
Article · Choco
To the catalog & beyond…
Sub-catalogs at Choco were slow, manual and impossible to scale. A look at why, and at the smarter, semi automated catalog we could build instead.
Sub-catalogs are created and maintained manually by our operations team. The process and the setup are slow, they have their limitations, they are hard to maintain and they don't scale. Sub-catalogs are also not an industry standard: vendors don't know about them and can't use them the way we do.
These hypotheses come up again and again in every recording, every user research session and every Miro board. The purpose of this article is to articulate my thoughts on the topic and to give some perspective for the longer run.
So, what is our current process?
Operations uploads a CSV file in the Admin. Every product in a sub-catalog has to exist in the main catalog; a sub-catalog cannot contain anything the main catalog doesn't.
Vendor users have no way to manage their listings beyond changing product availability: available or not. They can technically edit the product name, ID, unit, price and category, but this is rarely used, and everything else sits with operations and the sub-catalogs.
What exactly is a sub-catalog?
In a nutshell, it is a group of products made for a single customer or a group of customers, for a permanent or temporary need. In one form or another we will always need this.
Temporary needs can be based on availability: a seasonal product, Halloween or Easter. Permanent ones could be branded products meant only for a specific group, or products that are restricted or simply not intended for another group.
In other words, we have a temporary or permanent set of properties, attached either to the product or group or to the customer or group, and those properties can be permissive or dismissive. Intended or restricted, forever or until Friday.
Say we want to hide certain products from a small group permanently, and at the same time apply seasonal changes to all customers with those segments in mind. Today that is not possible. It takes workarounds: copying or removing products in a CSV and uploading it to every customer individually. That is what sub-catalogs are for.
To go a level deeper: for all these product and user properties to work together in a meaningful way, and they often live in different dimensions, they need at least a basic hierarchy. Between equal rights, force decides. If two actions contradict each other, for example a branded ketchup is blocked for a certain user while the same product is on a seasonal promotion, should that user see it or not? Our setup can't even consider the question. The CSV upload has no conflict handling, not even for duplicates, and no other validation.
How does this show up on the buyer's side?
For technical reasons, buyers select products in bulk: they pick several products and the action only happens when they tap save. If one of those products has a duplicate ID, which this complex manual CSV process makes easy, the whole action fails. And you can imagine every other way wrong data can creep in.
That already raises questions. What happens when I remove a product from a sub-catalog? Is it still in the main catalog and visible to the buyer? What about removing or archiving products directly from the catalog?
We don't have a management tool on the web that vendors could use for their own customers, which would also reduce the workload on our side: duplicate, replicate, turn one setup into a template and apply it to another vendor's products or customers. Today all of it is done manually by operations in the Admin.
Will vendors use it, and what about integrations?
We need this regardless, for our own internal management, if we want to manage customers and products faster and smarter. And yes, it should work alongside integrations. Ideally we could take data and setups from our bigger vendors and replicate them for other users.
We can build the tool for internal operations first, then open it up to vendors and see if and how they use it. The only difference is scope: Choco has access to all vendors and their buyers, while a vendor user only sees their own buyers.
How can this scale?
Let's repeat the sentence from before:
We have a temporary or permanent set of properties, attached to the product or to the customer, that can be permissive or dismissive.
It is easy to imagine those properties being replicated and applied to other buyers, and even that alone would be a massive improvement on what we have. But for the sake of argument, can we go a few steps further?
Can we imagine a system where these permissions and decisions are semi or fully automated? Where properties and functions are mapped out with hierarchical rules and conflict avoidance? Where taxonomy exists not only for products, groups and categories, but also for actions? A dynamic distribution of large amounts of data between vendors, their products and their buyers.
Take a simple function like grouping unit variants, in two setups. In the first, humans do everything: the user selects a few products, taps "make variants", done. Humans alone decide whether a banana and a beer can be variants of the same product. It is a bit like a sub-catalog, just faster and shareable. Is that only error prevention? Where is the business benefit?
In the second setup, properties are attached to users and products, and to their attributes like unit variants or categories, and those properties talk to each other. Even at the most basic level there is a rule that produce and drinks can't be combined. With a proper taxonomy, products are grouped not only by primary category but by subcategory: drinks, alcohol, beer, lager, dark lager, Munich Dunkel. Maybe also by producer, location or packaging. Thousands of articles sorted instantly, without errors.
Speed and accuracy are nice, but the real power is in combining attributes that work together and can be replicated between users. Modular, like Lego bricks that fit, only dynamic.
Imagine price and conversion factors as properties of ordering and invoicing units, already defined in a vendor's ERP, pulled into our web tool and reused as a setup for another user. Add a seasonal discount as another property and have it calculated instantly. Follow the same path and the system could automatically include or exclude products based on user settings, or power product recommendations and upselling.
The next task at hand
We don't need to jump into the distant future. The relations between these product and user properties live in different dimensions and need a minimal, functional hierarchy. That is the next task at hand.
We can keep improving the current solution: better CSV uploads with conflict merging, ways to block sub-catalog workarounds. But the real question is whether we want to change an industry and reduce food waste without a smart, online tool to manage our own business. And why not start thinking about it now, before we have a bazillion products in our database and need to remodel everything?
In my opinion, this should be the long term vision for the catalog: a smart, semi automated system that is also reflected on the buyer side, which I have barely touched here.
When you build a rocket, you don't start with satellite navigation. You start with the engine and the fuel tank. But without a bigger vision, we will only replicate what others have built, instead of being a driving force into the unknown.
