Product Recommendations for Electronics and E-commerce Stores: Accessories That Fit, Brand Affinity and More
Open a phone on most electronics stores and the row under it offers more phones, often the same phone in two other colours. Buy it, and the thank-you page offers phones again. The shopper has already chosen a phone. What they need now is a case that fits it, a charger and a screen protector.
Electronics has two things most catalogues do not: products that have to fit each other, and shoppers who stay with a brand. This guide covers the recommendations that work for electronics stores, and for any shop that sells add-ons: where each one goes, and how to tell which one is selling. It is the companion to product recommendations for fashion stores, and it uses Personyze’s recommendation engine; the practices hold whatever you use.
| Widget | Where | What it shows | What it learns from |
|---|---|---|---|
| Accessories that fit | Product page, cart, thank-you page, email after an order | Cases, chargers, cables and add-ons made for this exact product, priced below it | What shoppers bought together, checked for fit |
| Their brand first | Any widget, as a Match the visitor rule | The shopper’s own brand ahead of the rest | The brands they viewed and bought |
| Similar items | Product page, for shoppers still choosing | Other models of the same kind, at a similar price, each once | What shoppers viewed together |
| Time to restock | Home page, cart, email | Ink, filters and batteries when they usually run out | How often shoppers rebuy each item |
Why “more of the same” fails in electronics
Most recommendation algorithms count which products turn up together, in the same visits or the same orders. In an electronics store that goes wrong in three ways:
- A purchase ends the search. Nobody buys a second TV the week after the first. “Customers who viewed this also viewed” keeps offering TVs to someone who has just bought one.
- An accessory either fits or it is useless. A case for last year’s model, a mount too small for the screen, a battery for another camera. Orders mix them: a family buys a new phone and a case for the old one in the same basket, and the data says the old case “goes with” the new phone.
- Shoppers stay with a brand. Someone whose phone, watch and earbuds are all one brand is shopping inside that brand. A list that ignores it leads with a brand they never buy.
The fix is to know what each product is, and what it fits. Personyze reads each product’s picture and title once and records it, then blends that with what real shoppers bought and viewed. Bought together turns into accessories that fit, viewed together turns into similar items at a similar price, and the shopper’s own history decides which brand comes first.
It is tuned per industry, too. Personyze works out the kind of shop from its catalogue and adapts: an electronics store gets accessories that fit the exact product, in the same brand ecosystem; a fashion store gets complete the look; a furniture store gets the rest of the room; a beauty or pet store gets the refill before it runs out. Nothing to configure. Every version is laid out in recommendation engine algorithms. The seven practices below are the electronics version, and most of them hold for any store that sells add-ons.
1. Accessories that fit, never another phone after a phone
Once a shopper has chosen the main item, or bought it, the most useful row is what makes it work: accessories made for that exact product, each costing less than it. After a phone: its case, a charger and a screen protector. After a TV: a soundbar, a wall mount and an HDMI cable. After a camera: a memory card, a spare battery and a bag. Never another phone after a phone.
The best raw material is what your shoppers actually bought together, the logic behind customers who bought this item also bought and frequently bought together. But only after it is checked for fit:
- Made for this product. The case for the model on screen, not for last year’s; the battery for this camera; the mount for this size of screen. Compatibility is the whole point of an accessory row.
- In the same brand ecosystem. The charger, earbuds and watch that belong with this phone’s brand.
- Never the same product again. Not the phone in another colour, not the TV the shopper has just bought.
- Items bought with everything stay out. Gift cards pair with every product, so they say nothing about any of them.
- Add-ons priced below the main item. A $39 case beside a $799 phone is an easy yes; a second phone is not.
- The most-bought pairs first. Rank pairs by how many shoppers made them, so a combination many people bought leads one that two people happened to buy.
CARD
Personyze does this check by itself: reading each product tells it that a case is made for another model, or that two phones are the same phone in two colours. Put the row under the product, in the cart for what is already there, and on the thank-you page and in the follow-up email for what was just bought. Why the add-on belongs right beside the main item is in the benefits of bundle sales.
2. Their brand first: brand affinity
Electronics shoppers keep to a brand. Their phone, watch and earbuds work best together, their chargers already fit, and they know the menus. A shopper who always buys the same brand, say Apple, should see that brand first: on the home page, in the category they are browsing, and among the similar items under a product.
That is brand affinity, and it comes from the shopper’s own history: the brands they viewed and bought. In Personyze it is one rule in Match the visitor, “Brand from their own history”. The algorithm still picks the products; the rule moves the shopper’s brand to the front.
- From their own history. What they viewed and bought, so it works for a shopper who has never filled in a form.
- A re-order, not a filter. Other brands stay in the row, after theirs. A shopper ready to switch still sees the alternative.
- A first visit keeps the algorithm’s order. With no history there is no brand to put first, so nothing changes.
- Any brand field. If your catalogue keeps the brand under another name, such as Manufacturer, the same check runs on that field.
3. Similar items for shoppers still choosing
Not every shopper on a product page has decided. Someone on their third phone is comparing, and a case for a phone they may not buy comes too early. For them, show similar items: other models of the same kind, at a similar price.
This is where “viewed together” earns its place. Shoppers who view items together are usually comparing alternatives, the logic behind customers who viewed this item also viewed. Shoppers who buy items together are setting something up. Feed each widget from the signal that matches its job:
oror+++- A similar price. Alternatives stay in the price band of the item on screen, so a $799 phone is compared with other phones near $799, not with the most expensive one in the shop.
- Each product once. A phone in three colours and three storage sizes is one card, not nine. The choice is made on the product page.
- Best match first, when you want the closest models at the front.
- Their brand first, with the rule from practice 2.
Put it higher on the page than accessories that fit, where a shopper who is still deciding sees it before scrolling. If you want both rows on product pages without crowding them, a targeting rule can show similar items to shoppers with nothing in the cart, and accessories to everyone else. New models come out all the time, and a model nobody has viewed yet still gets similar items: Similar Items (AI), once switched on, reads titles, brands, descriptions and categories, so a product has neighbours before its first visitor.
4. Time to restock, for what runs out
Many electronics are bought once, but what they use is bought again and again: ink, filters, batteries. Time to restock brings each of these back to a shopper when they typically run out, learned from how often people rebuy each item.
- Learned per product. Ink keeps its rhythm and batteries keep theirs, from how often your shoppers rebuy each one, not one rule for the whole store.
- The product they already chose. The ink set they ordered for their printer, so it still fits.
- Not before it is due. Until then it stays out of their recommendations, and the printer itself, bought once, is never offered again.
- On the home page, in the cart and in email. In email the picks are made when it is opened, not when it was sent; see open-time email recommendations.
It is the same idea as the grocery replenishment example, applied to the consumables an electronics store sells.
5. Match the visitor: their brand, then a price like theirs
Brand is the first of two rules that do most of the work in electronics. The second is a price like theirs: items near what the shopper usually looks at, so someone shopping mid-range is not shown the most expensive TV first, and someone who buys premium is not shown the cheapest.
Stack them in order, brand from their own history first, then price like theirs. The algorithm picks; the rules re-order its own top picks for each shopper and never fill the widget with random items.
The same rules work wherever recommendations run: on the site, in recommendation emails and in feeds. The live version with three shoppers side by side is under Match the visitor, and the Match the visitor guide has a recipe for each kind of site, electronics included.
6. Put each widget where it does its job
- Product page. Accessories that fit under the product; similar items higher up for shoppers still choosing. Never the product itself.
- Cart. Accessories for everything in the cart, priced below the main items, never what is already there: the last moment a $19 screen protector is an easy add.
- Home page. Their brand first for returning shoppers, and time to restock when something is due. Popular items on a first visit.
- After the order. On the thank-you page and in the follow-up email, what completes what they bought. Never more of it.
Keep every list full. When one source runs thin, a new product with few purchases or a shopper with little history, the next one fills in, so a row is never half empty. The page-by-page setup is in the product recommendations playbook, and more ideas are in product recommendation examples.
7. Measure each recommendation source, not only the widget
An accessories row is fed by more than one source: what shoppers bought together, what is made for the product, and a fallback such as best sellers when both run thin. Measure only the widget and you cannot tell which of them is selling.
Personyze measures clicks and purchases for each recommendation source, so you can see whether bought-together pairs carry the row, and how often the fallback is doing the work. Whether the widget adds sales at all is a separate question: keep a share of visitors on the page as it was, and compare orders and revenue, not clicks. How to read that test is in how to read A/B test results.
Setting it up
There is little to set, and that is deliberate:
- Set up with AI. Connect your store — Shopify, Magento, WooCommerce, BigCommerce or Shopware, a product feed, or Personyze reads your product pages — and the guided setup reads your products, what visitors look at and your kind of shop, finds the spot on your home, product and cart pages, and places each widget in your store’s look. Nothing changes on your site until you publish.
- Nothing to configure. Personyze works out from your catalogue that it is an electronics store, and adapts by itself.
- Reading product pictures is one switch, in Recommendation settings.
- The cost comes first. Before you switch it on, you see the estimated cost in AI credits for your own catalogue, and you choose Best quality (the default) or Standard.
- Their brand first is one rule. Add Brand from their own history, then Price like theirs, under Match the visitor in the widget’s editor.
Connect your storeOr a feed, or we read your pages
It reads your storeProducts, views, your kind of shop
Where to show themHome, product and cart pages
Check it, then publishAs a real shopper would see it
Then check it as a real shopper before anyone sees it. Product recommendations, start to finish walks through the preview, the test against no widget and the results.
The checklist
- Accessories that fit under the product, in the cart and after the order: made for this exact product, in the same brand ecosystem, never for another model.
- Never another phone after a phone: nothing the shopper has just bought, nor the same product in another colour.
- Their brand first, from their own history, as a re-order rather than a filter.
- Similar items for shoppers still choosing: the same kind of product, a similar price, each product once.
- Time to restock for ink, filters and batteries, when each one is due and not before.
- A price like theirs after the brand rule.
- Add-ons priced below the main item.
- Clicks and purchases read per recommendation source, and the widget tested against no widget.
More to look at: AI product recommendations on an online store, page by page; recommendation algorithms explained; and for clothing, shoes and accessories, product recommendations for fashion stores.
Questions people ask
What product recommendations work best for an electronics store?
Four, each in its place. Accessories that fit under the product, in the cart and after the order: add-ons made for that exact product, priced below it. Similar items for shoppers still choosing: other models at a similar price. The shopper’s own brand first, from what they viewed and bought. And time to restock for consumables such as ink, filters and batteries.
How do you recommend accessories that actually fit?
Start from what shoppers bought together, then check each pair for fit: the accessory must be made for the model on screen or in the order, and in the same brand ecosystem. Drop the same product in another colour, accessories for other models and items bought with everything, such as gift cards, and rank what is left by how many shoppers bought it together.
How do I show shoppers their favourite brand first?
Add the rule Brand from their own history to the widget, under Match the visitor. It re-orders the algorithm’s own picks so products of the brands the shopper viewed or bought come first. Other brands stay in the row, after them, and a first-time visitor keeps the algorithm’s order.
What should a shopper see after buying a TV or a phone?
What completes it, not more of it. After a TV, a soundbar, a wall mount and an HDMI cable; after a phone, its case, a charger and a screen protector. Show them on the thank-you page and in the follow-up email, each priced below the item bought.
How do I know which recommendations are selling?
Clicks and purchases are measured for each recommendation source, so you can see whether bought-together pairs, products made for the item or a fallback carry the widget. To measure what the widget adds, compare against a share of visitors who see no widget.
Book a demo with a personalization expert
30 minutes with a personalization expert. Bring your stack, your goals, your skepticism. We'll show you what changes when every visit feels like the only one.