Here at LSB, we use data — including Instagram data — to create ideas that change businesses. It’s that simple (and that hard, when you roll up your sleeves and get into it). For instance, by extracting Instagram data from images, we find insights that reveal not only what consumers think of brands, but also how they use products.
In my last post, I covered how we’re leveraging machine learning and other innovative statistical techniques to extract insights from Twitter and Facebook. Those methods are primarily focused on mining text-based data (which, it should be noted, may be a little tougher to come by in the future). Now it’s time to up the ante by extracting image data.
For this post, we’re focusing on one of the most popular image sharing sites: Instagram.
Digging for data via image recognition
As Instagram continues to grow at an astronomical rate, harnessing the true power of Instagram data is more important than ever. There are many social listening platforms out there that have historically revealed surface level details about your brand’s Instagram content, like geo, filter and hashtag usage. But that kind of Instagram data is just the tip of the iceberg, and, as I noted above, some of that data is starting to be restricted as a result of backlash against data practices by Facebook.
Now, there is incredibly valuable data on Instagram in the images people post, not just in the meta data surrounding the images. Until recently though, it’s been nearly impossible to gather meaningful insights from Instagram images in bulk.
Previously, tasks like image recognition were reserved for those with access to super computers, but now, as long as you have the technical know-how, even lean operations can use innovative technologies to unlock the potential data hidden in images.
Here are three hacks you can use to unlock the data hidden inside Instagram’s photos:
1. Find out more about your category
Here in Wisconsin, cheese is big business. In 2016, the Dairy State produced over 3 billion pounds of cheese.
As much as we love to grab a chunk of cheddar and nibble on it, we’ve found that for our clients the best way to drive food sales is to show consumers how your products can be used in everyday life. To determine which dishes you should feature in advertisements, you should know which foods people are already consuming.
To demonstrate this, we collected images of food from Instagram with #WisconsinCheese. We then used image recognition to determine the specific foods people are using (and posting) Wisconsin cheese in. Here’s what we found:
Even though we consider ourselves Wisconsin cheese experts, there were some surprising takeaways:
- No surprise, deep-fried cheese curds were the #1 posted food.
- Don’t skip the drinks! Beer and other beverages were frequently included in Instagram posts featuring Wisconsin Cheese.
- Some of the first things that come to mind when we think about cheese are cheese and sausage platters, cheeseburgers and ham sandwiches, but we tend to overlook that vegetarians are also cheese heads. Not only is vegetarian food the #2 most frequent post, but also specific non-meat foods like noodles made an appearance on the list.›
- Cause mac and cheese is awesome (and a noodle). YUM.
- Cheese: good at all times of the day. Breakfast sandwiches, brunch and fruit also made the list.
That takeaway is that if you’re trying to sell a product to people, it’s important to find out how people already use it.
A great question to think about is: how does our product fit into existing habits? Some of the items on the list might not be top of mind, but they would nonetheless be advisable to include in future advertising efforts.
Although we applied image recognition technology to food images in this example, this method can be used to identify virtually any object.
For example, if you’re a camping supplies company, it might be helpful to know how many people share pictures of their dogs while they’re camping. If there are enough people including their dogs, you may consider promoting dog friendly products on Instagram.
Or, maybe you’re a furniture supply company. You could use these algorithms to see if more people post about their newly-renovated bedrooms, kitchens or bathrooms versus completely new construction.
2. Find out how influencers are promoting your products
Influencer marketing is one of the hottest marketing trends, but how can you mine those influencers and the content they produce to uncover the latest trends in your category?
That’s right, by using those images from Instagram. A great example is in interior design, where arguably the most important trend to keep up on is which colors are going to be “in” next.
To demonstrate, we pulled 1,000 images of rooms from interior design Instagram influencers like Alyssa Kaptio, Ryan Korban and Becki Owens. Then, we computationally extracted the color palettes they used in those rooms.
In the image below, every individual 3×3 box is the color palette of an Instagram photo from an influencer.
Can you find a color palette that you like?
A quick glance reveals neutral colors (whites, grays, browns) are used heavily. We can gain additional insights from this data by seeing exactly which colors are being used most often.
If you’re a furniture manufacturer or interior designer, you can use the data you mine from Instagram and color identification to get ahead of the next big trend, or give your customers a look that feels like it’s directly from an influencer’s Instagram.
3. Landmark detection
Objects and colors aren’t the only ways to leverage Instagram and understand what users and influencers are posting.
Here at LSB, we’ve developed an expertise in helping cities promote themselves in areas like tourism and conventions. And often, cities miss opportunities to promote their most prominent landmarks and destinations because they don’t (or can’t) track Instagram.
Take Chicago for example. We extracted Instagram posts using the hashtags: #DowntownChicago, #VisitChicago and #ExploreChicago. We then ran them through one of our algorithms that identifies landmarks and locations within the images. Here are the places that were seen most frequently on Instagram:
- Cloud Gate (“The Bean”)
- Willis Tower (Sears Tower)
- John Hancock Center
- Chicago River
- Jay Pritzker Pavilion/ Millennium Park
- Rookery Building
- Wrigley Field
- Trump Hotel
- Chicago Loop
- Chicago Theatre (Balaban and Katz Chicago Theatre)
- Harold Washington Library Center
- Flamingo (Alexander Calder statue)
- Civic Opera House
- Chicago Cultural Center
- Tribune Tower
- Chicago O’Hare International Airport
- Nichols Bridgeway
- Marshall Field and Company Building
- Marina City
- LaSalle Street Station
- The Field Museum
- The Art Institute of Chicago
As one might expect, Cloud Gate is the most visible tourist attraction in Chicago. It’s unique, conveniently located and is featured extensively in promotional material for the city. Other buildings and locations on the list may seem obvious, like Willis Tower, the Wrigley Building and the Chicago River, but many of the other buildings are surprising.
A quick comparison to the locations shown in Chicago tourism advertisements reveals they are missing some of their most Instagrammed landmarks like LaSalle Street Station and the Civic Opera House. These locations are significant and appealing enough for people to post them to their personal Instagram, so why not feature them more prominently when promoting Chicago tourism?
Conclusion: Approach Instagram data mining strategically
Insights from Instagram photos can be difficult to acquire since you have to know how to create your own data pipelines and analytical tools. But with some fine-tuned algorithms and cloud-based APIs you can find a treasure trove of data within Instagram by extracting image data.
Begin this type of research with clear questions in mind. Especially because Instagram data actually found within photos can be computationally strenuous and difficult to attain, you don’t want to waste time exploring your data.
Ask specific questions like:
- What are the complimentary products that people are posting with my product?
- What are the trends that matter to my business?
- How do I go about identifying them?
As Instagram continues to grow, brands that leverage insights from the platform are going to have a leg up in the market. Will your brand be one of them?
And hey, in case it’s not obvious, we can help with this stuff. Reach out and let’s talk.