Johannesburg, 12 May 2016
Kroshlen Moodley, GM of public sector at SAS Institute, is not a man who minces his words, particularly when it comes to government bodies using analytics to get the best out of social media.
"I believe government departments on all levels - local, provincial and national - are losing out on a huge opportunity to harness analytics and social media to better deliver services to ordinary citizens," Moodley says.
He says that, while some ministers and politicians may have a Twitter presence, this is often of more recreational value than making a difference in the lives of citizens. "And yet, the opportunity is huge," he says.
"Take a basic service like water provision. A number of government players - uMngeni Water, Joburg Water, Rand Water - have a social media presence, but many are very inactive. Lead SA's Operation Hydrate showed us that people can be motivated through social media to make a difference around water, or at least to raise awareness about their own water usage. Government can certainly learn from the private sector in making social media personal."
Himself an ex-government employee, he points to the City of Johannesburg's social media manager, Tumela Kumape, known on Twitter as TK, as an example to follow. "What I love about him is that he's in touch with people, his response times are good, and people ask for him. He is in your face, fresh with messages, and has permission to be sassy, but never disrespectful. Governments have to find that person to handle their social media. Don't let your junior intern run your social media - you have to find someone who understands your brand, and is astute about what's happening in the country, what the general sentiment is and what government is saying in response. This is where you can really tap into an advantage by working with sentiment analysis."
Also known as opinion mining, sentiment analysis uses a combination of natural language processing, text analysis and computational linguistics to identify the emotions and other subjective information in source materials, in this case, social media.
"There is a huge amount of data in social media. If you pass it through taxonomies, you can identify whether the general sentiment is positive, negative or neutral. Now, couple it with structured data and you can see how to position your messages going forward."
"Add a geospatial layer, and you could, for instance, pinpoint service delivery issues before they get out of hand. I believe governments are not doing enough to understand what's going on with the citizens, to deliver better service to them."
He is quick to discount concerns that a lack of universal Internet connectivity in rural areas could mean these systems can't work. "There is no such thing now. Devices like smartphones are here - IDC and Gartner talk about 20 billion devices in the next few years. We have people who have connectivity through smartphones, but we haven't taken advantage of that leap."
Moodley acknowledges that South Africa's 11 official languages add a layer of complexity, but says it's not insurmountable. "Yes, the taxonomies would be a bit more difficult to create. But then, a lot of people will be communicating within groups, so messages should be customised for groups. It's similar to how retailers use social segmentation and campaign management to target specific shoppers. It can be done."
He is also not put off by concerns about old government IT systems and a need for security hampering social media solutions. "A social platform is not dependent on your legacy tin. The social platform is really a different platform, it's a breath of fresh air. It's more about changing the mind-set of people to focus on the new things. The need for security is real, but if SARS and Treasury can find a way to allow citizens to interact with them via an open platform, other government departments can too. There are smart tech people who can create secure environments that still allow citizens access to the information they need, such as their utilities bill."
The advent of in-memory technologies, he says, means it is now possible to do predictive analytics on data sets "without building the warehouse".
"The current technologies of virtualisation, artificial intelligence and machine learning mean that it is now possible to allow users - government employees - to have faster access to data. You no longer need to be a statistician to analyse the data. You are now using a dashboard.
"The visualisation of the data is very powerful. You can see the gaps quickly and find ways of closing them. Previously, someone in a processing environment would have to request a specific report from IT, who would have to clean the data before delivering a report to you two weeks later. That's no longer the case - you can do it yourself and change the questions you ask on the fly."
Asked for a "real-world" example of this ideal world, he turns to the police. "Let's take a basic scenario, where the local police station in a small town has no computers, uses paper-based statements and doesn't have a tracking system or dash cams in its vehicles. You start by ensuring that the temporary manual statements and case numbers are captured onto a system, which gives a permanent case number. The investigating officer has access to a dashboard that allows her to visualise where the incident happened, where the victims live, what other crimes happened in that area over the last 24 hours. Combine that with data on where the police vehicles were at the time, and do it for a month. This becomes your structured data.
"Now, harvest some data off social media, choosing to include specific hashtags or keywords and opinion mining, and you give station management the ability to do dispatching and shift planning optimally by identifying very precisely the areas of high crime or potential service protests. Or catch corrupt cops by geo-targeting the vehicles that were in the area where people are complaining on social media about being asked for bribes.
"Then expand the solution to the district, then to the province, and then nationally. If Sanral, eNatis, the department of transport and local police were all connected and understood things like cars going through toll gates, and share information about crimes in the area, traffic patterns, specific high-crime time periods, and overlay social media information, they would be able to isolate and identify perpetrators."
But, can the public service afford this type of solution? Moodley says: "As a South African citizen, I don't think we can afford not to do it. SA is struggling to fund the social burden of 16 million social grants, unemployed youth subsidies and the like. But our fiscus is losing money through big ticket items - things like gold smuggling, VAT fraud, customs and excise fraud. If you start using analytics, you will identify syndicates through network analysis and link analysis. And that can make a huge difference to our books."
To find out more about the role social media analytics can play in government, download the SAS white paper: Applications of Text and Social Media Analytics.
Share