To create value from data, you do need a big data strategy – right?
Well, actually, probably not. According to Gartner, the ‘all-encompassing data strategy’ fails around 60% of the time. In fact, Gartner have even stopped publishing a Big Data ‘Hype Cycle’ (their measure of maturity and adoption of technologies).
According to Frank Buytendijk, research fellow with Gartner, ‘big-data’ has several different components and as a consequence, he argues, “Organisations need an overall ‘data and analytics’ strategy.”
With Forrester claiming only 12% of data is ever analysed, how do you ensure the right data is being analysed, by the right people to generate the right insights and solve the business challenges on the executive agenda – be that innovation and growth or end-of-life and cost reduction …., plus everything in between?
The 6 key challenges
Data itself, is valueless, and only becomes of value when insights and actionable outcomes are generated and executed upon in a timely manner. In order to deliver this, most organisations face 6 key challenges;
- Resource & skills – Data is siloed, so collecting, cleaning and organising the data is a difficult task. On average, data scientists spend almost 80% of their (expensive) time doing this, and not adding any insight. Most decision-makers don’t have the time or skills to undertake the necessary analytics to generate the insights. So rely on business analyst, IT, finance, anyone really – all of whom have priority conflicts, objectives and will often look out for them over the greater good.
- Accuracy – Spreadsheets are the tool of choice for almost all analysis, or its big brothers Access and SQL. However, 88% of all spreadsheets contain at least 1 error. So, when merging, linking and creating formulas across multiple spreadsheets already containing errors, and then building (or worse deciphering someone else’s) formulas, even small errors can mean significant, share-price effecting, inaccuracies.
- Insights not metrics – With the digital revolution happening at pace and businesses producing ever-more data, ever-more metrics are produced on the back of it. But creating actionable insight that creates business value is a separate thing entirely – it’s the difference between presenting ‘how many sales reps left last year’ and ‘why our top performing employees keep leaving?’.
- Agility - Historically, insights have been produced in a ‘batch process’ approach. Teams go off, collect data, analyse it, come-up with a solutions, propose it to management, who mull it over, ask questions which takes time to respond to and then decide how they are going to act. At best this cycle takes weeks, often months. In todays world, this timescale is useless and often dangerous for a business.
- Collaboration – it is important to be agile, but it is also important to ensure the impact of decisions across the organisation is known, understood and where appropriate, agreed.
- Cost – businesses may already have invested significant sums in systems and technology – 80% of IT costs occur after the initial purchase. How do you maximise current, embedded, technology investments and resource, and minimise new platforms, software costs and expensive analytics talent?
Practical solution steps
The challenges faced are diverse, and require a mix of technology, people (with varied skills) and process to address them. Each challenge could be addressed through building in-house or acquiring products from a variety of providers. However this approach can be costly, time-consuming and ultimately disparate in nature – not bringing a complete solution to bear.
This is why more and more businesses are partnering with 'analytics-as-a-service' companies. These agile companies build solutions that fit the business processes and technologies, that can start small and scale as required, therefore do not require any capital outlay or long-term commitments.
Two key benefits to ‘analytics-as-a-service’ are speed and choice. Analytics and insights can be delivered, shared, evaluated and approved quickly, before building-out as the business requirement dictates. In the event the partner doesn’t fit your business, you can quickly move-on to an alternative solution with no legacy to deal with.
Whatever your approach to data analytics, whether in-house or through partners, here are 5 steps to increase the chances of success and maximise the value extracted from your data.
1. Start with defining the challenge you are trying to address. Begin with a business problem. Problems are individual to a business and indeed individual decision makers. From wanting to understand customer behaviour, sell more and create loyalty, to changing cultures and behaviours, reducing costs or improve employee retention. Collecting business data is important, but data on its own won’t make you successful. In order to generate insights that enable action, you must first know the problem you are addressing.
2. Collect ALL relevant data. You don’t want spreadsheets, you want raw data. It doesn’t matter how many disparate systems you have, the important part is having the raw data that has not already been manipulated – eliminating the errors that spreadsheets are plagued with. Capturing data from internal systems should be relatively simple – ERP, finance, operations, HR, CRM, manufacturing etc. But don’t forget external data, which can be equally useful, dependant on the problem. From social media to weather history/forecasts – the more relevant the data, the better informed the analytics, the greater the insights and the more effective the decision making. There is little point in collecting data for collecting sake – if you don’t have a purpose for it now, its just noise and will be out-of-date by the time you might have a purpose for it (caveat – the exception to that is customer buying data, the more data you have, over a long period of time, the better behaviour analysis you can undertake).
3. Tools & skills. The key to analytics is doing it properly. Most analytics tools require skilled experts with a data science to use them effectively. Even the ‘new breed’ of self-service analytics requires some training on data manipulation, it's not that users can’t learn this, it's more they already have more than enough to do (likely explaining why adoption of self-service analytics is already in decline). Instead they go-to analytics specialists (who are already busy), or use what they are comfortable with, spreadsheets. (with their error rate)
A further consideration - in-house analytics experts want to spend time on challenging analysis, not repetitive reporting, or collecting, cleansing and organising data. With plenty of analytics roles out-there, these teams need very effective management and challenging roles, or they will be off.
Choose your approach well. Don’t start investing large sums in company-wide roll-outs of new technology, or building-out data scientist teams, as your talent pool will need a mix of capabilities and specialisms. Start small, learn, generate value and build-out.
4. Build-out the analytics – there are 3 inter-related ‘types’ of analytics, each can improve performance and decision making, and each can be sub-divided, however the summary is
- Descriptive – what has happened
- Predictive – what could happen based on previous patterns
- Prescriptive – what should I do
Descriptive analytics can generally be visualised in simple (but valuable) charts, and is important in all processes to ensure the data being analysed is accurate and all parties involved agree the output is robust. It simplifies huge amounts of data into useful ‘bite-sized’ nuggets of information at an aggregate level.
Predictive analytics can be simply described as adding trend lines to the historic data and performance. It’s a ‘crystal ball’ that visualises what could happen in the future based on past performance. To improve on this, statistical and machine learning algorithms can be built to pool historic data from across datasets and develop effective outcome analysis – good for root-cause, what-if, forecast corrective actions and even develop business strategy. Retailers use predictive analytics to identify your buying habits and offer you promotions.
Prescriptive analytics is the next step on. In essence, it uses simulation and optimisation to answer the question ‘what should I do’. Prescriptive analytics is a combination of data, mathematical models and business rules - exploring many possible actions suggesting action(s) based on the descriptive and predictive analytics. An example is self-driving cars – the car makes multiple decisions based on predictions of future outcomes – left or right at a junction based on destination and constantly changing variables such as traffic conditions, traffic signals, pedestrians etc. This type of analytics is comparatively complex and therefore expensive, but can have a major impact on business.
Start with the descriptive analytics. Get comfortable it is robust and accurate. Create context through merging data sets for enriched insights. Share the workings and outputs with impacted departments and generate the insights and actionable outcomes that allow for better decisions.
5. Measure the results
Having made your contextualised, data-driven decisions, ensure you are obtaining the anticipated benefit. Having built your analytics process, data and insights can be refreshed to match your requirements. Retailers may want live sales data from the EPOS system, manufactures may require machine operating performance and potential failures flagged live, whereas a monthly refresh may be sufficient for operations or procurement to manage costs or HR improve employee engagement. Whatever the frequency, the tools are there to enable you to react as quickly as your requirements determine.
Analytics has been around for years. However, the significant increases in the amounts of data businesses produce and the advancements in the technology available mean businesses have to re-think their approach or they will quickly lose-out in todays environment.
Following these five steps will allow business and individuals to make better, more effective and quicker decisions. Don’t get stuck in planning mode for your analytics approach – things are moving to quick for that. Agility is the order of the day, driven by agile analytics.