by Dennis Yu | Aug 23, 2019 | Advertising, Analytics, Boosting, Facebook Advertising, facebook analytics, facebook marketing and advertising, local advertising
5 years ago, Facebook had an auto-boost feature where they would automatically boost your posts. But they killed it because not many people were using it and because the system was boosting posts about site outages, sales that had already expired, and other things.
So this new version gives you a bit more control:
You can choose the default boost amounts (how long and how much to spend each day), the audience (from your list of saved audiences), and if you want to auto-approve:
Some things I don’t like about this re-released product:
- Only one post gets boosted at a time– I like to be able to put more money on winners, even to have them live forever. As a business grows, we would want to have a growing number of posts boosted evergreen (forever) as part of our Greatest Hits that live forever.
- The 60% threshold for “top posts” is arbitrary. Instead of getting 60% more engagement than our average post, it should take the top 10% of posts by engagement or all posts that meet a particular fixed engagement threshold (like 10% engagement/impressions or 10+ second average watch times on videos).
- It’s buggy– I’m not able to switch the audiences. And the reported engagement figures don’t make sense– how do I have no engagement, yet 144 engagements?
Have you had a chance to play?
by Dennis Yu | Oct 5, 2013 | facebook analytics
We promoted this post
(evidenced by the “sponsored” label at the bottom) and see suggested pages beneath it.
Some people say these suggestions are paid– they’re not.
But what’s curious is how Facebook decides what to show.
This would be the math behind lookalike audiences.
E-commerce players like Amazon use this technique to recommend related products.
It’s called collaborative filtering– what any first year statistics student can run. People who bought X also bought Y. Or people who watched this show also liked that show.
Are they looking at my activity to see what other pages I personally might like?
We’ve been meeting with folks like GoDaddy and Quicken Loans, who sponsor NASCAR teams. So the Miss Sprint Cup suggestion makes sense.
And I was in Miami a week ago– so the “Visit Miami” page seems quite relevant.
BlitzMetrics isn’t based in Miami or happen to have a stronger concentration of fans/traffic in Miami. So I’m leaning towards #2.
We have a mix of big brand and entrepreneurial folks on our page, so the SpringWise suggestion makes sense.
What do you think?
by Dennis Yu | Mar 13, 2013 | facebook analytics
Some similarities for sure in any set of companies that skyrockets to a massive user base and corresponding corporate baggage. However, I was part of the Yahoo! Data Team and can tell you that Facebook is FAR more aggressive with data than we ever were.
The General Counsel of Yahoo! refused to let us tie browser and log-in cookie data, which would have given us better targeting on the 80% of users who weren’t logged in.
We wanted to personalize ads in Y! Mail way before gmail– no, because users said it was creepy.
We wanted to roll out conversion optimization well before AdWords, but it would have dented next quarter’s earnings. So no, even though it would have been a smart long-term decision.
Yahoo’s earnings are driven by a strong display sales team– arguably the best in the world. But the world has moved to self-serve. Facebook hasn’t dominated self-serve like Google, but they are making promising moves.
I’d love to see Facebook fix critical analytics bugs
and roll out something stronger than Google Analytics. Never heard of Yahoo! Analytics? There is a reason for that.
I own no stock in either of these companies– just a guy who looks at things from the standpoint of the data and what we can do with it. I’ve never seen so much data available to marketers– so Facebook could
win in this area, if they can teach companies how to use it appropriately. What do you think? IS Facebook the new Yahoo!
by Dennis Yu | Feb 22, 2013 | facebook analytics
Email received from Facebook outlining the “bugs” in insights. Click for full view.
We’ve known that calculating reach and frequency metrics for Facebook is a tough engineering challenge, especially to do it in near real-time.
While Facebook has been vague as to which metrics are affected and to what extent, our guess is that the calculations of uniques by day and post are the main issue. Interactions, fans, and other aggregable metrics are easy to calculate– but deduping users across multiple sessions and devices is not.
Reach is really just unique impressions. And the sum of daily reach across a week is not the same as weekly reach, since people can see messages on multiple days. In the world of web analytics, we call this the non-aggregable metrics issue.
On February 25th, you can go into your analytics to see what’s changed.
Make sure you take a snapshot now, since they’re not going to keep the old, incorrect data for you.
Meanwhile, I’d rely upon other methods
to analyze Facebook data via the graph API and ads data.
by Dennis Yu | Jul 6, 2012 | facebook analytics
We grabbed 1,000,000 random posts from Facebook and tracked all interactions at a user level. We used the graph API, which is open to anyone. [Side note: if you’d like to learn how to pull this data, ping me.]
- Women tend to “like” more than “comment” on a post. Worldwide, 51% of post likes were from females, though this was 55% in the US.
- Women “Like” comments. In the US, 59% of likes on comments were female
- Males post on pages more frequently than females– at 51% vs 48%. The difference is gender unknown.
For more on Facebook Analytics or to get your sneak peak social dashboard, head on over to BlitzMetrics. Or come check us out at facebook.com/blitzmetrics, where we’ll answer your questions!