97040861251

Would the Apple keynote win a TV Ratings battle? Probs.

TL;DR, tech keynotes, like Apple’s + Google’s, have a live viewing audience size comparable to/greater than cable news airing at same time.

image

Some notes on how I broke down this estimate with little public data (I welcome any more data to fill this out more).

GETTING THE GOOGLE I/O AUDIENCE SIZE

In June, Google hosted a 9am keynote for their IO developer conference. Since they used YouTube, the concurrent viewer number was visible, and I took a screenshot.

image

GETTING APPLE AUDIENCE SIZE

That same month, Apple had its own developer keynote, with livestream.

It’s fair to say that Apple events seem more popular/highly anticipated than Google product events. Since Apple viewing data is not public, I decided to compare social share info (FB shares, FB Likes, twitter shares, LinkedIn shares) of both events’ keynote URLs, and extrapolate from that the size of Apple’s live video audience.

If my Apple number ends up being wildly innaccurate, I’d guess it much higher, not lower, coz I rounded down.

TV NEWS DATA

I got Nielsen data for CNBC, Fox Business, and Fox News (FNC), from TVNewser.

I compared the tech keynotes primarily with cable financial/business news coz it seemed most similar in format/content.
As the chart above shows, even the smaller Google keynote has a larger audience than the financial news networks during that time slot.

I also threw in, for scale, the number one rated cable news (not just business news) show in that slot, Fox News’ Outnumbered. Being such a big show, Outnumbered has significantly more viewers than the tech keynotes.

Or does it?

When you look at the “key demo" (viewers 25-54), even FNC’s Outnumbered gets beat by the likely Apple audience size, and the Google keynote gets competitive, too.

If you have access to any data you think could fill this in more, hit me up on twitter- @bluechoochoo

image

96668858671
[update: added Buzzfeed trends]
As blogged yesterday I’ve started extracting Tumblr trends with importio. (Long story.)
I’m not yet 100% sure importio actually saves me thaaat much work, but I’ve committed to really learning it (and its peer/competitor, Kimono Labs), before just scraping things myself (did i say, “scraping?” i meant extracting.).
In the screenshot above, I’ve created a lil’ dashboard in Google Spreadsheets of:
Tumblr Trends
Twitter Trends
Google Search Trends

[update: added Buzzfeed trends]

As blogged yesterday I’ve started extracting Tumblr trends with importio. (Long story.)

I’m not yet 100% sure importio actually saves me thaaat much work, but I’ve committed to really learning it (and its peer/competitor, Kimono Labs), before just scraping things myself (did i say, “scraping?” i meant extracting.).

In the screenshot above, I’ve created a lil’ dashboard in Google Spreadsheets of:

  • Tumblr Trends
  • Twitter Trends
  • Google Search Trends
88291593286

tl;dr: Vines now preview in the Tumblr (web) Dashboard. Instagram videos are still blank black boxes.

As previously documented, unlike rich media content from YouTube, Vimeo, and SoundCloud, Vines used to show up in the Tumblr Dashboard as blank, black boxes of mystery (and as a result, were vulnerable to being scrolled. on. by.)

But now Vines show up. In fact, Tumblr has made an interesting decision- rather than thumbnail as stills (like YouTube, Vimeo, native Tumblr videos do), Vine videos autoplay (with sound off). The effect is kinda GIF-like (on web).

Thanks, tumblr staff, for fixing this up. 

[Update: I heard from a Tumblr engineer on Twitter. Apparently there are two reasons why Instagram is a blank black box on Tumblr: 1) Instagram doesn’t provide video thumbnails 2) embeds are not https]

PS- i’m not gonna suggest it’s the reason for this tweak, but i’ve been noticing a lot of popular Vine tumblrs (like vinebox + weloveshortvideos) in the past 2 months. At this point, I discover more Vines on Tumblr than on Twitter or Vine itself.

unwrapping:

Vine Autoplays in Tumblr Desktop Dashboard:
I’m not sure what impresses me more: watching the first Vine from space or witnessing a Vine video autoplay in my Tumblr Desktop Dashboard. (My Tumblr mobile apps show “Video not compatible.”)

To add a Vine on Tumblr, create a video post and simply paste the Vine post URL into the top text box. For example, the URL to this milestone Vine is https://vine.co/v/MD1eEQEjM9u. As you are composing, the Vine will look like a video with play button. But when you publish, you’ll see the Vine autoplays on the Desktop Dashboard with volume muted. On my blog, the desktop browser immediately loops the Vine, also with volume muted. For Safari on iPhone and iPad, I have to tap the play button to start the Vine.

86341629666

I know this is an old piece but I wanted to let you know I enjoyed it. "Re blogs and content sharing on Tumblr: a personal network analysis"- I am very interested in the "tipping point" of tumblr blogs. ie how many followers to you need before your followers begin to generate exponentially? How do you find the "connectors" and "influencers" on Tumblr? Its all interesting stuff but quite hard to track! thanks- E

hautepop:

Digging into my Ask box and responding to very old questions - this one on this post doing a little network analysis of reblogs from Dec 2012

1. How many followers do you need before your followers increase exponentially?

That’s a tricky one - mostly because Tumblr doesn’t make follower counts public on people’s blogs or in the API, so it’s fairly impossible to do objective research. What I can say though is that follower numbers matter a lot less than your position within Tumblr social networks - that is, it’s about community.

When I say community, this can mean self-identifing communities such as TV fandoms or Feminist Tumblr (ok, lots of communities within Feminist Tumblr!) However, it’s not the self-identification as part of a community that’s what really matters. What’s actually important is the habitual pattern of who interacts with who through reblogging and commenting on each others’ posts. It’s community as defined by social network analysis.

Become a part of a really active community, put a lot of time into writing good posts, following people and reblogging - and your follower count could increase very rapidly.

On the other hand, having loads of followers doesn’t necessarily get you anywhere. I’m a great example! So I have 164,00 followers because I was Tumblr tech listed & recommended to new users on sign-up. That gave me 250 new followers per day. Now that’s stopped, I get maybe 5? Two main reasons:

  • Essentially all the people who followed me were newbies with zero network power (and most of them probably stopped using Tumblr really quickly), so they provide no network advantage to spreading my posts.
  • I don’t really participate in any Tumblr communities so I am relatively marginal within the network.

(My network power is much stronger on Twitter despite far fewer followers - but that’s another post…)

2. How do you find the connectors and influencers on Tumblr?

As I said, with no publicly visible follower counts, there’s no easy way to do this. You can do it through social network analysis, collecting a big pool of posts and analysing who reblogs who. However, you need proper social media research resources to do this (API access, some Python skills, and use of programs like NodeSQL or Gephi). This happens to be my job - but I work alongside a developer team to do the data wrangling.

But all is not lost! From the principles of network analysis we can draw out some rules of thumb that can provide pretty good proxy methods for finding these super-connectors in the Tumblr niche you want to be part of. So:

1. Look at a Tumblr post you like/admire, and expand the list of notes so you can see everyone who faved & reblogged it. Posts with 100+ notes are most useful here, as are ones where a bit of a “reblog conversation” has gone on.

2. Looking at the message body, make a note of:

  • original author
  • rebloggers who’ve added comments that have then become part of the ‘reblog tree’ discussion

3. Looking at the list of notes for the “A reblogged B” format, identify which Tumblr users seem to have been reblogged the most.

4. Repeat this for 10-20 more posts, so you’ve got a pool of 50+ names. Are any names starting to recur? This indicates they’ve got high “network centrality" - that means influence.

5. Take the list of names you’ve got (heavily-reblogged original authors and popular rebloggers) and go to their Tumblrs to see what’s going on. Check out:

  • The note count on their original posts (if they make any): what kind of engagement are they getting?
  • Their follower count (some people do list it in bios, and of course the more influential are more likely to brag!)

This will help you rank your list of names.

6. Iterate the process. Look at the Tumblrs of the top 10 most influential people you’ve identified, and explore who they’re reblogging, who’s mentioned in their posts, and who’s heavily reblogged in their Notes.

Add these people to your Influentials list.

7. Follow the people you think are really interesting. They might not necessarily be the most influential ones on the list - but you’ll have a better chance of making a connection to them if you think alike.

Bonus step 8:
To get influential users to interact with you, the best way is probably to reblog something they’ve written, adding an interesting comment or question on the bottom of it. They’ll be pretty likely to reply & check you out, creating the opportunity for them to decide to follow you.

People also use Ask boxes to make themselves known to others and build relationships - but you do need to have something more interesting to say than “Please will you follow me back?”

*

What do you reckon, Social Network Analysis Tumblr? How’s my heuristic? Anything I’ve missed out?

Pay special attention to section two, How do you find influential Tumblrs? (paraphrased). hautepop gives some great rules thumb here, but it’s a question so many people have. Tumblr does not list follower count, unlike Twitter and Pinterest. So what are your options?

I’m reblogging this post completely here (Tumblr doesn’t give you a choice with Ask reblogs, unlike other post types), but stay tuned for my next post, where I share how you can “rank” Tumblrs, without knowing follower count, and without using Tumblr’s API.

#suspense

84488790781

Recommendation Engine tutorial

Marketers, if you’re interested in this topic, even if you don’t intend on setting up infrastructure on your own, I recommend clicking through to their blog and watching the 44 minute tutorial. You may think, “I’m a marketer, I don’t need to get this stuff.” Then how come you’re studying social network analysis? (There’s a bit of overlap! #gotcha)

In the tutorial, you’ll learn a bit about recommender systems in general. Also, if you are interested in your own set-up, he gives really cool options for every step of the way. You don’t haaaaaave to use the Mortar service. You can use free/open source stuff for every part.

mortardata

A few weeks ago we made an announcement: our recommendation engine, which our engineers and data science advisors built in consultation with select partner companies, is now open source and available to all. The response from the community has been tremendous—tens of thousands of people read our blog post announcing the news. And many of them dove right into our Github repo and tutorials and got to work.

If your company is among the many that use MongoDB, we’ve just made the path to generating personalized recommendations even smoother. Our step-by-step tutorial on building a recommendation engine is peppered with callouts that highlight information specific to Mongo users. We have an entire section of our documentation site devoted to analyzing MongoDB data with Hadoop, which is at the core of our recommendation engine. And our CEO, K Young, recently showed MongoDB users how it all fits together by hosting a webinar that walks through the process of building a recommendation engine with Mongo data.

Our goal in building the Mortar recommendation engine was to create a product that any data scientist or engineer could use, regardless of budget or experience with Hadoop and its associated technologies. So we made all our code available for free, and we wrote an obsessively detailed tutorial that is as informative as it is easy to follow. Our integration with Mongo (and extensive documentation thereof) makes the recommendation engine even more broadly applicable. If you missed the webinar, a video recording of the presentation is embedded below, along with K’s slides. So give it a look—and then dive into our tutorial to start building a powerful recommender with your Mongo data.