Another Way of Thinking about Keywords: Semantic vs Syntactic

When we think about keywords, we’re often focused on how they match to search terms in a process-oriented approach.

For instance, you might think about broad match as showing for a lot of potential queries, some good and some bad, and that you’ll have to manage the search queries and negatives closely when using that match type. Or you might think that exact match is going to show for exactly what the user typed in, but you won’t have a lot of impressions for that term compared to the phrase match variation.

We often associate the word match types with control.

Control of exactly how your ads will show is useful in a world where search queries are repeatable and predictable; however, we’re starting to enter a world of widely varying queries.

What’s changing?

Voice search.

As voice search continues to rise, the sheer volume of queries will also rise. We see some people ask 10+  word questions to their phone and others who still just state the search query they would have typed into a laptop.

The other change that is happening is a contextual change. Consider this scenario:

I have a song stuck in my head. I know some of the lyrics, but the rest are eluding me. I just want to remember who sings it so I can get it out of my head.

Now, I can search for “Who sings this song with these random words in it: <lyrics>?” The search engine could give me two possible answers:

1. My entire rambling search query, in an attempt to  match the documents to my words.
2. The answer. This song is sung by <artist> and here are the lyrics and how you can buy the song.

The first result is syntactic. The search engine takes what I say and tries to match that query to documents. The second is semantic. The search engine understands my question, and instead of just repeating it, shows me the answer.

Now, we can expand this to match types.

Syntactic matching is matching search queries to keywords based upon the actual words the searcher typed into the engine. This would be exact and phrase match.

Semantic matching is matching search queries to keywords based upon the intent of what the searcher typed into the engine. This is broad match.

Modified broad sits somewhere between the two based upon how many +signs you use in the keyword.

When we think of the buying funnel, we often move through a few stages:

• Awareness: We want a wide range of matching coverage, so let’s show for contextually related information. This is semantic match.
• Interest: We want to show for queries that are closely related to someone in the consideration phase of buying. This is a mix of semantic and syntactic search.
• Purchase:  We know these products and services. The queries are specific. Let’s show for syntactic based matches.

When you consider where you are trying to attract users throughout the buying funnel, you might also consider how different match types can help reach those objectives. If you need more information on match types, please see this article.

As voice search changes how people search, and the search engines start to move beyond just the actual words a user types into an engine and start to determine what they really want to find, we’ll see broad and modified broad grow in usefulness.

While broad match can still be too broad in many cases, its uses and usefulness are growing.

Over the coming years, we’re going to see machine learning get better at understanding intent. We’re going to see voice search continue to increase as a percentage of query volume. And eventually, we’ll see a world where broad match understands the semantics of a user’s intent and starts to become a great match type once again.