The chart above illustrates the emergence and resolution of a flame war[1] on Waitrose’s Facebook Page last November.

The horizontal axis represents sequential Posts on Waitrose’s Wall while the vertical axis represents the individual contributors to the "conversation" (really it was more of a barney than a conversation.) Each blue dot plotted on the chart represents at least one comment posted by a specific contributor on a specific post.)

So the more blue dots in a column mean the more unique users have commented on that post; the more blue dots in a row, the longer that unique user has continued engaging with the overall conversation (or to put it another way, the greater their appetite for the fight.)

The flame war in question more or less dominated Waitrose’s Facebook Page for more than a day and a half; accounting for 70% of all Posts and 72% of all Comments until it finally ran out of steam.

Much as I’d enjoy going into them, the ins and outs of the matter have little bearing. For the sake of this post, I’m only interested in what the numbers tell us about how Page Admins should deal with these emerging crises when they appear on their Facebook Walls.

Because, as it turns out, the accepted wisdom may be misleading.

Accepted Wisdom

Both Crisis Management and Social Media experts will tell you that it’s essential that a brand under attack does all it can to take control of a crisis situation as soon as possible. As one commentator on the Waitrose incident wrote:

The key to successfully managing an issue and ensuring that it doesn’t become a crisis is to keep communicating.

(Sarah Williams, Senior Lecturer in Public Relations, MMUBS)

This is relatively easy to say, but in practice, it’s often hard to identify emerging crisis situations in Social Media platforms.

What did Waitrose do?

As it happens, Waitrose Admins did comment on one of the early posts in the sequence.

For most of the sequence, though, they were more notable by their absence. After apparently ignoring the situation Waitrose admins finally weighed in at the end of the day with three identical posts in rapid succession:

Why did Waitrose post the same message three times?

Here’s an apposite quote from Stuart Bruce speaking in Communicate Magazine:

The volume of comments on the page makes it extremely hard to follow. One sensible thing Waitrose is doing is to repost its response, as if they did it just once and left it then it could be easily missed.

Stuart is one of the smartest chaps in this space; but he’s wrong here. I’m not trying to show him up, instead I’m using his quote to illustrate how easy it is to get Facebook Community Management wrong. It should make us all feel better to know that we’re in good company.

Why is Stuart right?

Let’s look at how he’s correct first. There was a lot of activity on the Page. The Facebook Wall is strictly chronological, and each time Waitrose posted the message it was pushed down and off the screen by subsequent user posts. By re-posting the message, they maintained a dominant role on the Page[2].

Why is Stuart wrong?

Here’s that chart again.

This time I’ve highlighted four areas. You’ll recall that the the vertical axis represents the individual contributors — the gradient represents new contributors joining the conversation. Together, the four highlighted areas represent 95 new contributors adding their voices to the mix.

The impact of the admin Posts outweighs the impact of the user Posts by an order of magnitude, as illustrated by the following chart:

In this particular instance, there were a mere 4 admin posts compared to the 65 posts from users. But those admin posts had a disproportionate impact on drawing new people into the conversation (more than 12 times as great: an average user post attracted 1.88 new contributors, an average admin post attracted 23.75 new contributors.)

New contributors have more "energy" for the fight (and will revive and rehearse old arguments and conflicts) so I think we can say that on the whole new contributors to a flame war are bad no matter how good their intentions.

You’ll recall that the rows represent the continued presence of a given contributor in the ongoing debate — so we can highlight the legacy of those initial three admin posts:

Why does this happen?

Facebook Pages are inherently asymmetric.

1. People interact with content in their News Feed and rarely visit Pages (for many of our clients, the figure is as low as 0.5% of Active Users.)
2. Posts made by an admin will be sent out to appear in a proportion of the Page’s Fans’s News Feeds[3]; reaching (in Waitrose’s case) tens of thousands of users.
3. Posts made by Fans will only be seen by visitors to that Page, and a (vanishingly tiny) proportion of their Facebook friends’ News Feeds.

Taking this into account, let’s retell the Waitrose Flame War story as follows:

A small number of people had a fight on Waitrose’s Facebook Page. None of Waitrose’s customers was really aware until the end of the day, when Waitrose admins broadcast three identical messages to as many of their Fans as possible.

What should they have done? What should I do if a Flame War breaks out on my Page?

What’s the lesson here? I’d say that — when managing Facebook Pages, concentrate on the News Feed, not the Wall.

Here’s what I’d suggest:

• Do not post Status Updates unless you want to reach a wider audience than is already aware of the issue.[4]

• Instead, comment on every new thread — even if it’s only a holding statement. These comments will be seen only by those who are already involved. Bear in mind that – as threads expand beyond 50 comments, Facebook pages the output and that as a result earlier comments may not be seen by later viewers.

• To avoid copy-and-paste fatigue, you should to maintain and link to a single page with the latest details. A blog would be a good place for this sort of content[5]; it really doesn’t need to be on Facebook. Whatever you do, keep it away from the Wall (publishing a Note, for example, would be an error.)

• Consider switching the default view to ‘Only Posts by Page’; but monitor carefully: it is possible that users may begin to add their complaints as off-topic comments on existing brand Posts.

9 days of activity on the ASOS Facebook Page

This chart (click for bigger) represents 9 days of activity on ASOS’s Facebook Page. Compare it to the Budweiser flow and you’ll see how focused ASOS is on Customer Service by comparison.

And yet their post frequency is high: an average of more than 3 posts per day. They promoted 11 individual links during the period which between them delivered over 120K clicks through to the ASOS site.

So Customer Service is only half the story; there’s a robust DR element here.

Here’s a brief summary:

Automated sentiment analysis uses a combination of two approaches to determine sentiment: lexical analysis (which looks for emotionally-weighted words) and machine-learning (which relies on a “training corpus” of manually scored documents to predict the emotional content of new documents that it processes.) Continued human intervention (“training”) in the machine learning process may improve results, but makes it more or less meaningless to compare the results over time.

No two sentiment analysis tools on the market agree closely; with most disagreements occurring and around “sentiment-neutral” comments. Since these form the bulk of the content, there is much room for disagreement.

Sentiment analysis on short content (e.g. Tweets) lacks sufficient context for accurate judgments, whereas analysis of longer content often lacks sufficient relevance (e.g. the search term may only be mentioned in passing, and the sentiment score refer to a different object)

Human-based/manual sentiment analysis also faces reliability challenges. There is often as much disagreement between two human analysts as there is between two automated systems. Worse still, research demonstrates that the same person – when presented with the same text on different occasions – may score it differently each time.

Conclusion: Sentiment Analysis may be used to guide customer service engagement, but should not be used as a KPI

Have a glance at this chart showing three years of search traffic (click for a live version) 

In previous years, search traffic has more or less followed the announcement – but for the (still mythical) iPhone 5, search traffic began to grow well ahead of time (a situation — as Scott Thompson has pointed out to me — probably only exacerbated by the delayed announcement.)

Bloggers and journalists are increasingly using social signals (search & twitter volume) to determine their editorial policy. After all, search and social are a good indicator of what interests your audience – and major traffic drivers.

In this case, however, the process created a vicious feedback loop. Even the more sensible commentators and analysts found that — when the conversation was about the iPhone 5, there was no virtue or value in writing about it in any other way.

I used Google to count the posts that mentioned “iPhone 5” in the title on some of the top tech blogs, MSM and online news sources — the numbers tell a story.

The thing we see on all the brand Facebook Pages that we’ve analysed so far is how much the conversation is controlled by the Page Admins.

Recently I was invited to talk about these at Warc’s “Social Media: Beyond the Hype” conference and at Webit Congress 2011 (Webit is the digital industry’s get-together for CEE. It’s huge: very energetic, very exciting and great fun.)

Below are the slides, and below that, my script. I know that it’s hard to read the two together. Actually, someone told me that I mumbled a bit during the presentation (I wasn’t used to my Madonna-style headset mike to be fair) so the script is for them (and for anyone else who was too polite to complain.) You can download the script here.

NB: I write my scripts (like all my stuff) in a kind of lazy version of Markdown. This works great, but can make for a fairly ugly layout.

Script

0

Ask audience to hold up iPhones, iPads, MacBooks.
How many of them are fans of Apple?

1

So there’s this social media revolution going on around us. Everything has changed. We need to change the way we think about everything.

Only this doesn’t help advertisers allocate their budgets. And if you’re an advertiser or a planner, the way you allocate your budgets will play a big part of your ultimate success or failure.

Everything may have changed, but we still need to be able to compare apples with apples.

We can’t just pretend that social media valuations exist in a vacuum. We need to find ways that we can relate them back to familiar metrics.

One of the biggest questions that we have to deal with is “what’s the value of a fan”? This is the most divisive questions in Social Media planning.

2

Let’s take a step back. What do we mean when we say “Fan”?

3

Here’s a definition from Kevin Kelly – from a brilliant blog post that you should all read when you have time.

Read Quote

That’s instructive, but unless you’re a musician or an artist that’s probably not relevant.

4

What motivates people to become a fan of something? Briefly, fandom helps us tell a story about ourselves, about who we are and where we belong. They are badges we adopt to identify ourselves as individuals, and as part of a group.

5

Of course, there’s a whole spectrum of fandom. What you should notice here is that most people just don’t care very much.

Another way of looking at it is that you can only reasonably be a fan of one or two things. Otherwise you’re not really a proper fan, are you? For example, how many football teams could you support and still call yourself a supporter?

This is interesting for marketers; we’re mostly used to targeting and engaging with the fat belly of the distribution curve, and not the mad bits at the edges.

6

Here’s an interesting thing. Those people with the mad love create the space for the ordinary fans to exist. The more mad fans you have, the more you’ll find that your average fan population grows.

You could argue, for example that the existence of a few hardcore Apple fans makes it possible for lots of other people to call themselves Apple fans. After all, no-one calls themselves a Dell fan, or an HP fan.

Have you noticed how people will take their iPhones out in meetings or at a bar or restaurant and put them on the table? No one does that with a Nokia.

So these Superfans create the space for this behaviour.

It would be good if we could create a measurement for that; if we could understand the correlation and causality.

7

This is all very interesting, but that’s not what I’m here to talk about. I’m not talking about True Fans or Superfans, but about Facebook fans. Statistically speaking, Facebook fans aren’t True Fans or Superfans. So why am I talking about Facebook? What’s so important about Facebook?

8

Well, for one thing, Facebook is very big. It’s currently valued at around $100 per active user, and it has around 800 million users.

That whole “value per user” is all very late nineties, isn’t it? All very dot com boom. A bit devalued.

But Facebook’s numbers are BIG

9

Just looking at Europe, Facebook accounts for 85 BILLION European user minutes per month. If you were to go back in time 85 billion minutes, then the Neanderthals and wooly mammoths would still be around. Every month.

10

And those users consume 163 billion page views.

That circle there is Google. And that one is Vkontakte.

11

So brands are understandably interested in how we capture a share of that audience attention.

About 4 years ago, Facebook introduced advertising and brand pages.

They told their users “we’ve built a way for you to connect to things other than people”. What they meant was “now you can tell people who you are based on the things you like.” Fan Pages were originally intended to be a bit like t-shirts, mugs and tattoos.

Brand Pages are free to build. When Facebook introduced them, they effectively set brands’ Facebook marketing strategy for the next 4 years.

We were going to make people Like our Page.

12

So now, four years later, Brands are beginning to ask this question: “I’ve got all these fans. Which is great. But what are they worth to my business as a marketing asset?”

13

Now I could just tell you that every business case is different.

This is a great answer that you’ll hear from time to time. You can’t fault it. It’s true.

But it’s not very instructive, is it?
If I tell you this at any point in the presentation or the Q&A you have my permission to throw stuff at me.

I want to give you some framework for thinking about and answering the question; three simple general approaches. And I’ll try to tell you how to make the most of each of these approaches.

14

The Pages may be free, but it costs money to recruit fans and it costs money to maintain that community.

So we know what our cost-per-fan. That’s fairly simple.

15

But when it comes to public Value of a Fan studies, there’s a massive discrepancy. How can this be?

Actually, let’s get rid of these two — even though analysts refer to them as “Value of a Fan” they’re actually just sales referral values; nobody became anyone’s fan in those calculations.

16

When we started in this space, we asked our clients why they wanted fans. Here’s what it boiled down to. There are some good reasons here; and we shouldn’t dismiss the top one.

17

We can reduce those reasons even further to this short list. We want to value our fans by:
- the value of the Earned Media that they deliver
- the value of the sales that we can directly attribute to our fans
- and how recruiting fans improves brand measures like preference and likelihood to purchase

18

Let’s look at these in more detail

19

Here’s that Earned Media Value model that Vitrue published last year.

The calculation’s fairly straightforward – But I won’t waste time reading it out to you — it’ll be in your packs.

20

Instead I’ll touch on how that model falls apart.

Most importantly – brand posts don’t reach every fan. We find that only 5-20% of fans receive the impression.

When we’re planning media campaigns, we try to cap frequency. That avoids wastage and burnout. We’ve no control over how often Daily Average Users are exposed.

That’s OK, but it’s hard to value those subsequent impressions as much as we’d value the first ones.

Are 20 impressions against 50K users the same as (say) 2 impressions against 500K?

I’d like to take you on a bit of a detour now to look at how this works.

21

This is a very simplified model that shows how fans interact with Facebook Pages.

This is taken from real client data.

And here are some real client numbers.

What you need to see here is that users see and engage with Brand content on their News Feeds. They don’t really visit the Page very often.

And one of the key factors controlling whether they see that content is timing.

Are they online and active when you post your content?

How many other voices are competing for their attention? That’s partly other brands, but most of the competition comes from their Facebook friends’ photos and status updates.

22

So we pay particular attention to timing. Here’s some before and after charts that show how one of our clients has responded; We found that audiences were more active, and more likely to engage at weekends — so they’ve upped their posting activity on Saturday and Sunday.

Timing isn’t everything; it’s only half the story.

23

Facebook uses an algorithm called EdgeRank to try to make the stories that appear in your News Feed as relevant and interesting as possible. I hope you’re all at least vaguely familiar with this, but if not, I’m happy to chat about it afterwards.

Here’s the thing: Facebook owns the relationship between you and your Fans. That relationship lives as a record in THEIR database, not yours. And Facebook decides what’s interesting to your fans and what isn’t – not you. Facebook owns the content that your fans post on your Wall, and the rights to anything that you post there. If you haven’t read the Terms & Conditions you signed when you created your Page, now might be a good time!

For now, the thing I want you to understand about EdgeRank for now is that — if users don’t engage with your content, EdgeRank will – over time – show them less of your content, till they no longer see any of your content at all.

That’s how you can have 1m fans, but only reach (say) 200K of them.

24

The Vitrue model assumed that Pages were posting twice a day, every day of the week.

Our own research shows that our clients’ posts have an active window of engagement between 3-6 hours long.

That implies that we could profitably increase impressions by posting twice in a day.

And if we were chasing Earned Media Value, then more impressions = more value.

So there’s a lot of pressure to increase frequency.

What might the impact of that be?

25

Our research shows that it has a strong impact on impressions; if we double our post frequency, we more than double our impressions.

Reach grows too; less strongly, and there’s an upper limit – only a certain portion of your audience will be online in any given day.

26

But unsubscribes also increase. If we increase post frequency, we have to be prepared for fan attrition.

27

Here are our big problems with the Vitrue model;

• This is incremental media value. It’s not like we’re saving media budgets. So we can’t count that as real ROI
• Fan audiences are small relative to our total target audiences.

Even Coca Cola only has 35m fans — and in theory they have a global audience of 7 billion people. So that’s 0.5% of their audience they can reach through Facebook

Of course, because of EdgeRank they can only reach a minority of those fans.

And how much does Coca Cola NEED to advertise to people they call Fans? How much more Coca Cola can these people drink?

• Thirdly, impressions aren’t a proxy for incremental revenue growth. Media impressions need to be carefully planned to have any value. One impression is not the same as another.

28

That’s the BAD news. Here’s the GOOD. Social Media is all about taking advantage of homophily and social proof. What’s that?

29

We all prefer to hang around with people like ourselves — that’s homophily. You instinctively prefer the company of people who share your values and beliefs and preferences.

30

And conversely, a lot of our decisions are based on unconsciously copying our friends, their values, beliefs and behaviours.

This self-reinforcing behaviour is very important for advertisers.

31

Because it allows us to find more people like our fans. And it allows us to use social cues from fans to make our messages stronger, more noticeable.

For example, targeting friends of fans with increases our media efficiency by around 20%.

And running Facebook’s sponsored stories (where the message comes from the friend, and not from the brand) increases efficiency by between 300% and 400%

32

Here’s the output of some famous Nielsen research:
- Facebook ads improve brand metrics, but not as much as Facebook Ads that contain a friend’s name. It seems that peoples’ eyes may naturally be drawn towards their friends’ names, and dwell there a little longer.

And if there’s an organic impression, a spontaneous, untargeted earned impression, that seems to improve brand measures further.

(11 mins fluent)

33

So here’s my advice about how to make the Earned Media Valuation work for you;

• Don’t focus on fan growth at any cost.
• But DO focus on the fundamentals, and on what makes social interesting. Look at how you can encourage sharing.

SHARING is very important. Focus on fans as a means of reaching friends of fans.

34

Let’s take a look at the Direct Response value.

35

After all, survey data shows that a good proportion of fans Like a Page so that they can be alerted with news and deals. We should be able to convert some of that interest to value.

36

Here are some casual examples – not from my clients. Tesco regularly posts deal links on its Facebook Page. They seem to drive a lot of traffic; some of that has to convert. So Tesco can begin to work out some kind of Average Lifetime Value for their Facebook fans.

Obviously it has to be Average Lifetime Value because Facebook owns the relationship; Tesco can’t segment or target. But it’s still good.

37

Here’s a lovely example. ASOS, an online clothes retailer, got 130K clicks in a single day when they ran their live sale last weekend. That’s pretty impressive.

You’ll see that they were offering dresses for £1 – but believe me ASOS aren’t in this business to be generous fools. You can bet they were making money off the back end.

38

Here’s how you make Direct Response values work for you.

Track everything.

Make sure that you use unique links, and that those links pass campaign data to Google Analytics (or whatever site analytics package you’re using.)

Never use the same link twice. Track source, message, campaign.

And make sure that you continue to optimise your Earned Media metrics.

Focus on being able to compare your performance here to other non-social channels. How does your fan base compare?

39

Of course, if you’ve not got an e-Commerce (or a publishing) business, this isn’t going to mean anything to you.

You’ll want to focus on brand measures.

40

1-to-1 engagement is great. Here’s an example from some activity we ran for our national bakery client, Hovis earlier in the year.

That personal touch is great; it’s really meaningful, and has tremendous impact.

Only no-one sees it.

I’d argue that this is a customer service role. Customer service is a brand attribute, it’s part of the product that people offer. Tony Hsieh at Zappos built it into his business; everyone in the company has to spend some time on Customer Service. As a result, he was able to cut back his marketing costs massively.

Customer Service is great, but it’s not what I’m talking about.

41

So we’ve been running a series of major surveys across Brands “Leased Spaces” — Twitter, YouTube, Facebook.

And what we find confirms others’ research: people who engage with a brand show improved measures against brand metrics compared to people who only visit those spaces.

Which sounds like common sense; and which continues to leave us open to the question – “Is this a causal relationship? Do they engage because they like you, or do they like you because they engage?”

We’ve got smart people working on that question, but we have access to a level of granularity that’s useful when it comes to planning activity.

42

And that granularity tells us that what we REALLY want people to do is share our content or recommend our product.

It’s about making a public declaration.

Psychologists will tell you that people want to behave in a consistent manner (it’s all to do with preserving a sense of self.)

Previous behaviour isn’t just a predictor of future behaviour: if I have committed to something on a small scale, I’m likely to eventually commit to that same something on a larger scale in order to be consistent to myself.

So asking people for a small public commitment is a fairly secure method of confirming their brand preference.

Essentially we can create fans simply by asking them to identify themselves as fans.

This is social media at its BEST.

43

What do you need to make the Brand Metrics Valuation work?

You need to invest in market research. And you need to be able to compare your social media research with your press and TV and PR market research. Otherwise, I’d argue, you’re just taking a shot in the dark.

And – again – you need to optimise your earned media metrics; with a strong focus on how people are SHARING your content and messages.

44

Now — just to make things a bit interesting, I think it’s worth pointing out that someone doesn’t need to be a Facebook fan in order to share your content.

These are four examples of brilliantly shareable content. They all use the same method, more or less; they allow the user to make the story about themselves, and not the Brand.

Remember those tattoos from the beginning of the presentation?

Remember Facebook’s initial reasons for allowing users to connect to Brands?

To tell stories about THEMSELVES Not to tell stories about the brands.

Mugs and T-shirts and Tattoos.

These four examples all make great use of Facebook. And only one of them (the Aviva case) insists that you become a fan first.

They all care more about the Share. About the Brand action.

45

So where does this leave us?

46

Let’s take a look at this survey data again. We should notice that only two of the reasons given are “Truly Social” — the others are a bit self-centred.

Arguably, the other reasons could apply equally well to (say) subscribing to an email list. It’s all old-think. Not new-think.

These people are subscribers; only they’re subscribers whose data you don’t own.

Look at the Terms & Conditions of your Facebook Page. Facebook – quite rightly – owns the data, and the relationship.

Through EdgeRank, Facebook controls which of your fans see your content, and which don’t.

Should this make us think?

47

On the left is the traditional Social Media funnel. If we follow that, we recruit people in at the top of the funnel, keep them engaged, then occasionally try to sell them something.

On the right is an alternative way of looking at things. What if some brands are making a mistake with all this fan stuff? What if we unbundled the user journey?

For example, what if we DIDN’T insist that they became a fan before they engaged with our content, or bought our products, or recommended us to a friend? What then?

Here’s the big question – does every brand need to maintain a day-to-day relationship? Because in this competitive world, in a world where your Facebook Page automatically becomes a new customer service channel that needs maintaining in public, and with EdgeRank breathing down your neck, day-to-day is what it takes to maintain a Page

48

Here’s more for you to think about:

Facebook has changed almost beyond recognition since they introduced Pages 4 years ago. Remember that graph? Back then they only had 50m users. Not 800m. Back then, MySpace was worth $65bn. Facebook was barely a fraction of that.

I’d argue that Facebook has begun quietly to change the rules. And smart marketers need to change along with that.

49

Here’s how Facebook used to be: you had all your leased spaces, Pages, Places, Apps INSIDE Facebook, where your audience was.

Today, Facebook provides a platform that lets you connect Facebook to your Owned spaces OUTSIDE Facebook.

You can pass data and communications back and forth between the two spaces. You don’t even need to have a Facebook Page to make Facebook work for you. In fact, if you’re thinking about Fan recruitment, that may even be holding you back!

50

So it’s not about building your brand on Facebook. It’s about building Facebook into your brand’s products, services and communications; and about understanding how you tie your Paid and Earned Media together.

51

Before I dismissed the valuations from EventBrite and Spinback because they weren’t really about fans.

What they are is a valuation based on incremental sales delivered through the Facebook sharing mechanisms built into their sites.

These are reports of REAL money being made by REAL businesses, not interesting models and theory.

Was I wrong to dismiss these?

52

I’d like to leave you with a thought.

Is a fan someone who likes you on Facebook?

Or is it someone who engages with your content?

Or someone who buys your product?

Or someont who promotes your content

Or your product?

Or is it someone who by using your content to promote themselves becomes themselves compelling content for your campaign?

Why not ALL OF THESE!?

Here’s what I say. A fan isn’t worth ANYTHING unless they DO something.

Focus on what you want them to DO.

You define and control the relationship. You decide what the value is.

Make sure that you’re in control, and not some 4 year old model of how we might do business in the future.

I’ve spent what seems like a great deal of time tinkering with a combination of open source and homespun SNA tools to answer interesting-sounding questions about Twitter social networks – but this seemed like a simpler method would suffice. And it does; I thought I’d note it down partly as an aide memoire, and partly in case I needed to tell someone else how to go about it.

It does rather assume that you’re using a Mac, or something else with a Unix-like environment. There may be ways to do it in a Windows environment that are just as easy, of course.

Collect your data

Twitter’s followers ids method offers a fast and straightforward way to grab a list of all followers* of an account. You don’t get the usernames, just the numeric IDs.

The format is http://api.twitter.com/1/followers/ids/mediaczar.xml (obviously, you’ll want to substitute your own username.)

So from the command line you can do this:
curl http://api.twitter.com/1/followers/ids/mediaczar.xml -o mediaczar.xml

Do this for the two accounts you want to compare.

Clean your data

I use BBEdit as my text editor. It’s total overkill for my requirements, but I’ve been a customer since the early nineties, and I can’t get used to not having it. The following instructions will work in any text editor. For each file:

1. Clean the <id>...</id> tags; I use the regexp <[^>]*id[^>]*> to search and replace these because it will pick up the <ids>...</ids> tags that wrap the list as well – although this isn’t strictly necessary as they’ll be removed in a later process
2. Sort the file (BBEdit has a “Sort Lines” menu item, but you could almost as easily use the sort command

Find the common elements

One of many useful Unix commands, comm can select or reject items that are common to two files. Here, we want to select them, so we ignore columns 1 (followers unique to username1) and 2 (followers unique to username2):
comm -12 username1.xml username2.xml > commonfollowers.txt

Finished

It’s the sort of thing that could be plugged into a shell script pretty easily; something like this works:

#/bin/bash
echo "getting common followers for $1 and $2\n"
# Collect your data (and do a little cleaning)
curl http://api.twitter.com/1/followers/ids/$1.xml | sed 's/<[^>]*id>//g' > $1.txt
curl http://api.twitter.com/1/followers/ids/$2.xml | sed 's/<[^>]*id>//g' > $2.txt

# Sort your data, then remove the original files
sort $1.txt > $1_sorted.txt
sort $2.txt > $2_sorted.txt
rm $1.txt $2.txt

# Find the common elements
comm -12 $1_sorted.txt $2_sorted.txt > common_followers.txt
rm $1_sorted.txt $2_sorted.txt
echo "done\n"

Another may be our unholy desire for traffic. After all, I’m writing this blog post in response to today’s Nestlé-Greenpeace-Facebook storm (if you’re coming to this story late, Sam Ismail’s post offers background). More to the point, at least four bloggers cleverly promoted their coverage of the events in the Facebook comment stream.

Several social media mavens got so involved that they joined the side of the protesters against Nestlé: twittering animatedly, changing their avatars, publishing lists of brands owned by Nestlé (and therefore eligible for boycott), or by referring the community manager to the Cluetrain Manifesto or Gary Vaynerchuck videos.

Examples of similar foul-ups pepper our presentations, mine included: I’m presenting on crisis management on Tuesday. All my case studies come down to two things:

• people inside your business fucking you over through ill-will or ignorance; and/or
• people outside your business fucking you over through ill-will or ignorance

Put that aside. As an industry I’d argue, it can seem that we’re obsessed with failure. And we’re nasty with it. Other than politics or the pro-wrestling circuit, I can’t think of another industry so ready to criticise one another in public. At each new crisis, the web buzzes with our speculation, exaggeration, misinformation, chinese whispers, and rumourmongering of which the events of today were just an example.

Social Media Pogroms

Why are social media gurus so ready to criticise each other in public? I suspect that our readiness to criticise the brands and agencies for whom (under different circumstances) we might work is in fact an artefact of social media. This is a an example of social behaviour. Compare it to mob behaviour. Playground behaviour where netiquette is ignored. We want to join in the fun; even when the fun turns out to be kicking someone when they’re down.

Karma

Kerry Gaffney and I once came up with a rule-of-thumb: “don’t be too quick to jump in because there but for the grace of God go you.” It sounds sententious, but we reckoned that it would pay off in time; people would, we felt, be less likely to jump on us when we inevitably #failed.

Here’s what I think: until we can grow up and leave the playground, we won’t be able to help our clients gain the confidence they need to make the right decisions in this area.

Individuals should try to come to terms with the fact that they work in marketing; and stop trying to moonlight as anti-globalization anti-corporation cowboys. Learn to love what you do, for fuck’s sake.

And I think we should go a step further. We should stand up for each other. Only a very few (and highly honourable) people stood up for We Are Social when they took an unnecessary pasting over Eurostar. I’m ashamed to say that I wasn’t one of those; while like many others I left messages of support on Robin’s post, I didn’t help to fight fires on TechCrunch or elsewhere.

And that brings me to the last idea: where we can, we should probably try to help clients in distress — even if they aren’t our clients. That might help our industry look professional and responsible. It’s a bit of a big ask: “why,” you might ask, “should I spend valuable business cycles helping someone else’s client?” But wouldn’t it be more constructive than spending those same few meagre cycles adding fuel to yet another social media fire?

Post Script

Incidentally, in the unlikely event that you’re interested in my take on the Nestlé situation: a bunch of social media gurus and pre-existing anti-Nestlé activists probably won’t do too much damage to the brand. Anything to do with Facebook has a whiff of slacktivism to me. Within a week or so there’ll be another stinking piece of Social Media carrion around which we can snarl and posture. Also, a campaign to knight Eddie Izzard seems to have more traction in the non-guru social media world.

After my first foray into using bonuses to engineer better results, I found that I needed to pay over a hundred bonuses. It rapidly became clear that paying these using the web interface would be nearly impossible, forcing me to look at the command line interface tools a little faster than I’d been planning.

Amazon provides a pretty full toolkit — but right now I’m only really interested in the grantBonus.sh tool. This meant I could (fairly quickly) knock up an improved workflow.

1. Download all completed HITs as a CSV using Mechanical Turk’s Export Results button;
2. Use a spreadsheet to approve and reject each HIT manually (need to improve upon this);
3. Export the sheet as a CSV and use Mechanical Turk’s Review Offline button to load it back into the system and automatically process the basic payments;
4. Append bonus values in a new column;
5. Export the sheet as a tab-delimited file;
6. Use a perl script to build a list of shell commands (I’m guessing if I were more comfortable with shell scripts I could skip this step); and
7. Run that list of shell commands

Not – of course – as simple as it could be. Each stage could be automated more; and at some point will come the moment when I don’t need to use the web interface at all.

Anyway — here’s that perl script.

#!/usr/bin/perl

# A quick & dirty script to make paying bonuses on
# Amazon's Mechanical Turk less of a nightmare
# Requires a tab-delimited export of Workers'
# results edited to include a new "Bonus" column
# appended at far right
# Also requires MTurk CLI tools from http://bit.ly/cUhscK

# author: Mat Morrison
# date: Tuesday March 9, 2010

my $count;

# load data from the file arguments specified at the command line

open(INFILE,  $ARGV[0]) or die "can't open file to read: $!";
print "\tReading from $ARGV[0]. ";

if ($ARGV[1]) { 	#has an output filename has been specified?
	$outfile = $ARGV[1];
	} else {		#if not...
	print "No outfile specified - reverting to default\n";
	$outfile = "multiplebonuses.sh"; #write to a default filename
	}

print "\tReading from $ARGV[0] parsing to $outfile\n";

open(OUTFILE, ">$outfile") or die "can't open file to write: $!";

print OUTFILE "#!/usr/bin/env sh";

while (<INFILE>) {
	chomp ($_);
	# read the relevant variables out of each line
	# there are lots of columns, but we only need three of them
	# easier to use tab-delimited -- otherwise random commas get in the way...
	my (undef,undef,undef,undef,
		undef,undef,undef,undef,
		undef,undef,undef,undef,
		undef,undef,$AssignmentId,$WorkerId,
		undef,undef,undef,undef,
		undef,undef,undef,undef,
		undef,undef,undef,undef,
		undef,undef,undef,$Bonus) = split("\t",$_);

	# if there's a value in $Bonus (i.e. a payable bonus)
	if ($Bonus > 0) {
		# print the command to the instructions list script
		print OUTFILE "./grantBonus.sh -workerid $WorkerId -amount $Bonus -assignment $AssignmentId -reason \"altruism or overperformance bonus as per http://bit.ly/97c8z5\"\n";
		#increment the counter
		$count++
	}
}

print "\tProcess complete: $count bonuses to grant in $outfile\n";
END

In some ways, the post you’re reading now could be another metapost — a post about first posts. But it’s really about new ways of working.

I know about Melanie’s first post because I’ve been carrying out some quantitative research using first posts. I took a user-generated list of UK PR blogs that I helped curate last October, and attempted to identify the date of the first ever post for each blog.

This is a task that’s almost impossible to automate. Getting the newest post is a cinch for a computer – the oldest post not so much. And yet it’s relatively simple for a human to perform the task – generally it’s just boring and repetitive (although I challenge you to find the first post on Jed Hallam’s blog, Rock Star PR). I’m not one of those people who enjoys repetitive tasks, so I decided to take this opportunity to set up the Magic Bean Lab’s first experiment; to test the efficiency of various alternative labour sources.

Method 1: e-lancers

I used Freelancer.com: a well-established e-lance and outsourcing marketplace I’ve used several times in the past. As we get more used to buying things that we can’t see over the web, the e-lance market has become a no-brainer.

I employed two e-lance researchers. I’ve found that running researchers in parallel on projects like these reduces the need for overmuch error-checking. Quality Assurance (QA) rapidly becomes the biggest overhead in any project like this.

My quick-and-dirty QA process runs as follows:

• Compare results side by side
• If the results agree, accept this as the correct answer
• If the results disagree, do some checking myself.

In the event, the two freelancers agreed in 88% of cases, and I only had to check the remaining 12%.

An obvious problem: if both freelancers agree on an incorrect answer, I won’t check it. This happened in approximately 4% of cases during this test (I’m working with a known set of data here, or — of course — I wouldn’t be able to tell that.)

Another problem which you wouldn’t see otherwise: it’s fairly time intensive. I have to post the project, wait for the bids to roll in, assess the bids and so on. The whole process took about 48 hours from start to finish.

Still, it beats doing the work myself, and it’s fairly scaleable.

Method 2: Amazon Mechanical Turk

I’ve been noodling around with Amazon Mechanical Turk for a few months now. Mechanical Turk, for those of you who don’t know, is named after a late eighteenth century hoax that — while it purported to be a machine that could play chess — was in fact powered by a person sitting inside a box.

Amazon describes its Mechanical Turk service as follows:

Amazon Mechanical Turk is a marketplace for work that requires human intelligence. The Mechanical Turk service gives businesses access to a diverse, on-demand, scalable workforce and gives workers a selection of thousands of tasks to complete whenever it’s convenient.

So instead of paying a freelancer a fee to perform all the research, I can split the job into its constituent parts that Amazon calls HITs (Human Intelligence Tasks) and pay a fractional piecework fee (say $0.10) for each blog on the list. This has two advantages:

1. It’s much faster. Instead of one person working through the list in sequence, the list is processed in parallel;
2. It’s cheaper. There seems to be a lower limit to the bids on Freelancer.com of around $30 per job; and
3. It’s extremely scaleable. Using Amazon’s API, I can embed humans into automated processes. More on this in later posts, I hope.

The only problem? In the words of one Turker I’ve interviewed, “There are few spammers in mturk who spoil the mturk community.” In other words, I have more QA problems. It seems that people are more inclined to be dishonest when smaller sums and greater anonymity are involved. I’m sure that a behavioural economist like Dan Ariely could explain this in detail — but for the moment, please let’s just accept that most people are more inclined to cheat you out of $0.10 than $100.

So here’s the first thing I tried.

• Run three Turkers in parallel
• Accept the earliest date received

Clearly there are problems with this method. But not as many problems as you might think. Results were approximately 80% accurate. And it took less than 2 hours from start to finish: that’s 46 hours faster than the e-lancer method. And remember it’s costing me less, although I’m paying for answers that the method deems to be inaccurate.

But surely I could do better.

Method 3: Amazon Mechanical Turk and “elementary game theory”

To be honest, I probably know less about game theory than you do. I know about the Prisoner’s Dilemma and the Three White Hats (which probably isn’t even game theoretic; only going to show how little I know.)

But I’m using “game theory” both very loosely and very specifically to mean “I know a little about how people try to game systems, and I’m going to do my damnedest to use that knowledge to hack their behaviour.”

So here’s what I told the Turkers:

Each HIT is being performed three times, and the results will be checked against each other:

1. When all three HITs agree, a $0.10 bonus will be paid to all workers.
2. When only two HITs agree, those two will be accepted as the correct answer. The third result will be rejected.
3. If all three HITs disagree the requester may consider rejecting all three.
4. Occasionally, it may be the case that it is very hard for you to find the correct date. We want to make it worth your while to find this. Please note this in the Additional Comments box together with the process you used to discover the date and we will consider paying a discretionary bonus of $0.50 on top of any other bonuses. For particularly hard HITs, therefore, the total potential upside is $0.80

Please, take your time, and get the right answer. It will be worth it for you, and worth it for the other workers performing this task!

My hope was that I would discourage the spammers. Here’s the logic:

Spammers know that they won’t get paid unless at least 1 other person agrees with them. For this to happen, either:

1. They’d have to be lucky enough stumble across the right date by mistake (at odds
   of around 3650:1 for blogs created in the past decade); or
2. They’d need to be accidentally matched by another troll (same odds.)

Incidentally, it turns out this “majority rule” is unpopular with the Turkers; mostly because they fear the spammers don’t read the instructions, and will still queer the pitch (which seems like a legitimate and logical strategy for spammers).

Nevertheless the task was completed in much the same time as the previous method; and with startling results. The answers received were 97% accurate. That’s more accurate than the freelancers (just under 96%), at a lower cost, and in a fraction of the time.

Comparison of the three methods

In the table below, Cost per Response is adjusted for accuracy.