When it comes to social media analytics, things could be not what they appear to be. And Instagram Insights is not an exception. In this article, we'll share with you some interesting insights about Instagram and HypeAuditor data sources and compare them to understand their analytics better and make the right decisions.
Almost all data in Instagram Insights and HypeAuditor are estimated. They aren't straight count — metrics are derived through statistical sampling or modeling. We both look at a portion of the data that is a good representation of the larger population included in the entire set of data.
Sampling allows calculating metrics quickly with a high level of accuracy.
Let's have a closer look at Instagram and HypeAuditor differences.
Sources of a data
HypeAuditor and Instagram have completely different sources of data. Instagram analyzes a sample of an influencer's followers who were active on Instagram over the last seven days.
That means that if an influencer bought a lot of bot followers, who are not active at the moment, they'd disappear from the Insights 7 days after they became inactive. Usually, bot becomes inactive as soon as it reached 7500 followings.
To get audience insights, HypeAuditor analyzes a sample of an influencer's audience who engage with the creator's content. The size of this sample is chosen by the machine learning algorithm and depends on a variety of factors, but this amount is always statistically significant.
If an influencer does not use any inauthentic methods to grow their number of followers and engagement, the HypeAuditor’s data is fairly the same as Instagram Insights. If an influencer uses some questionable mechanics, the number of errors might be higher.
If we suspect that the audience is not authentic, and a creator uses suspicious tactics of Instagram Growth, we'll make a notice about it on the report. In such a case, you'll see the following note: Audience seems inauthentic or low-engaged. We do not advise to rely on country/city/language data for influencers with an inauthentic audience.
To detect the audience's country and city, Instagram uses a sample of an influencer’s followers who have been using Instagram during the last 7 days. The location data comes from Location Access, a setting on a mobile device. It includes IP address or Wi-Fi connection and specific location information like the device's GPS signal.
Be aware that bot accounts are registered automatically using scripts. Thus these accounts do not have location information, or it's hidden using a web proxy. An influencer could easily buy bot followers with Europe or the US IPs to look more natural in Instagram Insights. It’s a bit more expensive than bots with Russian, Asian, or Latin American’s IPs, but it’s not hard to find.
HypeAuditor uses completely different methods. The machine learning algorithm chooses a sample of an influencer’s active audience and checks their account’s bio and recent geo-tags for location data.
If a creator uses only organic methods of growth, the difference between Instagram Insights and HypeAuditor is only 5%.
Audience age on Instagram Insights is also an estimated metric that calculated based on audience activity from the last 7 days. Unfortunately, Instagram does not provide their sources of data, but we suppose that they rely on info that a user provide themself.
At HypeAuditor, we use proprietary image recognition technology to detect the audience's age. The algorithm is trained on an extensive database of different human faces.
The technology tracks multiple facial features and gives an estimate of the age of each individual. In 97% of cases, the result is the same as in Instagram Insights.
According to the information in the Facebook Creator Studio, gender on Instagram Insights is an estimated metric, calculated based on audience activity from the last 7 days.
Source: Facebook Creator Studio
Instagram also doesn't provide their sources of data, but we assume that they also rely on information that a user provides. If a user didn't disclose their gender on Instagram, they might use data from Facebook, where users self-report their gender and select the preferred pronoun from "female - her," "male - him," and "neutral - them."
HypeAuditor detects gender by using natural language processing algorithms supported by an image recognition technology.
The average error rate of our algorithm is around 3% (it makes 3 errors out of 100 cases).
Whenever you make a decision, what data to use for your business, pay attention to the sources. The only thing you should care about, does this data help you to achieve your goals and make the right decision?
At HypeAuditor, we seek to assist advertisers by providing them data to make their marketing campaigns more relevant and effective. For influencers who do not use any inauthentic methods of Instagram growth, our data is very close to Instagram Insights. That’s why thousands of brands and agencies around the world rely on HypeAuditor data.