Start with Feed Builder alone. Add Data Builder when building the feed takes too long.
Feed Builder alone#
Feed Builder works out every row each time the feed runs: it loads each product, looks up stock, prices, images, custom fields and variants, and writes the row. That is simple and always current, and it suits most stores.
The cost grows with the catalog. On a test store with 14,499 products and 42,478 variants, a Google feed with variants took about 142 seconds per run. The same work happens every time any product changes, because a change rebuilds the whole feed.
Feed Builder with Data Builder#
Data Builder stores each product’s and variant’s finished row in a custom field. It recomputes a row only when that product changes, so a stock change on one product recomputes one product.
The Feed Builder feed then maps that one custom field with the Use Value as Feed transform. Each run reads the stored rows and writes them out without working anything out. On the same test store, the full feed of 46,218 rows built in seconds.
The extra steps:
- Run the field’s All Products task once after setting it up. On the test store this took about 11 minutes.
- Keep the field’s Queued task enabled so changes are recomputed.
- Put the CSV column line in the feed’s Header text, because the feed can’t see the columns inside the stored rows.
The ready-made channel files ending in -data-builder.xml set all of this up.
Quick guide#
| Situation | Use |
|---|---|
| A few thousand products | Feed Builder alone |
| The feed runs in under a few minutes and that’s fine | Feed Builder alone |
| A feed you’re still designing | Feed Builder alone, then switch |
| Tens of thousands of products or variants | Both |
| Frequent stock changes that would rebuild a large feed many times a day | Both |
| You need a filter that removes products before the row is built | Either: the filter is on the Feed Builder feed in both cases |