A Fresh Take on the Pinterest Smart Feed Algorithm

How To's, Tips

The new Pinterest Smart Feed algorithm is unlike anything the internet has seen before. It takes several complex search processes based off of user signals and interests and combines them to create the foremost visual discovery engine to date.

Ultimately, the algorithm determines quality, rates pins, assigns interests, and places visual content in front of consumers that are actually looking for it. The Smart Feed is fueled by Pinterest’s Taste Graph and how it assigns interests based off of quality and relevancy of every pin on the site.

To better understand how to market your pins and create a reliable process to suit the new Smart Feed algorithm, we’ll look at the process in reverse and explain what happens to a pin once it’s published on the site. This rearward take on a forward-looking process is refreshingly helpful for those looking to master the algorithm and create powerfully effective pins this year.

How do these new search processes change the way you should market your pins?

The new algorithm, which is the way Pinterest rates pin quality and sorts it under categories of interest, relies heavily on visual content and the textual data found underneath the image of the pin.

Your goal should be to make every pin as relevant as possible for it to be scored and sorted into an interest category and placed on the Taste Graph of consumers that benefit from the pin. The better you get at doing this, the more often your pins will appear in the Smart Feed of consumers that find it relevant and interesting.

Pinterest uses these pin histories to tailor relevant results that accurately predict what someone wants to see. – Allison Lips, Social Media Week

The better you are at predicting what your exact consumer wants, the higher chance you’ll have at structuring your pin in a way that gets it into the Smart Feed of the people willing to act on, or click, your visual content.

Effective Engagement with Interest Targeting

To market effectively you need to understand how Pinterest assigns interests to visual content. With this understanding, you can cater pins toward the specific interests of your target consumer.

All pins are extracted, normalized, and scored before being assigned an interest. If you want to cater your pins to a specific consumer you need understand what ties their personal interests to your product or service.

If you can approach your Pinterest marketing with this mindset you will be able to create pins that are hyper-focused on one specific type of consumer, which is exactly what the new Pinterest Smart Feed wants.

Here are a few tips on making effective pins that are precise and inclusive.

  • Improving precision (don’t guess what interests them, know)
  • Expanding interests per-Pin (understand how one interest can be enhanced by another)
  • Expanding language coverage
  • Improving space efficiency

According to Pinterest, they assign on average ~8 interests per-pin, per-language, with significantly higher coverage for more popular pins. The more you can improve the way your pin is scored and sorted into categories of interest, the more successful the pin will be.

Allow the Taste Graph to Guide Your Strategy

By incorporating the Taste Graph technology into our interest targeting, we’ve expanded the number of interests an advertiser can target to more than 5,000, a 10x increase. In early testing, many advertisers saw CTR increases over 50 percent and drops in CPCs over 20 percent. – Brian Johnson, Pinterest Head of Knowledge Engineering

The Taste Graph is made up of 3 components that work together to assign interests to the pins you post. Over 1 trillion data points have been collected to power Pinterest’s taste graph, and with over 100 billion pins and 200 million active monthly users, the Taste Graph expands and strengthens daily.

Let’s take a look at how this information is collected and what it means for marketing your pins.

Extract

With every pin published you need to ensure that it meets the requirements of having a rich source of textual data. The more specific you can get with your pin’s associated text, the better.

And by specific, I mean knowing your target consumer (a single consumer not a targeted audience) and exactly what they will type in the search bar to find their desired result.

That means you really need to intimately understand the specific interests, or taste, of your consumer if you intend to have your pin pop up on their Smart Feed where it belongs.

Pinterest lists the following as the textual aspects of a pin they assess to classify it:

  • Title
  • Description
  • Link text
  • Board name
  • Link alt text
  • Image caption
  • Page title
  • Page meta title
  • Page meta description
  • Page meta keywords

It’s important to note that pins with the same image all fall under one category (something called a PinJoin) and then all of the pins are textually-dissected at once to determine which of them are truly relevant, quality pins.

Normalize

At this point, Pinterest takes the pins and normalizes them by doing this nifty thing called lemmatizing.

A real tongue-twister, I know.

Lemmatizing is much less complex than you’d think, it is the process of grouping words by their inflected or variant forms. So, “car,” “cars,” “car’s” and “cars’” all lemmatize to the word car.

The reason Pinterest normalizes pins in this way is to make them easier to score based off of high quality, frequent text, and to discard random text and blacklisted text such as spam.

Please keep in mind that this process of normalizing is a very bland description of a system that contains layers and layers of processes underneath the hood.

By understanding people’s evolving tastes, preferences and interests, the Pinterest Taste Graph connects the millions of people on Pinterest to hundreds of billions of fresh ideas that are just right for them. – John Milinovich, Product Manager at Pinterest

Score

Now we’ve reached the scoring process where the algorithm decides which interests to assign to which pins. Here is a sample of how Pinterest assesses and scores a pin’s interest at this point in the process.

  • Word embedding (cohesive)
  • Frequency counts from Pin, board and link texts (popularity)
  • Normalized TF-IDF scores (uniqueness)
  • Category affinities–do interests belong to the same or similar categories (cohesive)
  • Position within text (importance)
  • Whitelisting (wikipedia titles, vertical glossaries/taxonomies, entity dictionaries)
  • Graph queries (cohesive)
  • Pluralization (normalization)
  • Blacklisting
  • Head queries (importance)

Ultimately, the interest that’s assigned to your pin will dictate where it falls in the Pinterest Taste Graph and the consumers who will see it in their Smart Feed.

Yes, it’s a complicated process, but it’s designed to direct relevant content toward interested consumers. And we all know that an interested consumer is more compelled to buy or act on a pin if it’s truly relevant to them.

Now that you understand the importance of marketing your pins with proper textual data you’ll be able to create effective visual content that the Smart Feed algorithm can empower and place in front of the exact consumer you want it to.

Don’t dread the slight changes that come with the improved algorithm and Pinterest Smart Feed. Instead, view the changes as tools to gain more traction on one of the fastest growing visual search platforms on the internet.

After all, 40 percent of all consumers rely on a visual piece of content to make a purchase decision. The better you can be at formatting your pins to suit the algorithm, the higher chance those pins will have at getting in front of your desired target consumer.

You can always create a Woobox account to practice creating textually rich campaigns that are directed at the interests of a single consumer. Creating the account and test campaigns is free so you can get comfortable with the process without spending money before embarking on the creation of an effective visual content marketing campaign for the year ahead.

Tagged with: , , , , , , , , , ,