Best Practices
The handful of decisions worth getting right before you scale up, and the mistakes that are easy to make and annoying to undo.
Before you go live
Saving a new Recommendation puts it live straight away, so run through this short list first. Ten minutes now saves support tickets later.
- [ ] Check the Live Preview isn't empty. If the preview says nothing matches, the storefront won't show anything either.
- [ ] Test your sparsest category. A block that fills on your biggest category can come up empty on your smallest. That's correct behaviour, but you should know it's happening.
- [ ] Test in a private window. You're logged in, your browser caches pages, and your visit has history: three ways to see something a shopper won't.
- [ ] Check for duplicate boxes. Your block and WooCommerce's both rendering is the most visible avoidable mistake.
- [ ] Check on a phone. The free plugin uses 4 columns on desktop, 3 on tablet and 2 on mobile. Make sure your product images still read well at that size.
- [ ] Read your heading as a shopper would. Does the block deliver what the heading promises?
- [ ] Confirm the status is Active. Only Active Recommendations render.
Start with two, not six
Every coordination limit ships off, so out of the box every eligible block shows. That's a sensible starting point, but it's not a goal: fewer, better offers convert better than a wall of boxes.
Build two Recommendations that earn their place. Measure. Add a third only when you can say what job it does that the first two don't. Six blocks on a product page isn't six times the revenue; it's noise, and each block gets less attention.
When pages start to feel crowded, turn on Settings → Recommendation Manager → Limit blocks per page and No repeated products.
Match the offer to the moment
Each step of the shopping journey suits a different kind of offer:
| Where | What works | What doesn't |
|---|---|---|
| Product page, high (after meta / after add-to-cart, Pro) | A better version of what they're looking at, or a genuine bought-together companion | Browsing grids. They're still deciding about this product |
| Product page, low (before or after tabs) | "More like this" browsing, related products | Hard upsells. They've scrolled past the decision point |
| Cart | Small, cheap, obviously related additions | Anything that needs real thought. They're about to check out |
| Empty cart | Best-sellers, popular entry points | Nothing at all. An empty cart with nothing on it is a dead end |
| Checkout Pro | One small add-on, shown once | A grid. Don't distract from paying |
| Thank-you page Pro | One focused offer, one decision | A grid of four. Trust is high; don't spend it on a menu |
Always add a sorting rule
With no sorting rules, matching products come back in random order, reshuffled on every page view. That's fine for a "discover something" block, but for most blocks you'll want a stable, sensible order.
Every rule you add counts the same, and they're blended into one order. There are no weights to tune: you shape the order by choosing which rules to include. Popularity on its own is a solid default.
Filters are hard limits; sorting is a preference
This is the most useful distinction when a block behaves unexpectedly.
A filter removes products completely. Add Price Is Over 50 and a 49 product can never appear, even if it's the perfect match and nothing else qualifies.
A sort expresses a preference. Sort by Price — Low To High and expensive products go last, but they still appear when nothing better exists.
Use a sort when you want a tendency. Use a filter only when a product genuinely must never appear: out of stock, wrong category, already in the cart.
It's also the usual fix for "my block is empty on some pages": you probably used a filter where a sort would have done.
Don't over-filter a bought-together block
This one's counterintuitive. A bought-together block's most valuable pairs cross categories: the camera and the memory card, the boots and the socks.
Filter to "same category" and you've excluded exactly the products that make the block worth having. For bought-together offers, keep filters light and let the co-purchase data do the work.
Know where your bought-together data comes from
The Frequently Bought Together and Others Also Bought Pro sorts, and the Store Statistics Pro selection method, all read co-purchase pairs that the Pro add-on mines from your orders.
- On the free plugin alone, nothing mines those pairs, so the Frequently Bought Together sort has no data and has no effect on the order. Add Popularity next to it so the block still has a sensible order.
- With Pro, a product with no co-purchase history still gets nothing from the bought-together rule, so the other rules decide. Pairing it with Popularity keeps your long tail covered.
Set headings that are honest
"Frequently bought together" on a block that's really sorted by popularity is a small lie, and shoppers notice small lies. Match the heading to how the block actually picks products:
| How products are picked | Honest heading |
|---|---|
| Real co-purchase data | "Frequently bought together", "Often bought with this" |
| Popularity | "Popular right now", "Best-sellers" |
| Same category | "More in this category", "You might also like" |
| Pinned by hand | "Our picks", "Recommended by us" |
If you leave Heading Shown to Shoppers empty, the block uses "You may also like". It's safe, but it doesn't do much work.
Make it Inactive instead of deleting
An Inactive Recommendation costs nothing and keeps its configuration. Delete only when you're certain; deleting can't be undone.
Analytics keeps the revenue a deleted Recommendation earned, labelled (deleted), because the money was real. But the configuration is gone for good.
Duplicate to experiment
The fastest way to try a variation: Duplicate it on All Recommendations, change one thing, and activate it. Duplicates are "saved as inactive", so nothing doubles up on your storefront by accident. Remember to make the original Inactive if both target the same position, or only one of them can win it anyway.
Change one thing at a time
There's no A/B testing in this release, so comparing one week with the next is your honest method, and it only works if each week has a single change in it.
A useful loop:
- Set a 7-day range in Analytics.
- Note views, CTR, conversions and revenue.
- Change one thing.
- Wait a full week, because weekday and weekend traffic differ.
- Compare.
Reading the numbers without fooling yourself
| Pattern | Usually means |
|---|---|
| Views near zero | It isn't rendering. That's not a content problem: start here |
| Views high, CTR under ~1% | The wrong products, or a heading that doesn't match them |
| CTR healthy, conversions near zero | Interest without intent. Usually price |
| Revenue concentrated in one product | It may be pinned, or dominating the sort. Check that's deliberate |
| Everything flat after a change | You changed two things, or didn't wait a week |
Turning on coordination on a small catalogue
On a store with a few dozen products, No repeated products plus narrow filters can leave later blocks on a page with nothing left to show. Those blocks are held back rather than shown half-empty.
Try these in order:
- Widen the pool. Use auto filters, or a broader filter. This is nearly always the real fix.
- Reduce how many blocks share that page, or lower Products to Show on the earlier ones.
- Turn No repeated products off if your catalogue really is too small for it.
Don't hide WooCommerce's sections unless you mean it
The WooCommerce Default Showcase switches under Settings → General hide WooCommerce's Related products, Up-sells or Cross-sells on every page. If your own block doesn't cover every page (because of a condition, a limit or a narrow filter), those pages end up with nothing at all.
They're the only way to switch off WooCommerce's sections, so use them once you've decided your blocks should take over, and check a few awkward products afterwards.
Keep an export
Settings → Export / Import → Download .json Backup takes seconds and gives you a file you can restore anywhere. Do it before a big reorganisation, before turning on Clean-up Data on Deletion, and before importing with Replace All.
Next
- Recommendation Manager: the coordination rules in full.
- Analytics: the numbers behind the loop above.
- Tutorials: these practices applied end to end.