Classifications

A classification enriches a dimension with attributes from elsewhere, without changing anything in collection: the product line and margin of each product, the agency and budget of each campaign, the section and author of each page. It becomes an analytics.cls_<name> view, which eodia insights joins to the others.

Dimension	What is classified	Join
path	pages	key = page_views.path (or sessions.entry_path)
utm_campaign	campaigns	key = sessions.campaign (or attribution.campaign)
utm_source, utm_medium, utm_content, utm_term	sources, mediums, contents, terms	key = sessions.source, medium, content, term
item_id	products	key = commerce_items.item_id
event	event names	key = events.name
property	the values of an event property	key = event_properties.value_text

Each join is also made on site_id: a classification belongs to one site.

Creating a classification

In the site’s Metadata tab, Classifications section:

• a technical name, in lowercase (catalogue): the view will be called cls_catalogue;
• a label and a description, which reach insights as comments;
• the dimension described;
• typed attributes (text, number or date), 40 at most: gamme, marge, fournisseur.

Importing a CSV

A CSV of 10 MB at most, separated by commas, semicolons or tabs (the separator is detected):

item_id;gamme;marge;fournisseur
KA2;Audio;0.38;Kerys
PL-ROUGE;Maison;0.52;Tissages d’Arvor
PL-BLEU;Maison;0.52;Tissages d’Arvor

The screen shows a preview of the first rows, asks for the key column and suggests a type for each column, which you correct. Then:

• Replace deletes the existing rows and loads the file;
• Merge adds the new keys and updates the existing ones.

Through the API

To keep a classification up to date from a catalog, with an eoa_… integration token:

curl -X PUT https://stats.example.com/api/v1/sites/$SITE_ID/classifications/$CLS_ID/rows \
  -H "Authorization: Bearer $EODIA_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "merge",
    "rows": [
      { "key": "KA2", "values": { "gamme": "Audio", "marge": 0.38, "fournisseur": "Kerys" } }
    ]
  }'

POST …/classifications/:cls/import also accepts the CSV itself. The full reference is in the API documentation, in the application (Documentation).

The cls_<name> view

Column	Contents
site_id	the site
key	the classified value
one attribute per column	typed: text, number or date; its label as a comment

The view’s comment states the dimension and the join to make: the insights assistant reads it.

-- Gross margin by product line, from the ordered items
SELECT c.gamme, sum(i.revenue * c.marge) AS gross_margin
FROM analytics.commerce_items i
JOIN analytics.cls_catalogue c ON c.site_id = i.site_id AND c.key = i.item_id
WHERE i.step = 'purchase'
GROUP BY c.gamme

The view is regenerated whenever the classification changes; an import only changes its rows.