NASA Predictive Analytics

@nasawebb: Final Report

Typology:

In order to assess what type of instagram posts are the most engaging for the audience we created a typology. This means that we arranged the posts depending on the type of content. Below are our seven typologies with definitions and examples:

1. NASA employee: These posts feature shoutouts of NASA scientists and employees. These are distinguished from lab technician posts by a more personal feature on the person's biography and contributions. Below is an example of a photo that would fall under our “NASA employee” typology.

2. Picture of space: Posts featuring space, not including artwork including space. Below is an example of a photo that would fall under our “picture of space” typology:

3. Picture of telescope: These posts feature an image of the telescope in the lab environment as well as images of its specific components. Below is an example of a photo that would fall under our “picture of telescope” typology:

4. Featured Artwork: These posts feature artwork submitted by users as well as art by more official artists. The primary source of this art comes from the hashtag #JWSTArt. Below is an example of a photo that would fall under our “featured artwork” typology:

5. Community based: Community-based posts include those centered around events. These seem to specifically focus on community outreach and education. Non-event posts focus on highlighting kids, young adults and remarkable figures in the broader community. Below is an example of a photo that would fall under our “community based” typology:

6. General science: General science posts often take the form of quick facts about both the Webb's capability but also about general space topics. Below is an example of a photo that would fall under our “general science” typology:

7. Lab technician: These posts feature lab environments and shoutouts of specific lab technicians as they work on the Webb satellite. Below is an example of a post that would fall under our “lab technician” typology:

Below are the top 5 posts, in order, based on our custom engagement metric (likes+comments/account followers).

Top 25 Posts:

Of the top 25 posts, 8 of them include the same three typologies, that being “NASA employee,” “picture of telescope,” and “lab technician.” In addition, 16 out of the 25 top posts for engagement featured a picture of the telescope. This is important to be aware of because it is indicative of what your audience is expecting, and likes to engage with from your feed. Another insight from the top 25 posts is that only 3 of the top posts were community based, and likewise, only 3 included featured artwork. From this we can gather that there may be a better way to engage your followers through these types of posts.

Average Like by Typology Item:

After creating our model using DataRobot, our categorical typologies showed us what content brought in the most engagement. Below, you will find every typology with their average likes.

● When a photo contained a NASA employee it averages 1812 likes, when it doesn’t it averages 1554 likes.

● When a photo is a shot of featured artwork it averages 1980 likes, when it doesn’t it averages 1578 likes.

● When a photo relates to general science it averages 2060 likes compared to 1522 likes when not relating to general science.

● When a shot has a lab technician in the photo it averages 1706 likes, when it does not it averages 1603 likes.

● When it is a picture of the telescope it averages 1652 likes, when it does not have a telescope it averages 1603

● When it is a shot of space the post average 1695 likes compared to 1613 likes when it is not of space.

● Community based post averaged 1634 likes, compared to 1627 likes when it is not.

Trends:

This graph is the average number of likes per month that this instagram account has been active for. This clearly shows that as time goes on, this account brings in more likes. This is very good, and shows that the more you keep posting, the more engagement you can expect.

Time & Day Analysis:

Content posted at 7 am, 8 am and 4pm gets the most average likes with 3091, 2210 and 2086 average likes respectively. Currently, the majority of content is published at 12 pm and gets an average of 1665 likes. The popularity of earlier posts indicates that @NasaWebb’s audience does the majority of their viewing before work or class. When looking at the top posts for content shared during this time, it appears as though visually appealing posts of the telescope and lab do best during this time period.

On a day-to-day basis, content released on Sundays does dramatically better than any other day with 4091 average likes. Fridays are the next most highest engaging day with 1911 average likes. While Sundays are the most engaging, the least amount of data is released with only post in our data, the majority of the content is shared on Wednesdays which only has an average of 1424 likes.

Predictive Analytics

Using the Elastic-Net Regressor (L2 / Gamma Deviance) model, we get an r-squared value of .64 meaning that we can account for 64% of the variation in the model. This is a fairly high r-squared value and indicates that there are several variables within content that impacts its audience reception.

The most significant of the variables was time, indicating that @nasawebb’s content has gained in reception by a large amount since the accounts creation. The increase in likes has been fairly consistent since 2014.

After time, the display URL is the next most significant variable. This indicates that the actual photo content of the photo matters. The URL structure represents content groups. Those defined with a display url of “3D,” “3D 3D,” and “e15” did best with coefficients of .51, .37, and .25 respectively. These are also the most frequently used. Conversely, the display url “com ig_cache_key” has the lowest coefficient at -1.

Another text variable, captions, is also predictive. Posts with captions that include “webb,” “telescope,” and “James” are most predictive of likes. While these words are among the most highly used, other frequent words like “space” and “tech” are not predictive with coefficients of -.37 and -1 respectively.

We see this trend repeated in the predictiveness of the tags “Webb,” “Webbtelescope,” and “jameswebbtelescope” with coefficients of .69, .33 and .33. Comparatively, the most frequently used tag, “NASA,” is not predictive with a coefficient of -.07.

The predictiveness of captions and tags that directly refer to the James Webb Telescope indicates that the page’s audience is specifically interested by and engaged in specific information about the telescope. It appears that users do not follow this account because of its general association with Nasa or with space, but because they are invested in the project itself.

What’s Working?

● Content that relates to NASA employees, featured artwork, and general science overall get the highest average of likes compared to the other typologies. Keep posting pictures related to these topics for the highest engagement

● Content posted on Sundays and around 7 and 8 in the morning performs the best. Post more content during this time frame.

● Content that contains captions or tags that are specific to the Webb Telescope do better than posts containing text variables related to generic space and science content.

● Keep posting pictures of the telescope!

● Keep posting about General Science!

● Keep posting on around 7am and on weekends!

● Keep using tags and captions such as Webb, #WebbTelescope

What Isn’t Working?

● Community based posts and pictures of the telescope get the lowest average of likes. If you are trying to keep your likes up, try to avoid posting pictures regarding these typologies.