Tag Archives: social media
Live note-taking at #IIeX in Atlanta. Any errors or bad jokes are my own.
Chaired by Seth Grimes
Automated text coding: humans and machines learning together by Stu Shulman
- It is a 2500 year old problem, Plato argued it would be frustrating and it still is.
- Coders are expensive, it’s difficult at scale, some models are easier to validation than others, don’t replace humans, no one right way to do it, validation of humans and machines is essential
- Want to efficiently code, annotate coding with shared memoirs, manage coding permissions, have unlimited collaborators, easily measure inter-rather reliability, adjudicate validity decisions
- Wanted to take the mouse out of the process, so items load efficiently for coding
- Computer science and HSF influence measure everything
- Measure how fast each annotator works, measure interacter reliability, reliability can change drastically by topic
- Adjudication – sometimes it’s clear when an error has been made, allows you to create a gold standard training set, and give feedback to coders; can identify which coders are weak at even the simplest task, there is human aptitude and not everyone has it, there is a distribution of competencies
- 25% of codes are wrong so you need to train machines to trust the people who do a better job at coding
- Pillars of text analytics – search, filtering, deduplication and clustering and works well with surveys as well, human coding or labelling or tagging which is where most of their work goes, machine learning – this gives a high quality training set
- If humans can’t do the labelling, then the machines can’t either
- Always good to keep humans in the loop
- Word sense dis ambiguities – relevant – is bridge a game or a road, it smoking a cigarette or being awesome
Automated classification interesting, at scale and depth by Ian McCarty
- Active data collection is specific and granular, as well as standardized; but it’s slow and difficult to scale, there is uncertainty, may be observer bias via social desirability, demand characteristics, Hawthorne effect [EVERY method has strength and weaknesses]
- Declared vs demonstrated interests – you can give 5 stars to a great movie and then watch Paul Blart Mall Cop 5 times a 6 months [Paul Blart is a great movie! Loved it 🙂 ]
- They replicate the experience of a specific URL to generate more specific data
- Closed network use case – examined search queries from members to recruit them into studies, segmentation was manual and company needed to automate and scale; lowered per person costs and increased accuracy, found more panelists in more specific clusters, normalized surveys if declared behaviors conflicted with demonstrated behaviors
- Open network use case: home improvement brand needed a modern shared meaning with customers, wanted to automate a manual process; distinguished brand follower end compared to competitive followers, identified where brand values and consumer values aligned, delivered map for future content creation and path to audience connection
Text analytics or social media insights by Michalis Michael
- Next gen research is here now, listening, asking questions, tracking behavior, insights experts
- Revenues don’t reflect expectations, yet.
- We’re not doing a great job of integrating insights yet, social media listening analytics is not completely integrated in our industry yet
- Homonyms are major noise, eliminating them needs humans and machines
- Machine learning is language agnostic, create a taxonomy with it, a dictionary of the product category using the words that people use in social media not marketing words
- It is possible to have 80% agreement with text analytics and the human [I believe this when the language is reasonably simple and known]
- Becks means beer and David beckham but you need training algorithms to do this, Beck Hanson is a singer, you need hundreds of clarifications to identify the exact Becks that is beer
- Beer is related to appearance and occasions, break down occasions into in home or out of home, then at a BBQ or club
- What do you say about a beer when they do a commercial that has nothing to do with the beer
- English has s a lot of sarcasm, more than a lot of other languages [yeah right, sure, I believe you]
- Break down sentiment into emotions – anger, desire, disgust, hate, joy, love, sadness – can benchmark brands in these categories as well
- Can benchmark NPS with social media
- Brand tracking questions can be matched to topics in a social media taxonomy, and there can be even more in the social media version than the survey version
Sharing is nice:
Ask first, listen later by Lori Reiser
- Traditional research often starts with the business, product iterations, product marketing, product fine tuning
- We should put people first so you hear about unmet needs and pain points
- Case studies
- Health insurance firm – young adults getting their first insurance plan, and new retirees moving away from employer benefits; had predefined assumptions but those were based on the six people in the room, fears of these people were financial and health and being bored not really insurance, how did they define good health and how could that be protected, did focus groups and bulletin boards, retirees weren’t worried about getting sick but rather that their parents would get sick, younger people were more worried about stress as their health issue
- Meat company – what did consumers need in terms of communication needs, saved qualitative for the end of the research, started with a survey of staff members and inspection organizations, realized they needed to formalize their email address to clients so that it didn’t come from Annie Pettit but rather from the company, realized that Mennonite members didn’t have email addresses [pay attention to that anyone who says they do probability sampling via RDD]
- Pharmacist – surveyed pharmacists as well as focus groups and the focus groups were after the fact, what does patient centered care mean, many barriers in terms of how pharmacists communicate with people given what doctors and other people want them to be able to say
- You can’t ask broad questions unless you go qualitative, open ends on a survey arne’t going to cut it; give your users permission to participat in the idea making, find the trendsetters and listen to them, use skilled moderation
Classified: Research that integrates to innovate by Mark Wood
- Technology has given us an identify crisis, people challenge are traditional beliefs, anyone can do DIY research, other people are jumping into our sandbox with new types of data
- Tactics to get our mojo back
- Understand consumer dynamics in a more connected world
- Leverage expertise in a data curation to improve SML capabilities, Think both/and between traditional survey and SM, Help companies navigate path to purchase, use data to help action
- Have to work with clients better and bring in other pieces of data so we can inform them better
- Bringing MR expertise into social media listening
- We need to take SM data to the full extent just not report facts and figures
- Must get access to this data first and then we can sample and structure and clean it
- SM provides an authentic voice of the consumer, has rich detail; surveys enhance social media structure to categorize themes and related back to real initiatives
- Use social and survey together to inform each other
- Need to identify lots of buzz with lots of impact, not just lots of buzz
- Use buzz impact matrix to understand which conversations have high buzz, high impact, and then dive deeper into those issues
- Identify all the touch points that need to be assessed, identify which have the highest reach, but there are big differences in claimed versus reality
- Need to move from measurement to action [the problem of the century!]
- Opportunities are greater than ever for MR to have an impact
Sharing is nice:
Live blogged at MRS in London. Any errors or bad jokes are my own.
The art of noise by Tom Ewing
- #IPAsocialworksinitiative – to guide effective use of social media data
- It’s all about measuring not counting, Ray Poynter wrote this part
- The insight guide is being written by Tom Ewing and he is interviewing people for it
- Five basic questions
- What questions do you need to answer, what decisions will betaken because of it
- What do you know already, existing research, behavioural
- What social behaviora are you looking for, what data sources, do you need aggregate data or granular level
- What is the relevant timescale, ad hoc or continuous, how do you know yyou’re finished
- Do you need quant metrics, what are they, it’s more than likes follows
- We have reached the peak of social tech
- Seth Grimes says the state of social insight is “confused,” it’s not really a mature technology, widely adopted, but not fully integrated
- It’s the early majority now, tactical, silos, unstructured text, the future is here but it is unevenly distributed
- We need to clear out the cliches, integrate with methodologies, integrate with insights and business and people
- Different poll results would arise if you use Twitter vs Facebook [i see this as a sampling problem, you don’t just take any church or school or company and say it is representative]
- UK Facebook data is closer to a census than a sample [but it is still very much a sample!]
- You can join insights with other methodologies [data fusion is ALWAYS wise, every method reveals different insights]
- Provide more file and predictive tracking systems and an ability to get granular and specific
- Who owns social insight? Is it marketing? Analytics? Insights? PR? Sales? It’s more about collaborating [why must one group OWN it, everyone can benefit hugely from it]
- Finding common ground – we’re all working toward the same goal, emotional understanding of people
- Limits of real time means it’s more tactical but there is indeed strategic value
- How do you get insight about a topic when you know nothing about it [oh goodness, I could go on about this. I was asked to draw insights about cars and trucks. I know NOTHING about them and couldn’t understand anything people were talking about. Expertise in the topic is a MUST! Put me on the food/outdoors/music research.]
- AI is already doing so much work that people would have done – collecting data, sorting and coding, analyzing, prepping, reporting
- Social insight begins with social practice, use new platforms like Whatsapp and snapchat, look beyond the brand level to the topic level, consider pictures and video, look for patterns at the network level
- Different things come out via text and images but you wno’t know unless you plan and look for differences in different types of content
- We are working towards general use, stratégie, cooperation, seamless integration among people, rich media as well as unstructured text
- Tom asks people to submit their social media case studies to him for publication in the book
Finding insight in 140 characters by Jake Steadman
- Jake [not from Statefarm] speaks
- Story about Star Wars, in 1977, need to tap into mainstream media, tap into core group of fans, went on TV and publications. FOr second movie he went to com icon. Fans went in the hundreds to see them in person for Q and A.
- Today, mobile is eating other media time.everyone has a mobile phone. December of last year, new Star Wars came out. Still need media and interest and fan base. THey launched on Twitter. That tweet drove all the other media. Drove coverage in print. Access to the stars still applies but they did it more directly and personal on Twitter. Stars did this on the red carpet, backstage, White House visits, all on Twitter. 1.2 billion tweets on opening weekend.
- All social data can drive insight
- There are culture issues and there are tools.
- Rise of the machines, machines allow you to become storytellers and consultants, it allows you to be a leader not a methodological policeman
- Let’s you do same day insights and be agility, we need to get over our obsession with precision, forget statistical precision [agreed, don’t create precision with decimal places, know when your data is directional and don’t make it more than it is ]
- Social is the democratization of data, client side is more like to be using social, agencies are more likely to be using big data, agencies need to be more for less and quicker
- Look for your soggy fries – customers were starting be less loyal with a restaurant, you could run a large research project, or you could look at social data immediately, found many uses of the word ‘soggy fries’, it didn’t need precision but rather it needed recognition
Panel with Jenny Burns, Christopher Wellbelove, Jess Owens, James Devon
- Some people jump at social media because it’s cheaper than things like focus groups [be cautious, cheaper means you need to spend more time thinking and analyzing it]
- Lots of people are analyzing social media without ever using it, you need to use the tools to really understand them [the speaker has now gone through three pairs of google glasses, and might be tweeting as we speak 🙂 ]
- Social media is still treated as a slightly dangerous cousin
- 67% of people using social media puts it really into late adoption, it is not new nor early adopter territory
- It says more about the market research industry being years behind to be so late in using this technology
- Privacy and ethics still matter, some companies get individual permission to review Twitter accounts
- People need to understand how their data is being used, whether it’s aggregated, Facebook data is fully anonymized and aggregated and great for ethics and privacy
[I kind of like have a couple presentations followed by a separate panel]
Sharing is nice:
Keynote: Why Social Media “Likes” Say More Than You Might Think by Dr. Jennifer Golbeck, Human-Computer Interaction Lab, University of Maryland #ISC2015 #MRX
- how much do two strangers online trust each other? you need to know about the people themselves first.
- social media data can give you accurate data, even when used in a simple way
- can predict introversion, political leaning, procrastination, personal health habits from your posts
- likes on facebook are always public, can use this to predict many things
- top four likes most indicative of high intelligence include science, thunderstorms, colbert report, and liking the page for curly fries. For low intelligence, the like was for the page “i love being a mom” But these don’t truly relate to intelligence.
- homopholy – we are friends with people who are like us, our traits are more common among friends than among random people.
- you just need one smart person to like a page and then it spreads like a virus to all your friends
- we’re really good at predicting sexual orientation, even in places where being gay could get you executed
- how does a store like Target figure out you’re pregnant, not from pregnancy tests, – an extra bottle of vitamins, a bigger purse, brightly coloured rugs together are a statistical connection
- how can you know what these statistical combinations are?
- can your social media network figure out which friend is your spouse? you know which friends are friends with each other and nothing else. Often it is the person with the most friends in common. But look at social dispersion instead, eg., your sports friends, your school friends, your family friends, your work friends. Dispersion is who is connected to the most of your groups – that is the spouse about 75% of the time. the other 25% is a good indicator that youre going to break up your relationship
- can predict postpartum depression from social media content.
- how easy is it to get data?
- takethislollipop.com – makes a video out of your social media accounts, try it if you think all your settings are private
- we leak a lot of data, we don’t realize how much data apps are pulling
- it’s not all bad – apps recommend products you might like, google uses it to make the web easier to interact with
- we don’t have control over it – don’t know who has data or what they’re doing with it
- even though the algorithms are smart, we still don’t buy whatever the algorithm says we might like. but they tell us things we would never think about but we actually want
- treat your business algorithms as useful but realize some are wrong or shouldn’t be acted on, it’s one more piece of advice
- [jen needs no slides, she changes every 5 or 10 minutes just for the heck of it, impressive]
- how do you figure out which people like the targeting or don’t like the targeting
- you don’t want to be the creepy facebook stalker guy, it needs to be handled with care
- Hillary or Rand: Trusting the crowd with your vote
- 69% of US adults say that Cost Of Living is an important issue
- 52% of US adults say that Crime is an important issue
- 58% of US adults say that Terrorism is an important issue
- Writing the Lyrics of Qualitative Transformation
Sharing is nice:
I came across this interesting infographic on the e-strategy trends site. It made me think about the types of selfies that I take. Most are at conferences with other attendees and I rarely have a product in hand. But, our clothes are visible, our smiles are visible, and someone’s logo is likely to be in the background whether it’s a vendor’s logo or the conference’s logo. Social listening researchers have been waiting for a tool like this for a long time and it looks like we’re finally inching forward just a little bit quicker now.
— Annie Pettit (@LoveStats) September 19, 2014
— Annie Pettit (@LoveStats) September 19, 2014
— Annie Pettit (@LoveStats) September 19, 2014
— Annie Pettit (@LoveStats) September 19, 2014
- What’s Gonna Kill You? An Infographic That Actually Works #MRX (lovestats.wordpress.com)
- In Honor of Infographics. #MRX (lovestats.wordpress.com)
- Frame this Gorgeous Mercedes Benz ad on your living room wall #MRX (lovestats.wordpress.com)
Sharing is nice:
If you subscribe to the Myers Briggs theory of personality, introversion/extroversion is a personality trait that looks like a normal curve. Half of people are introverts and half of people are extroverts. Indeed, this trait is a continuum such that people aren’t simply one or the other. Everyone has some degree of extroversion and introversion, and the ratio differs from person to person. But let’s press on.
How is social media killing social? Well, people are spending more and more time pulling those little electronic devices out of the purses and pockets and typing on them. Instead of meeting up with their friends, going to a coffee shop, and chatting in person, people are simply pulling out their phone and texting, or tweeting, or facebooking those conversations. And instead of lengthy, well-laid out conversations, they’re sending out quick sounds bites and pithy remarks all with the intent of getting a chuckle out of someone else sitting on their couch waiting for a pithy remark to respond to.
Is this reality? Maybe. If you’re an extrovert. But if you’re an introvert, it could very likely be a completely incorrect picture. You see, introverts don’t require the constant noise of five people talking over each other to entertain themselves. Introverts put up with all that noise because half of the world is extroverted and they don’t know how to be quiet for five minutes. Introverts have no need to constantly flap their lips and make noise and listen to themselves talk. When possible, introverts will simply avoid all that nonsense. It’s easier, it’s quieter, and it doesn’t create unnecessary white noise.
If you think about it, social media creates a quiet place where people can talk to each other without interrupting five other people who are also trying to fill their air with their noise. Social media creates a space where introverts want talk. A place for people who would have never talked in a crowd before can now share their opinions in a calm and quiet place. Social media creates a place for people who might have never shared their opinion before to share their thoughts with many different people. Social media is a tool that just so happens to increase the sociability of introverts. A place for them to meet people they would have never met otherwise. A place to make friends with people they would have never met otherwise.
Is social media killing social? Maybe if you’re an extrovert. But absolutely not if you’re an introvert. And the world could stand to listen to the other 50% of people for a change.
Sharing is nice:
What makes a great webinar, you ask? Well, of course there’s a top ten list for that!
- Write a script: Don’t you dare read an in-person presentation. No! No! But with the tight timelines of webinars, you can consider this option. It will ensure you stay exactly on time and cover every important point. But re-read points #2 and #3.
- Engage don’t read: It’s ok to read a script but it’s not ok if I can tell you’re reading. Read as if you were actually talking. Keeps the ums and ahs that you would normally say, though perhaps cut back on them if you’re an abuser.
- Voice modulation: Nothing makes people tune out quicker than a monotone voice. Make sure your voice rises and falls and pauses and speeds up as you normally would. If you can’t fake it, then bring some colleagues into the room and present your webinar to them.
- Don’t go under time: Don’t run short. Webinars are often only 30 minutes total and you should use up the entire 20 minutes of speaking time (allowing for 10 minutes of questions). 15 minutes feels too short and almost like I shouldn’t have bothered to break up my schedule for you. 20 minutes feels like you wanted to give me as much as you could.
- Don’t go over time: Your audience was kind enough to loan you 30 minutes, not 32 minutes, and certainly not 35 minutes. I don’t care how excited you are about the questions coming in. Close on time. You can follow up on individual questions and answers by phone, email, twitter, or otherwise.
- Budget time for each slide: Assume your system will take 1 to 5 seconds for viewers to see your slides change. Make sure your talk budgets 5 spare seconds on either side of each slide. In other words, don’t rush through 100 slides in 20 minutes. A good guide is no more than 1 slide per minute.
- Forget the transitions: Fancy slide transitions and builds are great for in-person presentations but are fodder for failure in a webinar. Every image change adds technology time to transfer the image across the wires and adds one more permission point for glitches. Simple=Fewer problems.
- Make a point: Again, assuming technology will glitch, make sure a written point is made on each slide. Pictures are pretty but when your phone cuts out for 10 seconds, the entire point of a slide can be lost if there is no guiding text.
- Have a back channel: Whether it’s Twitter and a hashtag, your online community, or the webinar system itself, give the audience a back channel to talk among themselves. Let them talk to other people who are just as interested in your talk as you are. It’s networking and education all in one.
- Get feedback: If you can’t bear to listen to or watch the video afterwards, ask your colleagues to be bluntly honest. Could they tell you were reading? Was your voice interesting? Did you time the slide transitions and your words well?
Sharing is nice:
It’s very easy to stick your foot in your mouth when you go online. You can get caught up in the moment and before you know it, you’ve said something you wish you had never said, something that you’ve never imagined you could say. With that in mind, I have a few rules that I try to follow. I’m sure that on occasion I’ve slipped, but here is what I strive for.
- I won’t use profanity or crude language. I have used a choice word now and then but to write it in the online space where it will live on in perpetuity doesn’t work for me.
- I won’t name people on a website if they aren’t already using the website. In other words, if a friend doesn’t use Twitter, then I won’t talk about them on Twitter. The only exceptions are very public people, like Barack Obama, Pete Cashmore, or Katy Perry.
- I won’t criticize individual people online. Ideas, yes. Companies, yes. Organizations, yes. People, no.
- I won’t release personal details of other people online. That includes other people’s email addresses, the city they live in, names of their kids.
- I won’t share pictures of children that people make fun of, even when no name is attached to the photo and the child is seemingly anonymous. The kid probably didn’t give permission for the photo, they’re not old enough to give permission to use the photo, and they’re not old enough to understand the consequences of distributing the photo. (Think memes here.)
- Twitter etiquette: what you can and cannot say (telegraph.co.uk)