PIM is boring. It’s what you do with it that’s fun!

Two years ago I was fortunate enough to be asked to speak at Akeneo: The Product Experience Company 's Unlock 2022 event in Paris. As a customer, I’d been responsible for implementing one of the largest instances of Akeneo globally, and in doing so we’d done some pretty interesting (and unique) things with how we were using PIM. I was asked to share some of these insights.

On stage, I spoke about how we’d developed Order App before we rolled out PIM to help with change management, how we’d built a customisation to allow us to create parent and child products from scratch within Akeneo to speed up product creation, and about the foundational decisions around multichannel / localised publishing that would later lead us to build Thundercat.

And I also said that PIM was boring.

You said what?!

Given the event was focused on shifting the conversation from PIM to PXM, my comment wasn’t overly controversial, but it was still an interesting one to make on stage in front of the C-suite and co-founders of a PIM SaaS vendor:

[With PIM] the information management part of it is the dull part of the equation… what you do with that data is the fun bit and where the benefits come from

So what did I mean by it?

In short, I meant that the effort to manage product information isn’t by itself that exciting. It’s more of a necessary foundation than an end in itself. If you want to see the benefits of PIM (and have some fun along the way) then you need to use that information to do something.

(I should add at this point that I do actually find the process of developing product data models and configuring PIM within a tech ecosystem to be fun, but that’s the problem solving / build part not the ongoing information management)

Having fun with product information (and seeing the benefits)

When I made my comment, I already had a shortlist of interesting things to do with product information. Given that was early 2022, the advances in AI over the past two years fuelled yet more potential options.

I’ve highlighted 10 of my favourite opportunities below - each of which I’ve helped to build, implement or research at some point. I’ve also included a few tech solutions that I know of where I think they’re relevant - though its worth considering that each business is different, so the ideal options will vary also.

1. Tag attributes with AI

Whether through a 3rd party SaaS provider like Vue.ai , Pixyle.ai or Dressipi , a multi-modal AI agent or a self-trained model - using visual tagging can speed up the tagging of attributes to products, whilst doing so reliably and consistently to a high level of completion.

Also, put simply - better attribution leads to better results of the other nine ideas that follow below.

2. Categorise for discovery

With a large catalogue of products, or even if it isn’t that big - using rules or machine learning to categorise products into relevant merchandising categories is an instant win for product data.

If you’ve got a set of criteria in mind, tools like DecisionRules.io can simplify the setup of the rules engine with an API that you can call from within your tech stack.

3. Generate descriptions

This list wouldn’t be complete without mentioning something generative. Tools to assist with product descriptions have been around for a while, though with the advances in GenAI in the last two years, they’ve come on even further.

There’s too many options to mention them all, but Textual.ai has some interesting configuration features. It’s also worth noting that Akeneo now includes generative descriptions within its latest version, so you may not even need to look outside your PIM.

4. Automate publishing

If you run a global e-commerce business, you’ll inevitably have products that are limited to certain markets or channels, whether temporarily for a special launch, or permanently because of trading restrictions.

As long as the information to determine these is held in PIM, then you can use it to help with automating and restricting publishing as well.

5. Identify trends

Product information doesn’t just help with getting products in front of customers - it also helps understand what’s happening once those sales start coming in.

Patchy or overly precise attribution makes spotting trends difficult, but with reliable data and the right analytics tools you can easily see what’s hot and what’s not within your assortment - and the drivers for this.

6. Dynamically price

Similar to trends, if you know what your products are - and can predict when they’re due to go out of style or season, you can use this to help price them dynamically to maximise your margins. Companies like Sparkbox.ai and Peak offer solutions in this space.

7. Plan ranges and assortments

Whatever the tools of choice, knowing the current mix of products and how they’re performing allows for better planning decisions on ranges and assortments. Richer product attributes provides the data for this planning to go deeper than just categories, and gives you a common language with which to communicate this with buyers before orders are placed.

8. Suggest buy ratios

Different versions of products sell at different rates, to different consumers, in different locations. As an example, in clothing the sales of the various colours and sizes of a product sell at markedly different rates, and often this is about more than just the broad category of product.

With the right tools (such as Order App) to hand, predictive analytics can be applied to products based on their attributes at the point of ordering, allowing these ratios to be suggested to buyers and helping to reduce markdown.

9. Tailor content

One of the core benefits of PIM is having centralised product information and a single source of truth. But a single truth doesn’t mean showing the same information to every customer, irrespective of channel.

Tailoring content allows product experiences to be configured in the right way for each of your sales channels. You might even consider generating different descriptions for different channels, managing detail and tone of voice to cater for different segments of your customer base.

10. Recommend products

The simplest product recommendations work on a 1st order, product-to-product basis. Essentially “if you buy this product, we’ll recommend you that one”. These rules can be built manually or dynamically - but if products are low in volume, then this both reduces the available data to learn from and increases the chance of stock-outs for recommended items.

Learning about 2nd order relationships is therefore key, matching “products like this” to “products like that”. To do this requires the algorithms to understand more about each product and to build the relationships between these attribute values, rather than the products themselves.

Focus on what you’re going to use the data for

Whatever you take from the ideas above, what’s hopefully clear is that beyond getting data into PIM, the real fun and benefit comes from what comes next.

It’s also worth saying that the same thought should go into the business case that justifies PIM in the first place. By focusing on what you’re going to use the data for, you can evaluate the benefits that PIM will help to bring.

And when you do that, PIM becomes a bit less boring.