It was a Tuesday afternoon in a Sharjah-based auto parts trader's office, and the sales coordinator had three phones on the desk. A customer on WhatsApp wanted brake pads for a 2018 Land Cruiser. She was scrolling her own catalog on her own phone, screenshotting product cards one at a time, cropping them, and sending them across. Eleven screenshots in, the customer stopped replying. Her supplier list runs past 4,000 SKUs. The catalog she had painstakingly uploaded to WhatsApp held 500 of the fast movers, and she could not find a way to get more than a handful of them in front of a buyer without turning the chat into a scrapbook.
That is the wall almost every product business hits with WhatsApp commerce. The problem is structural: WhatsApp was never designed to be a searchable storefront, and the limits in the platform quietly enforce that.
The short answer up front
WhatsApp catalog message limits are simple and unforgiving: Meta's catalog documentation caps a multi-product message at 30 products, and its Cloud API inventory guide allows one catalog per WhatsApp Business Account. For a 500-product inventory, the working pattern is therefore not "send the catalog" but "send the right 30". You break the single catalog into collections inside Meta Commerce Manager, expose those collections through an interactive list or button menu, then send a multi-product message containing only the products in the branch the customer chose. Anything deeper than that goes to a catalog link or a hosted web view.
That is the whole answer. The rest of this piece is about why that shape works, where it breaks, and how to build it without turning your WhatsApp number into something Meta throttles.
We are Learnmind, and we install AI communication systems for UAE service businesses: traders, clinics selling product lines, salons with retail shelves, and distributors who sell to other businesses over WhatsApp. The Sharjah scene above is an illustrative composite of several engagements, and the pattern below is what survives contact with real inventory.
Auditing what you actually need in the catalog
Before anyone touches Commerce Manager, we sit with the sales team and ask which products actually get asked about in chat.
The answer is always a fraction of the master list. Pull six weeks of WhatsApp conversations for a hardware distributor whose master sheet runs to thousands of line items and the same few categories come back over and over. Every serious enquiry starts with a category question. That changes what you build: the real deliverable is a decision tree that ends in a small set of products.
Why uploading everything is the wrong instinct
Since only one catalog exists per WhatsApp Business Account, everything you upload lives in the same container. A customer who opens your catalog link and sees 500 unsorted items leaves. Think of a container port: the yard stacks thousands of boxes, and the crane lifts only the containers named on today's manifest. Your catalog should run the same way, with depth held in the yard and a small, named selection lifted into the chat on request.
The fields that decide whether a card converts
Catalog items need a clean image, an exact title, a price and a retailer ID. The retailer ID is the piece people skip and later regret, because it is the handle your automation uses to assemble multi-product messages on the fly. If your retailer IDs match the SKUs in your inventory system, you can generate catalog messages from stock data. If they are random strings, someone will be maintaining a mapping table by hand forever.
- Title: lead with the attribute customers search by (vehicle, size, model) instead of your internal naming convention.
- Image: one product, plain background, legible at thumbnail size. Chat cards are small.
- Retailer ID: identical to your ERP or stock sheet SKU, no exceptions.
- Availability: keep it current, because a customer who taps an out-of-stock card blames you.
Splitting one catalog into collections that mirror how people ask
Collections are the only real segmentation tool you get. Inside Commerce Manager you group products into named sets, and each collection can be surfaced on its own. This is where the 30-product ceiling stops being a limitation and starts being a design constraint that helps you.
Our rule: no collection intended for a multi-product message should exceed 30 items, and most of ours hold between six and twelve. Thirty product cards in one chat message is a lot of scrolling on a phone, and somewhere past a dozen cards the customer stops evaluating and starts feeling overwhelmed.
Naming collections in the customer's language
A cosmetics retailer's internal taxonomy might read "Skincare / Actives / Retinoids". Customers type night cream. Collections named the way customers speak make the menu step self-explanatory, which matters because the menu is the only thing standing between the customer and 500 products.
Worked example: 520 SKUs, three taps to a product card
Take a home appliance trader with 520 items in Commerce Manager. The structure we would build:
- First tap, a list message with 6 rows: Kitchen, Laundry, Cooling, Small Appliances, Spare Parts, Talk to Sales.
- Second tap, a list message with up to 10 rows: inside Kitchen, that is Refrigerators, Ovens, Hobs, Dishwashers, Microwaves and so on.
- Third tap, a multi-product message: 8 to 12 refrigerator models from the matching collection, sorted by what actually sells.
- Anything beyond that: a single catalog link plus a human handover.
Three taps, and no message ever carries more than a dozen cards. The 520 products are all reachable, but the customer never confronts 520 of anything. It is the same logic a freight forwarder applies at a transshipment port: tens of thousands of containers sit in the stack, and the load plan for any one vessel names only the boxes that sail today.
Choosing between multi-product messages, list menus and a plain catalog link
Three delivery mechanisms, three different jobs. Getting them mixed up is the most common build error we clean up.
Multi-product messages are for the final shortlist. They render as tappable cards that push items into a WhatsApp cart, so they are the only format that carries a customer toward an order without leaving the chat. Cap: 30 products, drawn from your single catalog.
Interactive list messages are for navigation. Meta's Cloud API reference allows up to ten rows across all sections combined, with a text-only header and no images, which is exactly why the format works as a wayfinding layer. They are cheap to render, fast to read, and they are how you compress hundreds of options into a few rows.
The catalog link is your overflow valve. One link, one catalog, full browse experience inside WhatsApp. We send it when a customer says they want to look around, when a request spans several categories, or when the enquiry is genuinely exploratory. It costs you nothing in card limits and it never breaks.
When to stop automating and hand the chat to a person
A customer asking for brake pads for a specific model year is past the menu stage. They want the person who knows whether the OEM part or the aftermarket one is the better call this month. We build catalog flows so that any message containing a model number, a quantity above a threshold, or a question about compatibility routes straight to a human with the conversation history attached. The automation's job is to handle the first two taps and leave the closing to your salesperson.
That is a house position we will not budge on. AI should clear the mechanical work off the desk so the expert can be an expert. A bot that insists on walking a wholesale buyer through a menu when he has already named the part is not helping anyone. Keep the tone professional and plain throughout; buyers spending real money read cuteness as amateurism.
Keeping the flow inside Meta's opt-in and messaging rules
Catalog messages are still messages, and the rules that govern all WhatsApp Business Platform traffic apply. Product content sent as an outbound business-initiated message needs an approved template and needs the customer to have opted in. Meta's Cloud API send-messages guide spells out the alternative: when a customer messages you first, a 24-hour customer service window opens, and while it is open you can send free-form interactive content, multi-product messages included, without a template.
That difference decides your architecture. The reliable pattern is customer-initiated: the customer messages your number, and the entire catalog navigation happens inside the resulting service window. Outbound catalog promotion to a cold list is where accounts get into trouble, both with quality ratings and with cost: under Meta's per-message pricing, in force since July 1, 2025, every delivered template message is billed, while free-form messages inside the service window are free.
Pacing, throttling and new templates
If you do run a marketing template that opens into a catalog flow, expect Meta to meter its delivery while it watches how recipients respond; the mechanics are set out in Meta's template pacing documentation. We have written a full guide to template pacing because the surprise catches people mid-campaign. Send the first batch to your warmest segment. Positive customer feedback is what releases pacing: once a threshold is hit, Meta holds further sends while it gathers quality signals, good signals scale the template out to the full audience, and negative feedback such as blocks and reports pauses the template and drops the held messages.
What the automation rules do and do not forbid
Product businesses ask us regularly whether automated catalog responses are still allowed under Meta's current policy direction, usually because they read something alarming. The short version is that the rules target general-purpose AI assistants offered as the product, not a business answering its own customers with automation. We unpicked this in detail in our piece on the chatbot ban myths, and the compliance specifics for UAE operators sit in our breakdown of the platform policy change. A business's own catalog automation sits inside the rules; offering a general-purpose chatbot as the product is what the clause forbids.
Handling the parts of the inventory WhatsApp cannot hold
Some catalogs simply will not fit the model. Variant-heavy inventory is the classic case: one shirt in five colours and six sizes is thirty catalog entries, and it eats an entire multi-product message with a single product. Serialised industrial stock is another, along with anything priced per customer contract.
For those, we stop trying to force the catalog and send a hosted web view instead: a mobile page under your own domain, linked from the chat, filtered to the customer's request before they open it. The chat carries the conversation, the web page carries the depth, and the order comes back into the chat. Nobody has to browse 500 cards and nobody hits a product cap.
Keeping stock and price in sync
A catalog that lies is worse than no catalog. Where a client runs an inventory system with an API, we sync availability and price on a schedule and let the flow skip anything out of stock when it assembles a multi-product message. Where there is no API, we accept a weekly manual refresh and we tell the customer in the chat that prices are confirmed by the sales team, because an unfixable sync problem should be disclosed.
Verifying the flow works before you point traffic at it
Test with real thumbs on real phones. Here is the checklist we run before a catalog flow goes live on a client's main number.
- Card rendering: open every multi-product message on both an iPhone and a mid-range Android. Titles truncate differently and it is ugly when it happens.
- Depth check: pick ten random SKUs from the master list and confirm each is reachable in three taps or fewer, or is covered by the catalog link.
- Cap check: confirm no collection wired to a multi-product message exceeds 30 items, including after your next stock upload.
- Out-of-stock behaviour: mark one item unavailable and verify the assembled message skips it cleanly.
- Handover: send a message with a model number and confirm a human gets it with the full thread, inside working hours and outside them.
- Cost per enquiry: pull a week of message data and count how many paid template messages an average catalog enquiry triggers. Free-form navigation inside the service window is free under Meta's per-message pricing, in effect since July 1, 2025; if enquiries are triggering paid templates, the tree is too deep.
The signal that the build is right is boring: fewer screenshots in your team's chat history, and enquiries that arrive already narrowed to a category. In the scene this piece opened with, the fix looks like the coordinator sending eight brake pad cards in one tap instead of eleven cropped screenshots, and the three phones on her desk going back to being one.
Quick answers
How many products can you send in one WhatsApp message?
A WhatsApp multi-product message can contain a maximum of 30 products. In practice we keep it between six and twelve, because long card lists get abandoned on mobile.
Can a WhatsApp Business Account have more than one catalog?
No. WhatsApp allows only one catalog per business account, so segmentation happens through collections inside that single catalog.
How do you share a WhatsApp catalog with 500 products?
Split the 500 products into collections in Meta Commerce Manager, guide customers to the right collection with an interactive list menu, then send a multi-product message with only that collection's items. Send the full catalog link for exploratory browsing.
Do you need an approved template to send a catalog message?
You need an approved template only for business-initiated messages sent outside the 24-hour customer service window. If the customer messaged you first, you can send catalog and multi-product messages free-form within that window.
Start this week by pulling a month of WhatsApp enquiries and building your first collections around the categories customers actually ask for, none larger than 30 products. If you want the whole flow designed and wired for you, collections, menus, handover points and stock sync included, that is exactly what Learnmind installs.
