The past few posts have taken a stroll through a variety of buzz research tips and techniques. What this post attempts to do is to grab all the things we've looked at in the previous posts and integrate them into a coherent framework for conducting simple buzz research.

1. Develop a hypothesis

There's a temptation with buzz research to just say "I want to monitor what people are saying about my brand online" without really thinking about what you want to get out of the exercise. It's important that you develop a hypothesis which can be tested by measuring variables generated by naturally occurring online conversations. For example, responding to customer service queries on Twitter will lead to a decrease in the number of negative tweets relating to our customer service. Having a clear hypothesis will guide how you gather your data and what variables you choose to measure. Generally I would use buzz research in relation to a particular tactical need which lends itself to a straightforward pre measure/post measure methodology. It's also worth noting that buzz research by itself will probably not generate enough evidence to draw proper conclusions often you will use it in conjunction with or as a primer for other research.

2. Recognise data sources

Your hypothesis will influence where and what data you choose to collect. For instance if the hypothesis is related to a particular service like Facebook you may choose to only analyse data from Facebook. Alternatively your research could be country specific so you may choose to look at data across services (Facebook + Twitter + Company Blog + Email + Fan forums etc) but only if it is posted from or by a user from a particular country. Sometimes it may be easier to use an aggregator like socialmention to get an overview of social media services.

3. Extract the data

Once you've decided what you need to measure and where you can obtain the data, you need to extract the data in to a manipulatable format. A simple example would be getting comments from the company blog into an excel file. Sometimes there will already be a mechanism which will allow you to extract the data, sometimes you'll rely on good old copy & paste. There are some tools which help extract data from specific sources like IPO's twitter search exporter. There are also some more advanced tools like dapper which you can use to extract data from a range of sources.

4. Format the data

Often you'll be getting data from a range of sources which normally means it will come out in all sorts of weird and wonderful formats. The next stage is to bring all the data together into a single dataset ready for analysis. For example the structure of your final dataset may contain the following variables: Text, Length of text, Date text was posted, Country, Source and it's our job to take the disparate data we've extracted and get it under these column headers. The simplest way is to use copy & paste but you will save a lot of time learning how to do macros in Excel; Yahoo pipes is also pretty handy when it comes to merging data.

5. Analyse the data

Now the data is in a suitable format you can analyse it and hopefully prove or disprove your hypothesis. There are some natural language processing tools which can help with the categorisation of text. However I would be wary of using tools where you do not understand how the results are calculated. Generally content analysis would be my weapon of choice when looking at this sort of data.

6. Conclusions

Hopefully your analysis has enabled you to prove or disprove you hypothesis so your free to go ahead and discuss what implications these findings have, highlight areas for future research and all that good stuff.

The end

Hopefully this and the related posts have given an overview of how you could use a range of free tools to conduct buzz research fairly quickly. Sure, we've only scratched the surface and there are flaws with the approach I've outlined but I feel it does give a gentle, easy to follow approach to conducting DIY buzz research.