SAP HANA, Big Data and Data Science- Part 4

'Do it right the first time , every time' the BVA principle

Big Data is a very big buzzword. Global enterprises have responded with Gartner reporting that 72% of global enterprises have started some form of a big data project. The world around us is filled with big data software, products, applications and technical solutions and companies continue to invest into them from one leap of faith to the next. Gartner reported in 2012 that fewer than 30% of global BI project will meet business expectations. Their recent report on Big data tends to report the same numbers for Big Data initiatives. A recent paper by Kapow Software states: “Big Data projects are taking far too long, costing too much and not delivering on anticipated ROI because it's really difficult to pinpoint and surgically extract critical insights without hiring expensive consultants or data scientists in short demand.”. Kapow reports that around 10% of big data projects are actually effective in identifying the data they need for their analytics, and also that only 10% of the vendors actually provided effective business benefit focused guidance.

Just last week walked out of a meeting with an EMEA enterprise that has invested in sensors for monitoring traffic patterns in their retail establishments and for the last 8 months they have not been able to collate the right data to the point where they wondered if the sensors were creating random ghost images by sending a handful of their own employees that they managed to track.

Most stakeholders are misinformed about the value of technical solutions. It is fairly easy to find platforms, vendors, databases and applications as technical add-ons. However, what is becoming apparent is that deriving BVA, Business Value Attainment, from these investments is an entirely different story.

Most companies and many individuals are rebranding their current developers, db administrators as data scientists. As these companies step from one puddle to another they soon relies that data science is a totally different game and they only find that out when they get a data-science resource on board.

Just as an example think Shazam. An app where you can point your phone towards any song for 15-20 seconds and  an in a matter of 5 to 500 milliseconds it will identify the song and its group with an accuracy that is uncanny. Though there are many data points that are streaming and very random it is pattern recognition that makes this application work. All this happens with network latency and on your cell-phone (added latency). So the two ways to solve this problem were lets get the worlds largest and fasted systems and let them identify a pattern from their data store- took too long. Or let the data scientists find an algorithm to solve a business need. They key to this is that Avery Li-Chung exposed his fingerprint, pattern recognition algorithm that does all this in sub seconds. It takes 15 individual pitch changes that define what the song and its group is as clearly as a fingerprint is used to day to find a criminal. The key here is not the system, the data, or the technical details but the pattern and the business case- tell me the song on a mobile phone with very little network bandwidth. Step 1 is defining the business need and step 2 is identifying the pattern to meet that business need – the rest is history. This is data science 101.

Data science sits above the infrastructure, tools, HW, databases and your traditional developers. Pure data-science needs to be led by an end point- that being very concrete business benefits deliverables. For HW companies it is easier to sell HW than to provide a business solution. The same goes for infrastructure, SW, technical experts and the rest.

Before taking the Big-data and IoT leap we firstly ask planners to stop briefly at Security and then head straight to BVA guidelines. Focus on the business need and use everything else as a catalyst to meet those needs.

If you are in this situation then try some of these processes:

1. Identify critical business-benefits : that work with very large data sets, that IT says cannot be deployed just because the technology did not exist? i.e. tell me the song and group in 20 seconds; on my mobile device; simply by pointing my cell phone towards the music, with over 2.3 billion songs out there.
2. Conduct a ‘Decision Design Thinking’ workshop: with 80% business stakeholders and 20% IT. Identify 8-10 needs and then prioritize and deliver 1 with the highest business value in 18-24 weeks maximum.
3. Build an IoT matrix: If your enterprise is planning to cross the digital divide for streaming / connected data feeds then start working the Digital Matrix and plan on a foundation on proven scientific processes with business benefit as the prime goal. Getting digital once again is a process of business benefits and not deploying a bunch of sensors as they come.
4. Don not ‘Build and they will come’: Has been a defective technical solution from way back in 2004 when Gartner reported this defective process. Question if it has every worked in your enterprise and then follow your true north accordingly.
5. Never forget the Human Emotional factor: at the end of every IT cycle what are we truly aspiring towards is the fulfillment of a human need- the user. Better reports, faster data and volumes of selections may be nice but they are of little value if our human users cannot use the deliverables. Measure not your success by the number of reports delivered, or by the on-time / on budget aspects of the project finish but by the bottom and top line human benefits it provided- measure it by user success scores.

1. It’s not about the technology but the decisions it can enhance and the problems it can solve
   The statement above is not entirely true for example I cannot take a 1970’s HW and try to run big-data solutions on it. Technology is important and a mandatory foundation but it must never be the goal. As we always tell our customers for SAP HANA – ‘HANA is a strategic Business Solution and not just another technical install’®. So Start with Business then the technology to deliver those needs. Don’t start and end with the technology alone. If you simply go and install HANA as another upgrade then that is all you will attain.  Just like a democracy we mostly get what we vote for.
    
2. Agility is a fundamental key
   We were in a project where a SI partner had convinced the customer to conduct a 18-24 month global audit to understand their global analytics needs before building their decision platform. There are two fundamental defects with this approach. In 18 months the world will be very different so whatever lessons they learn will be redundant by the time it is reported. The second is no enterprise can afford 18-24 months to critical decision enablement. We recommend Build in small step, 90 to 120 days, and get to production within this time. Build iteratively and measure business benefit scores in each delivery iteration. Start with business, work through with business and end with business. Technology is simply a service to meet their needs. Most big-data products are fairly massive and inflexible unless contained and focused on a clear business deliverable.
    
3. Small steps towards a common strategic Goal
   There is little benefit in building a strategic big-data solution that will run for 18-24 months only for business to get a surprise after all that time. According to a BI Valuenomics and Gartner driven report in 2010 ‘98% of BI projects are declared successful in week 1 after go live, yet fewer than 50% remain so by week 10’.  According to our findings define clear strategic and midterm goals. The work on the 90-120 agile delivery programs as defined in #A. this is like buying a 13 year old child a Tesla when they can’t drive for another 3-4 years – things can only go wrong.
   In big-data it is -> the business direction focused start -> coupled with Data science know-how of finding patterns, or extracting small amount of data to meet defined needs from very large data sets. This would normally include sampling, variable compressions, defining and choosing the appropriate algorithms, and managing to shrink all that big-data into something that can rapidly, in microseconds, fit into an analytic model on your users laptop or smartphone. By putting business before the technology and small before the big architects will maximize their assets and investments with high user scores and business benefits.
    
4. The right Data Scientist
   If your company has a true data-scientist you are indeed a minority with a unicorn. Some companies, and individuals,  are currently rebranding their software developers and less experienced data analysts as their data science team. So when asked they say they have resources retitled as ‘Data Scientists’. These companies are speeding along the technology alone can answer all questions path. These companies will continue on a path of high stress, low business benefits and probably ending by telling business ‘ here is your big-data you do what you can with it’.
    
5. Start with renting then buy
   If you are unsure of the benefits of true data scientists then hire a group with proven BVA experience.  If you’re unsure what to look for read the prior three blogs on data science, 1 / 2 and 3,  and then see if these resonate with your go-forward planning. We have access to exceptional PHd level data scientists that companies can leverage on focused projects as the fastest way to find out their benefits is to actually try one. Data science some in two parts the Business Value Data Science in finding the small, medium and strategic goal sets. The second is the actual data scientist that will work with the team and build, test and define the true data science magic.
   Once you get a taste of what a true data scientist can do only then you will understand why they are being termed as a unicorn and why competing with the Googles, Facebook’s and the like is probably going to be a very interesting endeavor indeed.