Services Offered: Influencer Discovery
In early April of this year, developers for third-party Instagram tools all woke up to a surprise. Overnight, their software simply didn’t work and their customers weren’t happy about it. These bad mornings got even worse as the reason behind all the issues became clear: Instagram had made some major changes to its API, effectively choking off some of the data other companies needed as part of their software’s core functionality. This did not go unnoticed in the influencer marketing community, where an entire industry of software platforms exist to provide exactly the kinds of data Instagram isn’t sharing anymore.
It’s not clear whether this is a permanent change. The sudden deprecation of features was part of a previously announced plan to migrate to an entirely new API designed to better support businesses and their marketing. It’s also meant to ensure user privacy, so their data will remain locked down for the next year. “Basic permissioning”—whatever that means—won’t be supported until 2019. And while it’s possible for anyone to turn their profiles into business ones, freeing their data to flow freely once again, many influencers aren’t really thrilled about this option, fearing it might alter the perception their followers have of them. The bottom line is, if you’re a developer whose software relies on a steady stream of incoming Instagram data, now’s a good time to think about a pivot.
On the other hand, if you’re IG analytics provider Deep Social, you’ve got it all worked out already. It’s actually baked into the business plan.
Founded in 2017, the company has a bit of a history before that. The initial idea was to create another marketplace, with their big differentiator being a robust analytics components. It soon became clear they were able to things with data that no one else was doing. The whole effort to onboard influencers for their marketplace was essentially just holding them back.
So what is that they’re doing that’s so unique? To put it simply, they’ve built what can be described as “the Google of influencers.” A lot of companies like to use this comparison, but generally what they mean by that is they want to be the first place people go when they want to search for influencers—and so they’ve loaded their database up with millions of them to search through. Deep Social fits both those criteria, but they’ve taken it a step further by actually creating a search engine that ranks results by actual relevance (as opposed to keyword matches). We’ll get into how that works in practice down in The Details—for now what’s important is how they’re doing this when Instagram isn’t giving them all the data they need.
Deep Social’s got an AI driven predictive analysis engine parsing through all the data they’ve collected so far, which serves as the foundation of the software’s predictions of future trends. Of course, to be able to do this accurately, you need a lot of data—and Deep Social definitely has that. Before Instagram went full Wonka and shut its doors to the outside world, Deep Social spent a year feeding their AI system the billions of data points associated with Instagram’s 43 million most popular users—hashtags, usernames, images, captions, comments, all of it. From all this input, the AI is able to identify trends and forecast how those trends will continue over time. Because of this, when you look at the data for a given influencer on Deep Social, you’re not seeing a precise snapshot in real time. You’re seeing a couple of months into the future.
But wait: the accuracy of this kind of predictive analysis can degrade over time if it isn’t constantly fed new data, right? Right. And while Instagram shut down access to individual users’ data, many businesses that are working with the new API haven’t been. With this, Deep Social can grab that audience and use it to compare to their predictions and correct them, if need be. This consistent correction of their forecasting keeps things accurate.
Note that this explanation is a major simplification of what’s really going on under the hood, but it’s enough to get an idea how it all works. For a more detailed explanation, you can check out this post on Medium, published by the Deep Social Team. Or, you can just take our word for it: it’s impressive software that should see the company providing accurate data well into the future, even if Instagram doesn’t let developers back in to scrape non-business profiles.
For all the R&D that went into Deep Social, as well as the processing power that has to go into parsing the many billions of data points, you’d expect platform access to be out of the reach of small businesses. It looks, sounds, and feels like a big business tool, but the pricing puts it well into any business’s grasp. Access to the platform is entirely free, and customers pay for credits instead of a subscription fee. You will use one credit each time you perform one of the following actions:
- Influencer Analysis (1 credit per influencer) — This is the heart of Deep Social, where you’ll stats and rankings for the influencer, plus pages and pages of data on her audience.
- Influencers Identified Export (1 credit per 50 influencers discovered)— No credits needed to use the discovery tool, but if you want to export the list you’ll need to use a credit for every 50 influencers whose names and data you export.
- Post Analysis (1 credit per 2 Posts) — You can still get post analytics with Deep Social, but only if they’re sponsored posts. There’s some fine print about the true cost of this, because the amount is actually .5 credit for every thousand users that engaged with the post. For smaller businesses, this should cover it, but bigger companies should be aware that this could end up using many more credits than just the one listed on their pricing page.
The first five credits are free—anyone can sign up, login, and give Deep Social a test run. After that, the cost of each credit is determined by how many you buy at one time:
- Bundles of 10, 20, 50, 100, 200, 500, or 1,000 credits — 99 cents each credit
- 2,000 credits — $1,320 (66 cents each credit)
- 5,000 credits — $2,310 (46.2 cents each credit)
- 10,000 credits — $3,960 (39.6 cents each credit)
Deep Social serves two functions—discovery and analytics—both of which are served by the massive amount of data the company has mined over its lifetime. But isn’t just the number of data points that powers Deep Social. It’s the way the AI engine learns the ways they’re all connected and begins to “understand” that data in human ways. You can see what this means by doing a simple hashtag search on the platform.
You can see by the tag cloud on the right that a search for the hashtag “foodies” also returned results featuring other, similar hashtags that are just as relevant. Now, this on its own isn’t necessarily groundbreaking—it’s the fact the AI learns how these are all grouped together into a discrete topic, and its understanding of the topic ensures highly accurate search results. To fully appreciate this, you can visit this site Deep Social put together, a visualisation of all keywords and hashtags on Instagram organised by topic. Here’s what the whole thing looks like:
Each color is a different topic, and the map is interactive so you can “fly” into each one to see the tags found within, how frequently they’re used, and how they connect with others. Remember that the number of words in here is somewhere in the billions, and it only takes a few seconds to pull in the tag cloud and relevant results. Then also consider that this map only shows topics and hashtags; Deep Social also has usernames, locations, audience demographics (gender and age), brand affinities for influencer and audience, and post analytics. When you’re searching for new influencers to work with, the AI takes all these things into consideration to bring you the most relevant results. You can throw any number of criteria at it, weight each one, and watch as the list gets narrowed to meet your demands. When you find a potential influencer, you can then analyse their profile. This is where you’ll see exactly how much information Deep Social stores about each influencer.
Here’s one example:
This is a good example, because we don’t need to do any research to verify what we’re seeing (they’d all be from fake news sources, anyway). We all know this guy well enough to judge whether the information we’re getting is accurate. So, first things first: we see that he’s got over 8 million followers and about 143,000 engagements (that’s over a 30 day period). Those seem like legitimate numbers, though that audience credibility score seems a little high. You’d expect with such a high profile account that there would be a larger than normal number of bots and hangers-on. Well, good news: that credibility score is only based on the engaged followers—those 143,000+ accounts that liked or commented on posts. That’s the more important segment of the audience to pay attention to, anyway. And in this case, you can look at any comment thread in a Trump post and see it filled with real people getting really angry at each other, supporters and detractors alike.
So far, so good. Let’s dive deeper, though, and see whether Deep Social gets anything else right:
If you look at the list of categories showing the influencer’s interests, it’s a little odd for a U.S. president. The Television category seems dead-on, but Travel, Tourism, and Aviation? Well, there are loads of pictures of him getting off airplanes and helicopters. Restaurants and food? Presidents eat a lot of fancy dinners, and those are on full display. Toys, Children & Baby? Has Deep Social’s AI gotten confused? No, as it turns out, there are some pictures and videos of his grandchildren at play. And if you noticed that Politics & Government isn’t listed as an interest, you probably already realised the AI understands this profile just fine.
As for the audience interests, that’s harder to confirm, but Deep Social did pretty well understanding the influencer—it’s got understand the audience just as well. But what about those popular hashtags and mentions? Those are from the influencer, not the audience, and isn’t weird that Donald Trump @mentions himself so much? And the answer is “No, it isn’t.”
There’s plenty more audience data to go along with this, including demographics—age, genders, locations (down to the city), and ethnicity. Psychographics like brand affinities can also get very specific. But how do we know how accurate this stuff is? We can’t really prove the pyschographics, but the demographic data is a little easier to corroborate:
An audience that is two-thirds male and 84% white? That checks out.
Sure, we’re having a little fun with this but as an exercise in explaining the power behind Deep Social, it’s instructive. You can see both how much data you can get out of it, and the ways in which the AI can make inferential leaps to give you more information. You could make the argument that in this case, the inferences about Trump’s inferences were pretty far off the mark from what his actual content and persona communicate. But let’s be fair, his account isn’t representative of what an influencers would like. This account isn’t even operated by him; it’s just a PR feed devoid of personality. The only common theme among the posts is Trump himself; an influencer is usually more focused on the topics that made her influential in the first place.
With such a powerful and thoughtful backend, it’s surprising that Deep Social didn’t go a little further to add even the tiniest bit of management functionality. Yes, the software was designed to be a complementary solution to other platforms—Deep Social can help prop up other software where data isn’t really the focus, like an IRM or Campaign Management platform. Still, though, the absence of any way to organise the influencers you’ve found is noticeable.
There are plenty of ways you can manage the data elsewhere using Deep Social’s export functions, but not having the ability to categorise influencers into a list (say, organised by campaign), or even the ability to save a search and come back to it later is frustrating. Especially because this is the type of feature that’s easy to add—here’s hoping they add it in the future.
The next year is going to be a turbulent time for the influencer marketing industry. Instagram became the number one social platform for influencer marketing precisely because of its ability to engage users like no other. Right now, measuring that engagement isn’t as easy as it used to be.
While many developers are scrambling right now to get on board with Instagram’s new Graph API and figure out a way to maintain their services, Deep Social is a step ahead most other platforms. Whether it was intentional or not, the fact is the unprecedented amount of data they were able to mine last year set up them up for success in the wake of Instagram’s kind-of-a-surprise policy changes. The result is an analytics and discovery platform so powerful it erases any doubt that you’re finding the right influencers. There’s still the matter of a way to organise these influencers to make the process even more efficient, but we’d recommend you sign up for the free trial and take it for a spin to see how deep it can go.
Services Offered: Influencer Discovery