Every category playbook I've written rests on the same assumption: the shopper is deciding whether they want your product. The hero's job is to get noticed, the stack's job is to persuade, and the whole exercise is an argument.
Replacement parts break that assumption completely.
The person shopping for a refrigerator water filter has already decided what they want. They want the thing that fits the fridge in their kitchen. They are not evaluating you. They are verifying a match — and if they can't verify it in about two seconds, they move to the listing where they can.
That single change rewrites the whole creative job. Persuasion is almost irrelevant here. Verification is everything. And after reviewing 50,000+ listings, I can tell you that almost nobody in this category has built their images for verification, because nobody briefed them that way.
This post covers the non-automotive half of the category: appliance filters, vacuum bags and parts, HVAC filters, coffee pods and refills, razor cartridges, printer ink, blender jars, pool and spa cartridges, mower blades, shower heads, remote controls, charging cables, hose fittings, toothbrush heads. Vehicle fitment is a different animal with its own data standards and its own widget — that lives in the automotive and car accessories playbook, which I'm not going to re-litigate here.
The four mechanics that make this category different
1. The shopper verifies with the image, not the title
This is the finding that should reorganise your entire stack, and it comes up constantly in seller forums in the same shape: when the image on the listing shows a different variant than the one the title specifies, buyers order based on the image, receive the item described in the title, and file a not-as-described claim. The model number was right there in the header. They didn't read it.
That isn't stupidity. It's how people actually shop. A part number is seven characters of alphanumeric soup that all look alike — DA29-00020B, DA29-00003G, HAF-CIN. A shape is instant. So the shopper glances at the picture, decides "yes, that's the one that comes out of my fridge," and clicks.
Which means your hero image is functioning as a fitment tool whether you designed it to or not. In most of this category, it's the only fitment tool on the page.
2. There is usually no fitment widget to hide behind
Automotive has Amazon Confirmed Fit, a garage, a vehicle selector, and an entire industry of ACES and PIES data feeds behind it — the standards moved to ACES 5.0 and PIES 8.0 in March 2026. PC components have Partfinder. Those categories have a machine that answers the fit question before the shopper reaches your creative.
For most of the rest of this category, there is no such machine. There's a title, some bullets, a compatibility list nobody scrolls to, and your images. The structured "compatible with" attributes exist and you should absolutely fill them — more on that later — but they are not rendered as a confident green tick next to the buy box on the average appliance-part listing.
Where there's no widget, the frame is the widget. Confirm what your own category actually renders rather than assuming a fitment stripe exists because you saw one on a brake pad.
3. A wrong-fit purchase is a negative sale, not a lost one
I keep saying this and it matters more here than anywhere.
A shopper who can't verify fit and skips you costs you nothing. An impression you didn't convert. That's the entire consequence.
A shopper who guesses wrong and buys costs you the click, outbound fulfilment, return processing, frequently the unit itself, the contribution, and a contribution to a return rate that past a category-relative threshold hangs a badge on your listing that depresses conversion on every future session — including the paid ones.
On a $19 filter carrying maybe $8 of contribution, a round-trip wrong-fit purchase is roughly $20 of damage. Not missed upside. Gone. To sell a unit you then had to take back and can't resell.
Which produces the strategic position for the whole category: a frame that costs you a sale by being clear is nearly always cheaper than a frame that wins one by being vague.
4. The purchase repeats, and the images decide whether it repeats with you
Filters, pods, cartridges, blades and bags are bought on a cadence. Every six months, every three months, every forty tanks.
That means two things most brands never account for. First, the buyer's second purchase decision is made by someone who now knows exactly what they need — a much easier shopper to win and a much easier one to lose on a detail. Second, the frame that made the first purchase verifiable is the frame that makes the reorder frictionless. If your images made them confident in March, they'll find you again in September. If they had to cross-reference the manual, they'll buy whatever's easiest this time.
The hero image: show the part, at real scale, with the diagnostic feature facing camera
Standard main-image rules apply and are not flexible: pure white background, roughly 85% frame fill, no text, no logos, no watermarks, no props, 1600px or larger so zoom engages.
Within those rules, three things decide whether your hero works in this category:
Fill the frame, properly. Audits routinely find parts sitting at 55-65% fill because the product is an awkward shape and the designer composed it with breathing room. A 2000px canvas renders around 280px in mobile search. At 60% fill, you have thrown away roughly seventy pixels of product that existed in your file, and you cannot buy more thumbnail at any price.
Point the diagnostic feature at the camera. Every part in this category has one feature the buyer identifies it by — the twist-lock tab on a fridge filter, the pleat pattern and frame colour on an HVAC filter, the blade mounting hole, the cartridge's connector end, the gasket profile. Most product photography shows the part from its most attractive angle, which is very often the angle that hides the identifying feature. In this category the identifying feature wins over the attractive angle every single time.
Give the silhouette a chance. If you sell a four-pack, show a single unit plus a clean indication of the count rather than four overlapping units that read as one indistinct mass at thumbnail size. The shopper is matching shapes.
One thing not to do: do not photograph the retail box. I see this constantly on consumables, and it's the same mistake as photographing a seed packet instead of what grows. The box is not what they're matching against. The part is.
Slot two is the fitment frame, and it is the most important image in the stack
Not the lifestyle shot. Not the brand panel. The fitment frame.
This is the frame that answers "is this the one that comes out of my machine," and it has three jobs:
Show the model list in type that survives reduction. A readable panel with the manufacturer part numbers and the appliance models the part fits. Not nine-point grey. Heavy weight, high contrast, on a solid field — thin strokes get interpolated toward nothing on the resize and then discarded by JPEG compression, which specifically targets exactly that kind of high-frequency detail. Your brand book specifies a typeface. It does not specify which weight survives delivery. That's a merchandising decision and it's yours.
Show the part in position. One photograph of the part installed — seated in the fridge door, slotted in the vacuum housing, in the HVAC return, on the mower deck. This does something no list of numbers can: it lets a shopper who doesn't know their model number verify by recognising the location. A meaningful share of this category's buyers do not know their model number and are working from memory of what the old one looked like.
Show the old one next to the new one. Two units side by side, same lighting, same angle, one visibly used. This is the single most under-produced frame in the category and it maps precisely to how the purchase actually happens — somebody standing at an open appliance holding a grey, scaled, slightly crushed version of your product, trying to decide if yours is the same thing.
If you can only ship one new image this quarter, ship this one.
The part-number frame: the only category where the most important element is a number
In every other category I write about, I'm arguing against text-heavy frames. Here I'm arguing for one.
Build a frame whose entire job is the alphanumeric. Your part number, large. The OEM numbers it replaces, listed. The appliance brands, listed. Nothing else in the frame competing for attention.
Two reasons it earns the slot. The shopper searching a part number is the highest-intent traffic in the category and they are literally pattern-matching a string. And on-image text at full resolution is read by the machine layer — Amazon's own AI surfaces, third-party agents, screen readers — which is a rare case where the same asset serves the human and the non-human reader identically.
One critical constraint: get the compatibility language right or you'll lose the listing. Amazon's guidance on compatible products requires a specific construction: your brand name, your product name, then "for" / "compatible with" / "fits" / "intended for", then the other brand. You may use another company's brand name in text to state true compatibility. You may not use their logo. "Brand X replacement filter" reads as a claim to be made by Brand X and that is how brand-name misuse complaints start. "Compatible with Brand X, model DA29-00020B" is the construction that survives.
This applies on images too, not just titles, and it's the thing most likely to get an otherwise excellent fitment frame pulled. Confirm the current wording requirements in your own category style guide rather than trusting a blog post — including this one.
The trust frame: aftermarket has a credibility problem and you can answer it
This category has a genuine, earned trust problem. Buyers have been burned by cartridges that leaked, filters that restricted flow, blades that were soft, pods that didn't seal. The shopper's second question, right after "does it fit," is "is this junk."
Three frames that answer it without a claim you can't back:
The construction close-crop. Photograph the thing the cheap version gets wrong: the gasket, the seal, the carbon block, the weld, the pleat density. Zoom-eligible resolution. This works because buyers in this category are unusually technical about a single detail — they know exactly which part failed last time.
The specification as a number, not an adjective. "Rated 300 gallons / 6 months" beats "long-lasting." "MERV 11" beats "premium filtration." "0.5 micron" beats "advanced." If the buyer's decision is technical, give them the technical fact — and give it to them in the frame, not only in the bullet nobody scrolls to.
The certification, only where it is genuinely yours. This is where I'd urge real caution. Certifications in this category are scoped narrowly: a certifier's listing covers a specific model under specific claims and conditions, not a product family and not a lookalike. If your exact SKU and your exact claim appear in the certifier's current public record, put it in the frame. If they don't, leave it out entirely. An unsupported certification mark on a live listing image is a different class of problem from a weak photograph, and it doesn't resolve as a return — it resolves as a complaint.
The quantity and cadence math frame
This is the closest thing to free conversion available in this category and it's missing from most listings.
Shoppers think in time, not in units. Nobody wants three filters. They want to not think about this again until spring.
- "3 filters — 18 months of coverage at the recommended 6-month change"
- "Pack of 12 — one year at one blade a month"
- "60 pods — about two months for a two-cup-a-day household"
- "4 cartridges — a full pool season for a 15,000-gallon pool"
Then show the pack-size ladder with what each one covers. This does two jobs at once: it moves buyers up a tier, and it prevents the wrong-size purchase that comes back.
Add the per-unit price if you're winning on it, because the deliberate buyer in this category is doing that arithmetic in their head anyway and you may as well do it for them correctly.
The what-actually-arrives frame
Cheap to shoot, disproportionately effective, and it prevents a specific expensive return.
Two panels: everything in the package, laid out flat and counted, next to what it looks like installed. Gaskets, O-rings, adapters, mounting hardware, the little plastic tool.
Trade write-ups on parts listings put the reduction in customer contacts from a proper what's-in-the-box frame at around 30%. Treat that as directional rather than a number to plan against — it's a vendor-side figure and nobody has published a controlled study on your catalogue. The mechanic underneath it isn't in dispute, though: half the "wrong item" complaints in this category are actually "I expected the adapter to be included" complaints, and one frame answers them.
And if an adapter is sold separately, say so at the same type size as everything else. Fine print that only a lawyer would find is not a disclosure, it's a return you've pre-purchased.
Eight subcategory rules
Appliance water filters. Twist-lock geometry facing camera in the hero. The installed frame is non-negotiable because buyers identify by location. State the capacity in gallons and months. Certification claims only where your exact SKU is in the certifier's record.
HVAC and furnace filters. Nominal versus actual dimensions is the single biggest return driver, and the industry convention where a "20x25x1" is actually 19.5x24.5x0.75 is invisible to a first-time buyer. Put both numbers in the frame. Show MERV rating as a number with a plain-words translation of what it catches. Show the airflow-direction arrow, because installing it backwards is a real and common complaint.
Vacuum bags, filters and brush rolls. Show the machine model list and the part in situ. Brush rolls in particular need the end-cap geometry visible, because that's the part that differs between models that otherwise look identical.
Coffee pods and refill cartridges. Machine compatibility is everything and the machine generation matters — a pod that works in one generation and not the next is a review waiting to happen. Show the pod, the machine it goes in, and the count-to-duration math. If there's a reusable filter in the box, show it.
Razor cartridges and toothbrush heads. Handle compatibility, shown as a physical connection close-up. These are high-frequency repeat purchases where the second purchase is the whole business, so make the reorder identifiable in one glance.
Printer ink and toner. Printer model list, cartridge silhouette, page yield as a number, and whether it's standard or high-yield stated plainly. Chip compatibility is a genuine functional risk and worth its own line.
Pool and spa cartridges. Dimensions with both the top and bottom fitting shown, because two cartridges with identical filter areas and different end caps are the classic wrong-fit purchase. Season-length coverage beats square footage for most buyers.
Mower blades, trimmer line, small engine parts. Mounting hole geometry and centre-hole dimension in the frame. Show the blade length measured, not stated. Include the deck sizes it fits.
The machine layer
Queries in this category are unusually literal and unusually long: part numbers, model numbers, brand-plus-model strings. That makes the structured data disproportionately valuable.
- Fill the compatibility and "compatible with" attributes to completion. A blank field is not a weak signal, it's an absent one — you're not in the filtered candidate set and you're not in the AI layer's consideration set either.
- Put the part numbers in text fields — bullets, attributes, Item Highlights — not only baked into a JPEG. A model reading your page with the pictures off should still be able to answer "does this fit a Samsung RF28."
- Keep the on-image copy literal and short. "Fits Samsung DA29-00020B" transmits to both readers. "Pure water, purely simple" transmits to neither.
- Alt text on A+ modules costs seconds per module and is the cheapest legibility work in the category.
Read one of your top listings in reader mode, with the pictures stripped, and see what survives. On most parts listings I've audited, the answer is a title and a wall of empty attributes.
Nine anti-patterns
- Photographing the box instead of the part. The box is not what they're matching against.
- The attractive angle that hides the fitment feature. Identification beats aesthetics in the one frame that has to be identified.
- Compatibility lists only in the bullets. Below the fold, past the buy box, invisible on mobile.
- Nine-point compatibility type on a gradient. Deleted by the resize and the codec before anyone reads it.
- Using another brand's name without "compatible with" or "fits." Not a style question. A complaint route.
- Using another brand's logo to show compatibility. Explicitly not permitted.
- Certification marks that don't cover your exact SKU and claim. The most expensive image mistake available in this category.
- Stating pack count without duration. The shopper is buying time, not units.
- Stacking OEM-equivalent claims you can't substantiate. "Same as OEM" is a claim. "Replaces part number X" is a fact.
The 25-minute audit
Run this on your top three SKUs by contribution, on a phone, in live search results — not on a monitor and not in a preview.
- Look at your hero for one second. Could you match it against a part in your hand? If your own team hesitates, a stranger has no chance.
- Swipe to slot two. Does it answer fit, or is it a lifestyle shot?
- Find the frame showing the part installed. On most listings there isn't one. That's your shot list.
- Read 90 days of return comments filtered to the not-as-described and doesn't-fit codes. It's a messy dataset — most buyers leave it blank and the rest are typing fast and annoyed. You're looking for the same complaint three times.
- Check your compatibility language against the required construction in every place it appears — title, bullets, and on-image copy.
- Count your configuration variables against your frames. If you sell nine model-specific variants and have one image set, that's the finding.
Four of those six findings are fixable from files you already own, this week, with no photographer involved.
FAQ
Can I put the compatible brand's name on my images? In text, to state true compatibility, with the required qualifier in front of it — "compatible with," "fits," "replacement for," "intended for." Not as a logo. Not as a standalone brand claim. And check your own category's current style guide rather than working from a blog post, including this one.
Should the hero show one unit or the whole multipack? One unit, clearly, with the count communicated cleanly. Four overlapping units read as one indistinct mass at 280 pixels and you've lost the shape match, which is the only thing the hero is for.
What's the highest-impact frame if I only ship one? The old-part-next-to-new-part comparison, shot at real scale with the identifying feature visible. It maps exactly to how the decision gets made — someone holding the used one, trying to match it.
My return rate is high and the images look fine. Where do I start? With the return comments, not the images. Find the complaint that appears three or more times, then find the frame that was supposed to prevent it. In this category, for the complaint costing you the most, there is usually no frame at all — not a weak frame, none.
Is it worth this much work on a $19 filter? It's worth more on a $19 filter than on a $90 product, because the purchase repeats and the return math is brutal at low price points. Two points of conversion on 4,000 monthly sessions is roughly 80 units, and the returns you prevent are pure contribution recovered, not incremental revenue you had to buy.
If you want to go deeper on the sequencing side, the image stack sequencing guide covers slot-by-slot logic, and product size and dimension images covers the scale work that underpins most of the fit frames here. For the returns-driven diagnostic side, reducing return rate through listing images is the companion piece to the audit above.