing methods — are exactly the parts SKARI automates. Your job is the interpretation, which is where the value is anyway.

Takeaway

Upload a file, and you're five minutes from your first real segmentation.

Complete Clustering Guide

Understand the algorithms behind the button

Data Normalization Guide

Why preprocessing decides your results

Advanced Clustering Techniques

Ensembles and multi-layer analysis for tougher data

5-Minute Workflow