Fake it ‘Til You Make it (Or Don’t) Part 4: Are We All Catfish?
By Gemma JoyceSep 20th
Published May 17th 2017
“Sorry to do this to you, Carly, but we have a pitch for this big brand owner at 2pm. Can you pull something together? Have a look on social media and see what you can find out.”
As a junior account exec, this is a sentence Carly is depressingly familiar with.
Where to start? A Twitter search, a quick look for trouble in the relevant Facebook groups, and a look at Google News all help. But she really needs access to good social analytics to build comprehensive, quantifiable reports.
And that is where Carly’s problems start.
If the brand is a popular one she will find social conversations dominated by noise, passing mentions, and spam.
Where is the insight? Where is the nugget that will win you the business? What is the one thing Carly can discover that your opponents won’t know about?
The clock is ticking. It is already 10.30am.
Solving this problem has been a big focus for the Stuttgart team at Brandwatch over the last year. We called it Speed to Insights. But we weren’t working alone. The work has been proceeding on three fronts.
The key piece of work, and one which underpinned everything, was to completely re-engineer the back end processing of Brandwatch’s data storage.
We called this the “Stability Project”.
Unlike a lot of social intelligence companies, Brandwatch crawls and stores its own data. This an epically vast – and expensive – undertaking; we index 55 million pages a day.
Our customers mark these mentions up with their unique categories, tags and rules and store their data in their own accounts.
The result is Brandwatch’s uniquely powerful capability to slice and dice data by brand or product names, campaigns, customer service issues, audience demographics, in fact by any category you can think of, and provide segmented views.
But this flexibility came at a price. The platform was suffering growing pains as customer numbers and data volumes grew.
The solution was to overhaul completely the crawler structure, the queues that wrote the mentions to the database and the database itself, moving from PostgreSQL to Solr.
[As a side note to this – when the Twitter data team visited recently and we told them about this, they leant forward. “Really? How did you do that?”
These kinds of data migrations are necessary to every company in the business, and they are uniquely painful. At one point during our data migration, we were discovering bugs in one vendor’s software which were only resolved when the software vendor patched its release – after three months of our solutionising.]
The result looks like this. That’s over four million mentions of Starbucks being searched through, filtered and analyzed in real-time.
Before the data migration this process took longer than we liked. Now the platform can eat a Petabyte of data for breakfast and not even burp.
Brandwatch has some super-powerful features, like its ability to mark up mentions in intuitive ways and create easy-to-understand reporting from the data.
But our research was showing that as much as super users loved Brandwatch, beginners could find it intimidating.
Carly has watched on as senior Brandwatch users made use of Rules to divide up customer service conversations from marketing conversations, but no one’s had the time to teach her how to do it.
The next stage of the process was bringing these powerful features to bear in ways that were accessible to Carly and other entry-level analysts.
The way we did this was to build a wizard.
The wizard asks the user what he or she is searching for, and updates the Query accordingly.
It even predicts the hashtags, keywords and Twitter handles that might be most relevant to the user. The result is a Query creation process which is much, much simpler.
By exploring the results, Carly can exclude and include based on her changes so she can begin to really understand what her changes are doing to her data.
Now Carly is pretty happy with the preview data she can see, and she can save her Query – even sampling it if the data set is too big.
The next stage is to get it all into a dashboard and work on it. The clock is ticking. 10.45am.
A summary dashboard is great, but it doesn’t break the data down into key categories. For a big brand owner, Carly wants to see which brands are giving trouble, and which ones might offer opportunities for an ambitious agency.
With the Dashboard Wizard, Brandwatch Analytics has come to Carly’s rescue with a series of simple steps that allows her to slice and dice the data based on information she has already given us.
This information allows us to suggest Twitter handles, hashtags, and terms. As Carly enters these, we create the Rules for her in the background. With the lightning-quick search function Carly can search within the data and create new categories of her own.
In her new Dashboard Carly can see key topics that are trending, not just at the corporate level, but brand by brand.
This is what you’d see for a search on Pepsico’s brands:
OK, all good so far. The challenge is then – how to share the results quickly? There is another meeting coming up a midday and it’s already 11.15am. Carly needs to create something that can tell the story of what she has found, and get it in front of the client in a way that lands and wins the deal.
That’s where Brandwatch’s PowerPoint export feature comes into play. At the click of a button, all the key components are loaded into separate slides that can be annotated, marked up and given the agency brand makeover.
At 11.45am Carly messages her boss to say that the research section of the pitch deck is ready in her inbox.
At 2pm, the presentation begins that wins Carly and her boss the new account.
All thanks to Brandwatch’s speed to insight? We like to think so.