STEM University — Prove skills by building

Your Work

A classification is a governed mapping F: K_B → Y from raw dimension keys to reportable values. Build the mappings, define the rule set, and verify the math against the event stream.

Date range

Preview slice: 20 of 20 events. Browsers seen: Chrome, Safari. Gate uses full stream.

Bulk import (.TAB) — browser → vendor

Classification Importer exports a .TAB; analyst fills in the classification column; import back.

ChromeSafariFirefoxEdge

Consolidate — category → group

Conservation invariant: Σ_y M[group=y] must equal Σ_k M[B=k] across covered keys.

ElectronicsApparelHome

Rename — SKU → friendly name

Pattern: SKU-Pnnn → Product Pnnn. Fill all 10.

SKU-P000SKU-P001SKU-P002SKU-P003SKU-P004SKU-P005SKU-P0061× in previewSKU-P007SKU-P008SKU-P009

Rule Builder — evar_1 → ChannelType

Add rules in priority order. Last rule should be a catch-all .* regex → UNCLASSIFIED. Test samples include paid_search, email, direct, organic.

Match typeMatch termOutputPriority

Metric by classified value

Compute M_orders[BrowserVendor = y, DR] for the chosen vendor over the full stream. The truth depends on the canonical mapping (Chrome=Google, Safari=Apple, Firefox=Mozilla, Edge=Microsoft).

Target vendor

Orders for that vendor

Workspace knowledge

For each classification category, pick the eligible base-variable type:

TrafficConversionMarketingChannel

Which tool fits which use case?

Static known table (fixed list of values)

Pattern-driven ongoing values

What do you do with the first 5 rows of an exported .TAB file?

Catch-all rule

Position in the rule set

Catch-all output value

Evaluation

Practice checks are unlimited and don’t count toward your grade. Final submit is one-time and generates your certificate.

Practice the workspace, ask Dr. Okoye anything, and submit when you're ready. The grade lives in the workspace, not the chat.