Algolia

How to structure and index your data set

• Searchable data
• Filterable data
• Display data
• Business metrics

The mapping is 1 to 1: 1 index per data type. So products, help, promos, all should go into a separate index each.

• For each DT, map out where this data is located in your system.
• What is the frequency of updates for each DT? Can we leverage any existing event system? In short, what should trigger a re-index? For instance:
  • when to update the stock level?
  • when to update the product list?
  • all other changes to the underlying data

Once index per store/language vs one index for all

If we go for one index per dimension (be it store or language):

• PRO: clear isolation per dimension
• PRO: flexible index settings
• PRO: optimal configuration and analytics
• CONS: slower indexing
• CONS: messier handling of settings, easier to make mistakes/not copy setting across indices

If we go for one index for all:

• PRO: fewer indices
• PRO: multilingual search is possible
• CONS: less fine grained analytics

Going live

Go through this checklist

Re-indexing

Algolia suggests to do a full reindex regularly but to do batched updates by default.

Open questions

• Why is sandbox useful?

TL;DR: I don't always know what I want, but I know what I like.