Google MUM: the coming SEO

SEO world never stops surprising us with algorithmic changes and new rules of the game.

Just when you think you are getting the hang of it, they do a 180-degree turn and what used to work stops working, or even becomes counterproductive for the positioning of your eCommerce.

Not all changes that are announced -or treacherously implemented- are equally far-reaching, but MUM threatens to redefine absolutely everything related to search.

Before explaining what MUM is and how it works, we should analyze how it got here.

If you know a little about how Google works, then you know that major changes are preceded by a sequence of previous changes that often seem minor or unconnected. Yet, everything is related and sets a direction.

In the case of MUM there is a predecessor that also has a short name: BERT.

In 2019 an update to the core algorithm that was to make the search engine more efficient through advanced use of Artificial Intelligence was announced.

The work that Google had been doing for years with so-called open source neural networks was put on the table.

In a nutshell: thanks to its technology, the search engine can now interpret language naturally and deliver more relevant results by "understanding" the context and not just the semantic framework.

This is a bit complicated to synthesize like this, but we will now go a bit deeper into it when we talk about MUM.

Once again, a friendly acronym for a complex concept. It stands for Multitask Unified Model.

As the name does not make much clear by itself, let's get down to the nitty-gritty.

It is a system that allows the search engine to answer questions using more resources, being more precise and conversational.

To give that answer, MUM will be able to use text, images and even videos. It will analyze all the content it has in its database, the context of the search, and it will not be limited even by language, since it is capable of doing so in 74 different languages.

That said, it may seem like an evolved version of BERT, and there is something of this.  But there is an important difference: although it partially shares its philosophy and framework, its multitasking capability makes it 1,000 times more efficient.

If BERT understands the language of humans, MUM is able to encode it and create a complex and natural interaction: it speaks to us using our way of doing it and with the authority of an expert.

By the way: speaking of authority: it is worth bearing in mind another of the historical workhorses that have been at the center of some of the latest updates. We are talking about EAT, an acronym for Experience, Authority and Trust and the driving force behind much of what has happened since August 2018. It is all spun out of Google.

Normally the best way to explain these issues is an example. In fact, we are going to use the same one that the search engine managers used during their presentation last May.

Imagine that you have a great love for climbing. You have recently climbed Mount Adams and you plan to climb Mount Fuji in the coming months.

You have previous experience but, logically, they are two different environments, different latitudes and even the documentation you can find will be in several languages.

Today, you will have to organize everything you want to know and perform multiple searches that will lead you to different pages and resources.

MUM aims to synthesize all of that into a single source, the one with the most authority and the one that solves the most related questions. In other words, we will be able to say to the search engine, "I have climbed Mt. Adams and now I want to climb Fuji next fall."

With this information, the search engine will be able to tell us how we should prepare physically, what kind of equipment we should bring, if we should ask for any kind of permission... and all this without being limited by the language of the search.

Another interesting example they offered is that of image search. You take a picture of your boots and ask the search engine directly if they are suitable for the climb you are about to undertake.

Thanks to technology like Google Lens, it is not only able to identify that the image corresponds to a boot, it can also infer what type of boot it is, characteristics, brand and model. It then analyzes everything and gives you an informed answer (can you imagine the potential impact for eCommerce?).

Google MUM has not yet been implemented, nor is there a definite date, but everything points to 2022 being its year.