Miravia Marketplaces: Pioneers and geniuses or serious mistake implementing AI?

Miravia is one of the main Marketplaces in Spain. Without putting data on the table, because it is not the purpose of this post, they will be within the TOP 15 in Spain in the few months they have been operating - They launched in late 2022.

They belong to the giant Alibaba and have a great economic muscle behind them, but... are they being pioneers with Artificial Intelligence applied to sellers' catalogs or are they making a mistake?

Miravia: First steps with AI

This summer, Miravia launched its first two content optimization options with artificial intelligence:

• Title optimization
• Creation of new attributes

Since I am pro-IA I looked favorably on this launch, although I had certain reservations from the beginning because, although everything always looks pretty at first glance, when you dig a little, something less pretty may appear.

Title Optimization

All of us professionals who are in the Marketplace sector know, thanks to the fights we have with Amazon, that creating good titles is of vital importance for the success of a brand.

So, a priori, the existence of an internal AI tool in the Marketplace that can help you generate better titles seems to be a step in the right direction.

And even more so when we know, for example, that many of the product titles that are sold via Amazon Vendor, the platform creates them automatically based on the different fields of the product upload Excel. Endless discussions with Amazon's AVS about this I have had with the La Liga Profesional de Fútbol project on Amazon Europe.

The problem is that it's all more complex than it sounds. So I will highlight several positive and negative points of the AI in Miravia:

• Being able to optimize hundreds of titles in 1 minute speeds up the time to market which, if done manually, would take months.
• The quality of the titles generated by AI depends a lot on the quality of the information given to the AI to generate them. So, if you have bad information in Miravia, you're never going to get good quality titles.
• There are a multitude of small factors to take into account that, if not taken into account, will generate poor quality titles.

Let's look at some examples of those "small" factors that AI must take into account:

1. Harmonize the writing style of metrics: gr or gram or Gram or Gram.
2. When to use uppercase or lowercase letters.
3. Identify when a given word is a brand name, so that it does not translate it. "Nike = Nike" "Key Health = Key health?"
4. etc.

Another important point is that each product subcategory needs to highlight different features in the title, description and bullets points so you must create pre-structures for the subcategories to get good information.

Let's take a look at some examples of creating appropriate title pre-structures for the sub-categories of coffee machine parts and food supplements:

• Brand | Coffee Maker Refill Type | Compatible Model | Material | Important Features | Most Relevant Keywords.
• Brand | Nutritional Supplement Type | Main Ingredients | Recommended Serving Size | Key Benefits | Most Relevant Key Words

These are some of the points to be taken into account for catalog creation using artificial intelligence, but the reality is that there may be more than 50 features to take into account.

What good are titles without keywords?

That's right, what good are they? Little use. If the AI is not integrated with the keyword database, in this case Miravia's, and you can't know which keywords are best for your product, you won't get the impact you want either.

Let's go into the details about AI-generated content in Miravia in some examples:

• The AI does not identify correctly the brand.
• Suggestions are near carbon copy of the original title - This is a fairly common mistake the AI makes, it recreates the information you give it instead of creating something original. This has to do with the Temperature you can give it and the prompts.
• The 2 titles that Miravia recommend you have radially different approach, but without having access to or the certainty of how much the terms are searched for, which title will be more effective? Which will convert more?
• Personally I like a space between the number and the measurement: 500 ml, not 500ml. One you can change by hand, but when you use the tool and optimize 200 products, you can't change by hand 50 or 60 products where it appears: "500 ml", "500ml", "500 milli", "500 milliliters".

Creation of new attributes or selling points

Here I will be more brief, I think this development is a good idea. I would simply take advantage of this development so that the attributes that are created are highly related to the existing filters for search in the platform.

Why? Because for Marketplaces where the keyword search engine is not yet super powerful, search filters are of great help to the user.

So the more attributes match with filter elements, the better for your product.

The new implementation: AI descriptions

Less than a month ago, Miravia launched the creation of descriptions through Artificial Intelligence, another step forward in its commitment to AI.

Look at the magic of this technology. Thanks to Miravia's technology you can optimize bulk descriptions for 1,000 products in 5 minutes.

But these descriptions have the same problem as titles:

- Low quality content.

- Problems to identify the brand name.

- Mix of uppercase and lowercase letters.

- In summary" paragraphs with redundant information of little value, etc.

- Loading... Do you see loading in the right corner of the next image? Is because the system has been stuck like this since launching more than a month ago. To see the AI Description you have to apply the changes and check the changes in the product page later on.

Conclusions

I hope you have come this far to clarify one important thing: "I am not picking on or criticizing Miravia in a bad way", I am trying to educate about the use of Artificial Intelligence for Catalog generation in Marketplaces or for eCommerce.

I'm trying to bring the hype down to earth so that many of you entrepreneurs in the eCommerce world can get the most out of this new technology.

Why is Miravia doing well?

Being a pioneer in something always involves taking risks that can either pay off in the long run or can be a flop. Miravia has taken a risk and has been a pioneer in bringing artificial intelligence to the world of Marketplaces. Why has it done so?

This is my opinion:

For those of you who don't know, Marketplaces, especially in their early days, need millions of products for users to find everything.

That's why, in their first years, they need big distributors and wholesalers to sell on their platforms. In this way, a single company of this style can offer for sale large catalogs of 10,000 or 30,000 products or more.

With this, what the Marketplaces achieve is to put as many products on sale as possible in the shortest possible time with fewer resources: fewer salespeople looking for companies to sell, fewer account managers dedicated to the management and support of the sellers, fewer accounts opened, etc.

The problem is that distributors and wholesalers have very poor quality catalogs in terms of product information. And very bad catalogs mean bad user experience, lower conversion, etc.

That is why what Miravia has done seemed and still seems to me to be a very good idea. Besides, the amount of data and conclusions they are obtaining will be brutal.

Last comments

• If you are a distributor, start using Miravia's AI tools for catalog.
• If you are a brand, don't use it, your listings need to have a very high quality. Better use a platform like Carpio where you can upload all the product information you want, technical information, commercial information, keywords, etc., and that will help you to generate titles and descriptions as you really want.
• With the help of professionals in this field, Miravia will quickly improve this system and if you need my help, I will be happy to help you.

Miravia is going to be very successful with AI. This technology is advancing disturbingly fast and although to us it may only seem to be Titles, Attributes and Descriptions, to Miravia it is user retention, conversions, increased average time spent on the website, etc.