Contextual analysis of tweets made during London riots

We have published a short whitepaper that analyses the tweets made over the weekend of the 6-8th August.  The tweets downloaded were filtered simply on having either the word ‘riot’ or ‘rebroadcast’, since this was one of the words used by the gangs to encourage other gang members to congregate at a particular site.

The analysis can be seen at our whitepapers section, and the results appear to indicate that in the majority of cases, mentions of specific placenames that were being targeted by the gangs were in fact being discussed on twitter some hours before the main attacks took place.  Analysis of the tweets themselves appears to show a mix of people voicing their displeasure at hearing of the proposed riots in their town or community and those actively encouraging it.  This shows that people were actively involved in trying to prevent the riots from happening, and the demographic of those tweeting against the riots is likely initially to be similar to those actively participating, since they are the first to hear about the rumours of the riots taking place – whether through the Blackberry messaging service or word of mouth.

Our analysis is possible since we can simply search in context with the word ‘next’ – as in what is the next target, or the next place to hit?  It is possible that other words could be used to identify other targets, and we will make available in the coming days some of the analysis so that you can search the data yourself, and see what placenames come up using the search words you would prefer to use.

Posted in Analysis, London riots | Leave a comment

True language independence for social media analysis

One issue facing many companies today is that their markets are spread across the world.  Nowhere is that more apparent than in Europe that has a number of countries and cultures, each with its own specific language.

This can make analysis of social media difficult, since translation tools are required for every market that you want to monitor.  For the majority of today’s analysis tools, this is a problem since rules need to be created for each country and language.  This is an expensive and time consuming task.  This means that although for the English language there are a plethora of tools out there, the same is not true for every other language, and for some languages there are no tools at all.

This highlights another difference with Brand Aura’s technology.  Since our tools are not based on NLP, and don’t require rules to be created for each separate language, it is a much simpler task to translate the result into another language, and one that you can understand and make sense of.

In fact our online demo shows exactly that.  You can see this for yourself, and become a truly language independent social media analysis guru by following the steps outlined below:

  1. Open the weblink in a separate browser
  2. Click on link ‘Tweets on financial topics’
  3. Either type in ‘tsb’ and press ‘Go’ or click on the tsb word link in keywords list
  4. Notice that the wordcloud shows what seems to be gobbledegook.  Certainly the words surrounding ‘tsb’ in context do not appear to be English.  To translate the wordcloud into English, scroll to the bottom of the page, and use the ‘Select Language’ drop down to translate the wordcloud into English so that you can understand it.  Note that if you want to translate to a language other than English that you can also do the same by choosing the language you want to translate to.
  5. You should now have a wordcloud that is partially translated – many words that were written in what was actually Indonesian are now translated to English.  You can see the original word by hovering your mouse over any word.  For example, the translation shows that the words ‘meeting’, ‘account’, and ‘committee’ were all translated from Indonesian.  You can click on any of these words to make this word now the centre of your analysis, to investigate further what is being said by the Indonesian community.  Notice that this will analyse the underlying Indonesian word and not the translated English word, which would obviously affect the analysis.  The subsequent wordcloud should automatically be translated, but if not simply use the dropdown at the foot of the page to translate to English again.

The above steps show how someone can analyse social media in any language that they want.  In fact it also shows how someone with no English could translate the English social media analysis into the language of their choice!

This is a truly powerful tool and is an extremely cost effective way of monitoring multiple markets in multiple languages using our discover tool.

Contact Steve Bone for more information and to discuss the package that best suits your needs.

Posted in Analysis | 1 Comment

Brand Aura Product Launch: discover

Brand Aura Product Launch

level 1 : discover

Good marketing campaigns start with the question “what do we know”?

Great ones start with “what don’t we know?”

When looking at new territories, new markets, new opportunities, product development, brand extensions or, well, anything new really, the blank sheet of paper stage is the hardest of all.

Primary and secondary market research is a key part of course – but we think millions of pounds is spent in a wasteful or unfocussed way because brand managers and planners don’t know what they don’t know.

Brand Aura discover is here to help.

By listening to the existing conversations on the internet right now in your chosen area and bringing you the results in a manageable, intuitive way, we help you conduct your basic spadework to the point that you are ready to shape a proposal or commission primary research into the field.

You can spend tens of thousands – maybe more – of pounds, euro or dollars to get the basic intel you’re looking for. Or you can use us, and have a better insight at lower cost.

By simply giving the team at Brand Aura some search terms to go on, we’ll build you a picture of what’s going on. We can do this over time and show you trending analysis if you want. And our data will allow you to ask the most powerful question for marketers – why? We’ll give you the context in which people are mentioning, say, Car Brand X, rather than simple word counts or basic sentiment analysis.

Costs start from just £2,000 for basic analysis of conversations on twitter over a defined time period. If you want us to reach deeper behind the sofa cushions of the internet, we’re pleased to do so, and we’ll work with you to generate a custom package. We think the value you’ll get and the money you’ll save on your other research will more than pay for itself.

Brand Aura discover is the head start you’ve been looking for.

You can play with a web demo version of the technology at this link, follow the instructions to explore the data set of your choice from the ones listed.

Contact Steve Bone for more information and to discuss the package that best suits your needs.

Posted in Brand Aura | Tagged , | Leave a comment

Social Media, the new exit poll?

An interesting article on Adweek today discusses how twitter can be used to determine the winner in US New Hampshire presidential debate.

The article shows how sentiment analysis can be used to give an overview of people’s instant reaction on facebook and twitter to the presidential debate.  The sentiment analysis appears to have been quite clear, with Mitt Romney doing better than both Michele Bachmann and Ron Paul amongst others.

A number of sites such as Likester analysed the ‘likes’ made on Facebook during the debate.  Another company Trendistic focused more on specific hashtags or keywords used during the debate in order to graphically display how the public increased the number of tweets made on debate itself.

The end of the article ends on a rather downbeat note: “language processing software is not yet advanced enough to get a read on what exactly Twitter users are saying.”

Evidently, the author has not seen our analysis since we have been doing exactly that for over a year – in fact a year ago almost to the day we were analysing social media data on the UK elections.  Not only did we predict the winner of the TV debates before the UK pollsters, but we could give explanations as to why the winner won.

More recently, we looked at the UK version of the X Factor, where not only were we able to predict correctly the final 1-2-3 but we were able to explain why the public voted the way it did.

So, interesting to note that Brand Aura is still the only company capable of delivering true contextual analysis of social media data, and that can give you true market intelligence as to why the public thinks good (or bad) of you and your products.

Posted in Analysis | Leave a comment

Can social media give true market intelligence?

To gather market intelligence, traditionally it is necessary to undertake expensive, and slow, studies.  Even a desktop study that does not actually interview anyone can be expensive and you can sometimes still be waiting weeks for the results.

Unfortunately traditional marketing campaigns have even more drawbacks.  For example, often it is only once the interview stage is complete that true market intelligence begins to appear – and you are often left frustrated at the incompleteness of the answers – why didn’t we ask them more questions about that?

Can social media replace the traditional routes to understanding a market?

You may think it is impossible.  I mean how can social media data replace a targeted question?  How can it replace interviews that are designed for a particular demographic?  How is it even possible to filter through the mountains of dross that are posted through social media – in what way would someone’s drinking habits (and subsequent hangover) be useful in undertaking competitor analysis?

All of the above points are valid.  However, if you can identify social media comments that are in fact about your brand or product, then it is an enormously rich source of information.  The demographic for a particular social media channel can be identified (albeit not to the same degree of accuracy as would be done offline), which solves that particular issue.

In terms of defining the questions to ask, the analysis of this data requires the marketeer to flip his/her thinking.

What do I mean by this?  Well, normally the questions that you would like answered are predefined.  A great deal of thought has to go into what the questions are, and to ensure that they do not promote bias one way or another.  Respondents are either phoned, emailed, or interviewed face to face.  The analysis then looks at the responses that are usually in the form of “Rate our product from 1-10 with 1 being terrible and 10 being fantastic”.

However with social media, your customers (or potential customers) are giving their opinions on the things that are important to them for free.  This has a number of benefits, but not least is the fact that they are more likely to be honest!  This means that the answers to questions are already there – you simply need to determine what questions to ask.

That sounds like an impossible task – how can you possibly know what questions to ask without knowing what the answers are?

The use of contextual analysis solves this particular issue.  Contextual analysis has one particular benefit, in that it would automatically ignore any comments not relevant to the topic in hand (i.e. ignores the dross).  Since the social media data is analysed in context you can ask a series of simple questions to drill down into the data and find out the issues that people find important.

For example, the first question could be “What issues are important for the customer in relation to  market X?”.  This question is posed by simply asking for words in context with a word that describes the marketplace for your product – e.g. “financial”, “ice-cream” or “printer”.

The analysis would then show the words that are identified in context.  To understand the data still further, you could then analyse one of these words in context with the first.  This allows you to drill into the data and find more information relating to these words.  In each case, when asking for the analysis of words in context, you are asking “What issues are important for the customer in relation to X?“.

This type of approach gives you a number of advantages, not least that it will be quicker, and less expensive, than traditional routes.  In addition, the marketeer can focus and concentrate on the important issue of better understanding the brand or product, and how it is perceived in the marketplace.

In summary, this blog post shows how the use of social media is an effective way to gather market intelligence.  In some cases it can be used to replace traditional methods, or at least, complement them so that the final interview stage can be better prepared, improving the overall ROI.

For an example of this process and an extended whitepaper exploring how this approach has been used, please go here.

For further whitepapers and more examples of what contextual analysis can do for you, please refer to the whitepapers section of our website.

Posted in Analysis | Leave a comment

Contextual analysis of social media for the financial sector

Brand Aura has undertaken a 3 month long analysis on the analysis of context in tweets made that relate to the financial sector.  We downloaded tweets over a 3 month period, that were in any way related to the financial market.

We then used this data to identify trends in the data, which gives an insight into what the public are saying about any particular topic.

However, since we analyse in context, we can more readily identify what issues affect a particular brand, or a particular product in the financial marketplace.

As an example, we could ask what is in context with the word ‘mortgage’?  And how does this context change over time?  In other words, what issues are the public discussing that relate to the concept of mortgage?  Are there issues that a brand should be concerned about?

We can go further than this, and ask a further question in context.  What do the public talk about when they discuss both the words mortgage and best in context?  Notice here that we are not saying what words appear when the public use both mortgage and best.  Our analysis will include tweets that perhaps don’t have either word, since it analyses what is in context.  (e.g. the tweets may include a common slang word for mortgage, or instead of ‘best’ may use other positive words like ace, great, awesome).

It should be obvious from the above how powerful this approach is.  If we don’t know a specific word that the public is using it doesn’t matter, since the analysis will automatically include it for us anyway.

We can see an example of the analysis when we look at the double context of the words mortgage and worst.  The figure showing the analysis is shown below, where the closer the words are to the centre, the closer in context these words are to the double context of ‘mortgage’ and ‘worst’.

Words in context with mortgage and worst

As can be seen, Yahoo Personal Finance comes out very closely in context with ‘mortgage’ and ‘worst’.  Is this due to Yahoo being negatively perceived by the public (they do not offer mortgages directly), or is this due to advice that Yahoo are giving, in particular with respect to bad mortgage deals?  Either way this shows how contextual analysis automatically identifies brands that are close in context with any particular topic.

Another example could be to ask a question that cannot be answered by standard sentiment analysis.  Sentiment analysis will tell you how many ‘positive’ comments have been made about your brand or product, and also how many ‘negative’ comments.  However what if you want to measure how close in context your product is to particular financial services?  For example, if you wanted to measure which banks did best when the public discussed words such as: interest, rate, payoff, return?  Normal sentiment analysis would not be able to tell you, but contextual analysis automatically identifies how closely in context the brands and products are to defined key words.

Context of banks against key search words

The above figure shows how four main bank brands score against those key words for one particular day in the analysis.  As can be seeen, Barclays scores highly in comparison to the other banks.  This analysis shows what happens on a single day, and our analysis can automatically track this over time, so that changes can be seen.

A full whitepaper is available that shows further analysis on this topic, and how this analysis changed over time, and can be seen in our whitepapers section.  The reader is directed to the whitepaper in this section to review the complete analysis that was conducted on the financial data.

However it can be seen even in this shorter summary of the analysis undertaken how contextual analysis gives a much richer and more complete picture of the public’s perception of the key areas in any given market, and the ability to track and compare competitors over time.

Posted in Analysis | Leave a comment

The challenge of language in analysing social media

You own a brand or product.  Either that or you are briefed with gathering market intelligence for one.

Social media is the new buzz.  Millions, no gazillions, of people are tweeting and blogging about your product every day.  Every minute of every day.  How do you analyse this avalanche of data?

One way is to employ someone – or a team of people – to read each blog, tweet or post.  This can be expensive.  And very slow.  Plus two people can read the same post and form different views as to how bad or good it actually is.  So not very scalable.

How about you use sentiment analysis companies?  Everyone else is doing it, it must work, right?  Well, sometimes.

If you know the key words for your market place – and by that I mean you already know precisely which words are considered ‘positive’ and which ‘negative’ – then sentiment analysis can work quite well.  You will be able to work out how ‘positive’ your product is doing, and how ‘negative’.

But what if things start to change?  Does it tell you why?

No, and the reason for this lies in how it analyses the data.  Sentiment analysis uses natural language techniques (NLP).  These techniques break up a sentence into its component parts – nouns, verbs, adjectives and so on.  It can be used to work out what the ‘object’ of a sentence is.  All that stuff that we learnt in High School English and forgot the moment we were out the door but do automatically every minute of every day when speaking and communicating with our friends, family or work colleagues.

Now NLP works really well if sufficient computer power is given to it.  And herein lies the problem.  To analyse millions (or even thousands) of tweets or blogs to any reasonable level of detail using NLP techniques requires exponentially more computing resources.  In other words, it simply takes too long, and cannot be done.  So the NLP techniques are stripped down to their bare bones, and the analysis essentially equates to key word searches.  Not so useful then.

This is part of the reason why we concentrated on developing new technologies for the analysis of text.  We analyse context and this is at the heart of our analysis.  Because of this, we are scalable, quick and accurate.

The use of our contextual analysis will therefore allow you to automatically work out what people are saying about your brand or product.  It will allow you to compare this against any competitors.  And it will also allow you to automatically track any changes in time.

You may ask how we can do this if we don’t use NLP techniques.  How can we track changes if we don’t break each sentence apart.  Well, that’s not what humans do.  When you have a conversation you don’t work out whether the word is a noun or verb before replying.  You know automatically from experience what type of word it is, and more importantly know contextually what words would come next.  This is more like the approach we take, which more closely mimics the human way to analyse text.

The best way to see this is to review the whitepapers we’ve produced which show the type of analysis we can undertake.  You can view them here.

Posted in Analysis | Leave a comment

What is the difference between contextual and sentiment analysis?

Many companies are now convinced of the merits of sentiment analysis.  It gives them an instant snapshot of the public’s view of their brand or products.

But what if the sentiment changes?  Does it automatically tell them why?  Well, the answer to this question is no, it does not.

What if the brand or product is a huge brand like the iPhone.  The number of posts that are positive will be so huge that any negative sentiment will be lost, swamped by the deluge of posts made that are saying good things about the iPhone.

As an example, when the iPhone 4 came out, there was a huge swell of positive sentiment about the new phone.  However, there was a significant minority who picked up a problem with the iPhone battery.  This grew in size as it was a design flaw in the model.  Sentiment analysis techniques will tell you the negative sentiment around ‘iphone battery’ but you would need to know to filter on ‘battery’ to identify this negative sentiment.  In other words, you would already know you had a problem.

Alternatively, you can monitor key words, but what if the issues that affect the public are not in those key words?  In fact this is often the case.

Contextual analysis solves these problems.

Since our analysis is driven by what the public say, rather than preselected keywords, then as the issues change, so does our analysis.  We would automatically pick up the fact that ‘battery’ was an emerging topic for the iPhone that was growing in importance, and also had a largely negative sentiment.

The use of contextual analysis puts you back in control of your brand and means that you can properly listen to your customers, identify any issues early, and resolve them effectively.

For more information, contact

Posted in Analysis | Tagged , , , | Leave a comment

How effective are your marketing campaigns?

So you have commissioned a new advertising campaign. It includes digital marketing as well as traditional channels. You’ve even got a new Facebook page and a strategy to get your customers to “like” you.

How do you measure the ROI? Ok, you are told that there are indirect measurements that will tie in activity against this particular campaign. But do they just take credit for purchases that were going to be made anyway?

This is not a new question. And the methods to examine the answer are not cheap and suffer the same flaws as the original market research pieces that were probably used to define this very campaign.

So can the question even be asked?

Well, perhaps we should try to break down what the question means.

Ultimately any campaign is about increasing sales, and so if we see the bottom line on an upward curve, then we could say that the campaign has been successful. But has it? Are the reasons people bought the product down to any particular marketing campaigns?

By using social media we can actually start to answer this question.  For example, we can monitor what types of words people use around a brand before, during and after a particular campaign.

Do the words change?  Does the marketing activity affect the words that are used?  i.e. if the advertising promotes the brand as ‘fruity’ does it start to be called ‘fruity’ by the public?

And how long does this last?  Does the advertising need to continue (at great expense) to keep this key product feature in the public’s eye?

Only contextual analysis gives you the ability to do all of the above.  The great benefit from this is that you as a brand owner understand your product better, what the public likes about your brand, and what it doesn’t.  And perhaps as importantly, what differentiates your brand from your competitors.

Posted in Analysis | 1 Comment

What do we do?

What can contextual analysis give you that sentiment analysis does not?

Well, we do not use NLP (i.e. Natural Language Processing) unlike everyone else, and we’re not trying to mimic the plethora of sentiment analysis companies that are ubiquitous these days.

We can tell you what words are important for your brand. We can do this based on comments made on twitter, or on forums, or even using your own data, such as emails that are sent to your complaints department.

Let’s put this another way. Sentiment analysis works ok if you already know what you’re looking for. However we live in a changing world, and what is an important issue one day is irrelevant the next.

We can work out what is important for your brand/product. We can do the same for your competitors. We can tell you what is different between you and them. We can also track this over time, so that you can measure the effect of any marketing campaigns that you do.

And we do all this in context. Which means that we’re more than a simple word search or word filter.

We’re also not limited to English. We can do any language – although we may need help in translating the key words that we find but Google is pretty good at that! :)

Posted in Analysis | Leave a comment