Return on imagination
Today I’m going to invent a new metric.
I’m sure everyone out there remembers when their high school math teachers introduced them to set theory and the hierarchical classification of numbers. Right?
complex
/ \
imaginary real
/ \
irrational rational
/ \
integers fractions
| |
whole integers
|
naturals
(courtesy Ask Dr. Math)
Most of these sets have fancy representations of letters to designate them. For example, ℚ is shorthand for the set of rational numbers (it’s ℚ because you can express any rational number as a Quotient, not because a mathematician wanted to mess with people). While there isn’t such shorthand for the set of imaginary numbers, you can always spot a member because it contains some multiple of i, where i2 = -1.
The point of this math refresher is not to make your eyes to glaze over (sorry about that), it’s to set up a point. Folks in marketing and communications are increasingly being asked to justify costs associated with participating in social media. People with responsibility for budget allocations and profits and losses want to know their return on investment (ROI), or return on marketing investment (ROMI), or a host of other abbreviations involving R’s and O’s and I’s. The problem, as beautifully articulated by Olivier Blanchard, is that asking “What’s the ROI of social media?” isn’t a meaningful question.
The question, then, is not what is the ROI of social media, but rather what is the ROI of [insert activity here] in social media?
Social Media and ROI: Some clarity. (Again.) « The BrandBuilder Blog
When it comes to measuring the return on investment of social media activity, the reality is that not all actions have an outcome that results in direct financial impact. Which is not to say an action may not have value, just that said value may not translate right into dollars and cents. Can you measure efficiency from a financial standpoint? Sure. If you spent X dollars and got Y retweets, then you spent X/Y dollars per retweet. If you spent Q dollars performing social customer support and measured a 5% increase in customer satisfaction, you can legitimately say you spent Q/5 dollars per point of improvement. Congrats!
The problem is that I see lots of people (and agencies) present financial efficiency as ROI. It’s not. The most charitable thing you could call it is ROi, for it exists as return on investment only in the imaginary domain of unicorns and Little Nemo.
Quick Reads and Reactions
I apologize for the recent lack of activity; I had some unexpected travel, a dying computer, and other distractions keep me from stringing together meaningful thoughts. In the short term, I wanted to share some recent content that caught my attention:
Privacy management on social media sites | Pew Internet & American Life Project
Pew’s hot-off-the-presses report addresses changing sharing behavior in social media. Their survey seems to have been fielded in April and May of last year, which makes me suspect the degree of access restrictions is even higher now. Yes, Facebook is not the entirety of social media, but this story syncs up well with my analysis of the huge drop in public Facebook content a few weeks back.
Scoop: Facebook To Speed Up Biz Analytics Tool Insights To Report In Real-Time | TechCrunch
Speaking of Facebook, it seems like this Wednesday’s fMC event is going to be a Really Big Deal. While the recent addition of People Talking About This and other engagement-based measures was a step in the right direction, the growing reporting lagtime was not.
How to Measure Mobile-Data Use? AT&T Has Two Ways - Digits - WSJ
The Journal tries to parse some confusing math. The bottom line is that data usage is exploding, no matter how you measure it: 
Content: The New Marketing Equation
You can always rely on the Altimeter Group to package the current social media wisdom into a format digestible by people not immersed in the digital space. Anyone asking themselves “What is my Pinterest strategy?” should read this before their local social media strategist embeds their palm in their forehead.
Digg Data Reveals What We Read But Are Too Scared or Embarrassed To Share | TechCrunch
I love that the URL to this article is “Be yourself.” The gaps between content consumption, content sharing, and content creation are well-established at this point (if you hear someone talk about 90-9-1, this is what they’re referring to) and it’s something that companies who peer into those chasms make noise about when they want to get some attention. Here’s bitly’s spin on it from a few months ago.
It’s not the big data, it’s the right data — GigaOm
Finally, I really wanted to highlight a quote from this article:
“The real purpose of big data is to enable big analytics. The most compelling companies out there, I think, are those that attack that problem,” Palmer told me this week.”I really do believe that big data is, in and of itself, a tool. The real story is more about big analytics. Once you aggregate the data you then have to ask really hard questions.”
A Data Stew to Serve Man
I feel a little late to the party in linking to this, but let’s pretend my analysis is so incisive that it merits the delay.
Almost every major retailer, from grocery chains to investment banks to the U.S. Postal Service, has a “predictive analytics” department devoted to understanding not just consumers’ shopping habits but also their personal habits, so as to more efficiently market to them. “But Target has always been one of the smartest at this,” says Eric Siegel, a consultant and the chairman of a conference called Predictive Analytics World. “We’re living through a golden age of behavioral research. It’s amazing how much we can figure out about how people think now.”
What’s not included in this article is almost as fascinating as what is. I’m genuinely curious about what factors went into determining the focus here. Was it that the author only had one juicy source, whose output was direct mail marketing, and one substantial case study, involving TV commercials? Was it an attempt to differentiate from the Wall Street Journal’s “What they know” series?
Here’s a fun exercise: Search the text for the word “Google.” Zero results. This is especially odd if you remember the Google Search Stories ad campaign. Searching for “Amazon” gives you one quote in the 9th paragraph. Online advertising and ecommerce are mentioned in passing; there are more references to the web in the first few comments than in the article itself.
It really feels like this should have been addressed. Behavioral ad targeting is at least as sophisticated as any of the work described here. PRIZM data is good, but PRIZM and BlueKai data together is better.
Even looking strictly at offline tracking, there were some notable omissions. For true creepiness, consider what Target may already be doing: Think fun buzzwords like RFID or NFC. Tie a GuestID to second-by-second in-store location mapping and BAM! Baby, you’ve got a stew goin’.
Social media warrants consideration as an ingredient in this data stew (or maybe it’s more of a data gumbo?), but I’m a bit skeptical of how valuable it is at the present time. Consumer packaged goods are what are politely referred to as a “low interest category.” There are a handful of products that provide excellent segmentation cues (think diaper conversation on TheBump revealing a mom). But segmentation is a Hard Problem if you’re not Facebook, Google, or a data management platform. Getting access to those data sources is either expensive or closely guarded, which is why most social media listening exercises try to wriggle out of doing it. What’s more, social customer relationship management is not mature enough yet for Target to effectively offer a coupon to someone who (in a highly unlikely event) tweets “I love Kmart.” An analytics team lacks sufficient data to tell you the precise value of said coupon to get that tweet’s author to defect.
The reality is that there is (an assumed) level of privacy associated with browsing the pharmacy aisle or a conducting a Google search that is not present in social media. As convenient as it would make marketers’ lives if people publicly stated their intent at all times, it is not the world that you, I, or John Wanamaker inhabit. As he famously said, ”Half the money I spend on advertising is wasted; the trouble is I don’t know which half.”
Pinterest Measurement Trick
Found a really helpful tip on tracking Pinterest content this morning:
Discover What Are People Pinning from Your Web Site
When clicked, every image on Pinterest displays corresponding information like comments, “likes,” other images in the same board and more. The info I find interesting is the area on each pin that shows what other pins came from a specific web domain. Take a look here for instance. You can see all images pinnned from socialmedaiexplorer.com from all users on Pinterest. You can see right off the bat that people enjoy the infographics here on the site. Most popular after the infographics is an image of Jason’s recently published book. Remember, each of these images could have been pinned from any page on socialmediaexplorer.com. Pinterest conveniently collects them all in one place for you.
Want to try it on your site? Type the following into your browser and replace “yourdomain.com” with your own web site: http://pinterest.com/source/”yourdomain.com”. You’ll likely find out something interesting about what visitors to your web site find visually interesting to them.
Let’s look at another example with the folks at FastMac: http://pinterest.com/source/fastmac.com/. Here we can see, out of all the products that Fastmac sells, 99% of people have pinned images related to their USB wall socket. Not only an image of the product itself, but the actual ad image on the product page.
The hollow emptiness in social media numbers - most accounts are fake or empty | ZDNet
I’ve been concerned by these same issues lately. Using tools like SimplyMeasured and SocialBro to analyze accounts for large brands, the proportion of followers that have 0 tweets, suspiciously identical account names, or sketchy bio info (do US-focused brands care about engaging with thousands of followers in Jakarta?) is shocking. And that doesn’t even take into consideration the legions of devoted Justin Bieber fans :-p
But more seriously, this is yet another reason to cultivate quality and not just quantity as you grow a social media community.
Why Radian6 is wrong for you - a Radian6 review (Part II)
The short answer: Abysmal Facebook data quality. For the longer answer, read on…
Image caption: I fight for the user
First, I want to reiterate the considerations I shared in Part I:
I am going to call out some specific scenarios. I freely acknowledge that the plural of anecdote is not data, but I believe these situations to be representative of serious issues, based on my personal experience.
And about that personal experience… I’ve been using social media data as a research input for the better part of a decade. I’ve led studies conducted across multiple countries, in multiple languages, in multiple use cases, for multiple Fortune 50 companies. A large portion of that time was spent at a Radian6 competitor, in various research and client service roles. I would understand if you thought I had an axe to grind against them. The reality is that this is a space with hundreds of competitors, and the issues I’m going to cover are by no means exclusive to one vendor. If you ask any of the product managers I’ve collaborated with, they’ll tell you I am an equal opportunity offender when it comes to criticism of tools.
I’ve been providing recommendations and insight to my clients based on Radian6 data since I joined my employer a year ago. So it’s at no small amount of professional risk that I call out Radian6 by name when I bring these issues to your attention. Hopefully, it also means they will take meaningful and rapid steps to address these problems.
Next up: Data access, data manipulation, and customer service. I promise it gets less snarky after I address data access. Really.
Harvard Researcher Uses Social Media To Predict Stock Market Volume
This isn’t actually a new thing, a hedge fund called Derwent Capital has been doing this for several years (and is still in business)
Why Radian6 Is Wrong for You - A Salesforce Marketing Cloud Radian6 Review (Part I)
The short answer: Abysmal Facebook data quality. For the longer answer, read on…
Mining Tweets for Public Opinion
Everyone seems to be gushing lately about how social media can serve as the world’s largest focus group.
Companies are high on social media for a number of reasons, but perhaps chief among them should be that social platforms provide the opportunity to create focus groups at a scale never before possible. Millions of people talk about all sorts of things online, and with the right systems and algorithms in place, it’s possible to decipher how they actually feel about the topics they’re discussing. If you want to know how the web-savvy world feels about a product, movie, team, you name it, millions of data sources should trump interviewing a few hundred people in malls across the country.
How social media is making polling obsolete — Cloud Computing News
While I agree with the general sentiment here (ha-ha), it does leave out one crucial point: the difference between prompted and unprompted opinions. You may get a read on the most passionate responses if you passively monitor social media, but then you’re ignoring the huge group of people out there who wouldn’t care enough to say anything…unless you asked them first. To get around this, you would still need to use an old reliable survey — or the new shininess of an MROC.
NetBase, one of the social-media-analysis companies, said that it, essentially, does have a sarcasm detector. For instance, company officials say they can detect that “Thank you [wireless provider name] for messing up my bill” is not a customer rave. … NetBase also claims it can handle slang and Internet-age codes. “We have assigned meaning to all of the emoticons out there,” said Lisa Joy Rosner, chief marketing officer for NetBase. “When something is sick, either they have the flu or it’s the hot new thing.”
“Sucks” doesn’t mean something bad… if you’re Hoover. But seriously, the major issue I’ve seen with automated sentiment algorithms is how they deal with comparisons. When people talk about two companies/brands/products so closely together, algorithms can get very confused about which one is preferred, and why.
Matt Pierson's answer to What are the tangible differences between Radian6, SM2, Meltwater Buzz and other SoMe listening tools? - Quora
I wrote this about a year ago and it holds up pretty well:
Pricing Structure: Options include seat licenses, price per keyword, and price per # of messages returned. I’m starting to see options where there is partnership with an IBM-type company where you can pay for a company to manage dedicated hardware inside your corporate network (this is really for enterprises, not mom and pop operations).
Content Types: Facebook, Twitter (with different amounts of access to the “firehose”), Blogs, Message Boards, Newsgroups, Video sites, Picture sites, Traditional media. To name a few.
Languages: It’s not all in English :-) Some tools offer in-line translation, others offer in-market analyst support, others let you fend for yourself, and still others provide no non-English data whatsoever.
Filtering: Most tools rely on the creation of queries. Modified Boolean operators are common (AND, OR, NOT, and then special ones like proximity) but there are also some that do more sophisticated semantic analysis. There are also different options for segmenting data into sets of sites or even individual sites, forums, or authors. This helps reduce the amount of spam and messages irrelevant to your listening objectives.
Workflow/Engagement: Some tools let you not just listen but also respond. Not to get too abbreviation heavy, but some even integrate with CRM and CMS.
Sentiment: Many tools offer automated sentiment algorithms (of varying levels of accuracy). Some offer the ability to manually assign sentiment to individual messages.
Interface: There’s a variety of visualization options; most of them revolve around a widget system. Some tools offer an API, in which case you can build whatever interface you want!
Support: There are some standalone self-serve dashboard options, but also more holistic offerings including analyst teams and strategy consulting.
You can see how someone like Forrester differentiates in their Wave reports; here’s the most recent one.
There are really on a few things I would add or change
Geolocation: This straddles languages and filtering. I haven’t encountered tools that can reliably do this well. On a site by site basis, the only obvious cue about a site’s location is the Top Level Domain. In theory, a site ending in .it resides in Italy, a .cx on Christmas Island, and so on. The problem is that nothing stops a site owner from registering under other countries’ TLDs. I would wager that only a small percentage of .me domains are by and for citizens of Montenegro. Having audience data from a comScore or Nielsen can help here, but what to do about a site incredibly popular across borders, like, say, Facebook? Tools are doing slightly better with Twitter and blogs. Profile data including location can help get down to the city and state level, but only if people don’t lie. Tons of people on Twitter claim to be on the moon, and how many accounts turned their profile pictures green and relocated to Tehran in 2009?
NLP (Natural Language Processing): Rather than a straight Boolean query, tools from companies like Brandwatch and Visible Technologies can discern semantically-related words and phrases beyond what you enter in a query. Since there’s just no way to account for every possible way people talk about a particular brand or product, this can be a huge help.
Data Collection Methods: One thing I didn’t include last year that I really should have is how posts get into a tool. Some companies leverage third parties like Boardreader to harvest data. From an operational standpoint, it makes a lot of sense to outsource this kind of work. The problem is that it makes companies dependent on a third party to update and add sources. Similarly, companies can choose to use an API to collect a site (assuming one is available). An API ensures that data is coming in quickly, and contains lots of metadata, but it can also impose arbitrary restrictions. The Facebook API, for example, does not allow the collection of posts made to public fan pages if the author does not make all of their status updates public. When I found the 90% decline in public status messages, I actually saw an even bigger decline in fan page posts. I can’t make any determination about whether fan page activity increased or decreased, since I don’t currently have access to a tool that scrapes non-API content.
Content Types: This isn’t a miss, so much as something that didn’t exist at the time :-) Google+. Sysomos has included some Google+ data for several months. I don’t have access to it, so I don’t know the quality, but people are going to want this data included since it has New Shiny Thing status.
