By Teresa Hooper
Big Data Small Data – Lost in Data
Data has always been with us, and people have always collected data in one way or another. In the past data was collected and processed by computer bureaus, punched into cards and fed into gigantic computers. In the 1980s there was the revolution of small computer calculators and even PDAs (the Psion Organizer in 1984) coming on the market. By the 1990s we had computers on our desktop.
Throughout all this time we were collecting data, but often we were not doing a lot with it.
One of today’s tablets (in the 2010s) holds 100,000 times the data that a punch card in the 1970s held.
The sheer volume of data you are exposed to each day can be confusing.
One framework that can be used when looking at data is shown if Figure 1:
Figure 1: A Data Framework
It is a good idea to apply this framework whenever you are developing new business models that are reliant on big data.
Expectations of Our World of Data
Data has grown exponentially over the last half century. And there is no sign that this growth is going to stop any time soon.
Some data expectations over the foreseeable future are summarised in Figure 2.
Figure 2: Our World of Data
There really are buckets of data.
Relative Sizes of Data
A guide to come common data measurements (and file sizes) goes as follows:
Kilobyte = 1,000 Bytes – Small audio book
Megabyte = 1,000 Kilobytes – Novel
Gigabyte = 1,000 Megabytes – Movie / Video
Terabyte = 1,000 Gigabytes – X-rays in hospital
Petabyte = 1,000 Terabytes – Years and years of data
Exabyte = 1,000 Petabytes – Large amount of data from multiple sources
There are three levels of data usage:
1. Descriptive – reporting on the past
2. Predictive – this data can be used to predict what may happen, e.g. budgets
3. Prescriptive – this models optimal behaviours and actions
Any data model has to embrace all three types of data. Figure 3 shows how the three types of data interact:
Figure 3: How is Data Used?
You work with data to provide an outcome. Data can be thought of as:
• Providing value to the business if used in the correct capacity
Most importantly, data needs to provide an outcome that provides value to your customer to build their loyalty.
Types of Data
Accenture suggests there are three types of data:
1. Raw data – data that can come from many sources but has no real relationship
2. Managed and refined data – you have managed to put it into some sort of structure inside a database, and therefore are able to report on it
3. Fully refined data – you have used the expertise from your company, outsourced consultants etc. to give you real insight into what the data actually means
The Process of Incorporating Data into Your Business Framework
Some Checklist Questions for Getting Started
1. What is your strategy for wanting to participate in big data?
2. What pieces of information can you identify or discover that may be useful?
3. What platform are you going to use? Will you store it in Excel, a database, or something totally different?
4. Who will be part of this journey (i.e. who puts this data together and analyses it)?
5. Who will rely on this information? How will it be put to use?
Listen out for suggestions from staff and other interested stakeholders. Form your partnerships. Also, start small and build up – set the right expectations.
The Human Aspect
People are an important part of making data work. The human aspect comes from all areas of your business, as you have people with different specialities and interests who have an interest in the same pieces of data.
This can be modelled as follows:
Figure 4: The Human Aspect
As can be seen from Figure 4 there are a number of different users of any organisation’s data. These include:
• The thought provoker, who is often the project champion (but not always so)
• The business analysts
• The data scientists (if you do not already have these in your company you can outsource this work)
• The marketing team who test the customer marketplace
• The IT department, who physically house the data, and who operate the data strategy
You cannot underestimate the human factor involved in processing Big Data.
You could have specialist people for each of these roles, or alternatively you could have 1 or 2 people with all these skills. It all depends on what you want to achieve. One possible problem, however, is that the data can possibly lead you to dead-ends – it is too easy to spend a considerable amount of time analysing data for no results. Therefore you need to have a clear strategy – and no fear of changing that strategy.
There will be sceptics. Use these people – they will keep you honest, in that they ensure that the data provides relevant outcomes for the dollars invested.
The stakeholders’ combined contributions to the big data ‘mixing pot’ enable you to tackle the data in an all-encompassing way.
The Data Framework
Putting a framework around your data is essential. Doug Laney uses the following approach:
Figure 5: The Data Framework
Volume represents both an opportunity and a challenge. With more devices and much more information we are faced with a huge volume of potential insights and discoveries. It can be a burden to manage – investigation and evaluation is key here.
Velocity represents the fact that there is a tremendous opportunity to collect data in real time. Thus, point of interaction opportunities (POIs) need to be put into place to enable this collection, and so that you know when things happen. Of course this presents quite a challenge – as being able to analyse in real time is preferred, but is in practice difficult to achieve.
Variety represents the fact that there is a considerable number of different types of data that come into existence every day. We need to understand how the different types of data we obtain complement or talk to each other. Again, the sheer volume of these different types can be daunting.
The Outcome represents the fact that if we can manage this process, collecting, managing and sifting the sheer volume, velocity and variety of data received, we have huge opportunities.
There are a number of challenges faced by organisations dealing with big data:
1. They need to identify what data is actually available
2. They may need to integrate multiple data sets from several sources. A decision has to be made, therefore on what data is needed and where it will come from
3. They need to work out how to store the data and then deliver it back with a structured meaning
4. There may be a need to change thought processes.
5. There may be possibilities to develop new offerings to your customers through the data. Two examples of this are LinkedIn and Google. Both companies use their data to get to know their customer base. LinkedIn uses their data to try and match up people with similar interests. Google is constantly revising their search algorithms and developing new products and services. Past examples include Gmail, Google +, Google Apps and most recently Google Glasses. In both companies’ cases, they are always prepared to take a different path if their data shows that the results are not paying off.
“IT/Data will solve all our problems and get our costs down….”
Data on its own does not solve the problems of a business or keep its costs down. It will cost dollars to build a data solution – how many will depend on what the business is trying to achieve. Every business should expect a return on investment from dollars invested, and data is no different from any other type of investment. It is all about what is good for the business – remember the main goal is to improve profit. Savings generated by data may be qualitative rather than quantitative.
Take the time to work your way through a strategy list for your business. You should consider the following points:
1. Do you have access to data?
2. Do you have a process to keep the data?
3. Do you have the right people to build the data warehouse?
4. Do you have the right people to analyse the data?
5. Do you know that your customers will want the products and services delivered / identified by big data?
6. Do you have the right people to implement the outcomes?
7. Is the culture of the business correct for big data?
Case Study 1 – A Transport Company
The aim was to reduce accident costs from $1M plus.
A particular transport company had a turnover in excess of $20M, but it had accident costs in excess of $1M, i.e. over 5% of its turnover went on accident costs.
Analysis of the accident data showed that the causes of the accidents were operational through driver error. This suggested that employing good drivers who were introverts would reduce costs, i.e. it would be advantageous to employ drivers who liked to be on their own for periods of time.
A further examination of the evidence suggested that most of the accidents were caused by drivers who were extroverts, who really wanted to get to the next stop so that they could socialise.
It was decided to introduce DISC (task profile) tests. To enable this they ran all of the drivers through the task profiles. They produced a past history for each driver. Relevant data relating to the trips the drivers took was accumulated: kilometres, hours, routes etc. They then recorded the accident information against each driver for the next 12 months.
It became very clear from the data that the introverted drivers were less likely to be involved in accidents than the extroverted drivers.
Having determined this, the company introduced a policy of only employing those who met the profile suitable for long distance driving. It did not take long for the culture of accidents across the company to change, and indeed the annual accident costs fell from over $1M to below $200,000.
Interestingly there were a few other clear and obvious statistics that were discovered. Some costs could be easily traced to particular drivers, so it was easy to analyse these particular costs. The average litres of fuel per kilometre was higher for the introverted drivers. Tyre costs improved.
The company built the data as it went along – looking for new outcomes and improvements to the overall performance of the company.
This is an example of playing with small amounts of data to discover conclusions. This was happening in the 1980s so use of data (big or small) is not new, it has always been done.
There was an element of a culture change in the organisation, from managers to drivers to mechanics. It was very much a case of those who were preparing the data being able to present it back in a format which could prove the results and therefore get the buying-in from the employees.
More recently, of course, there have been truck computers giving this kind of information, having it at the fingertips of drivers and executives, instantly through GPS and the like. However, the qualitative data provided by the task profiles did enable other revenue-generating outcomes to be discovered.
The Data Journey
There is definitely a data journey. Data is fed through a typical data warehouse environment as follows:
The journey begins with the raw data, e.g. business analysis, the team, ideas, ERP, a loyalty club. This raw data goes through the data warehouse which refines if, using such tools as SQL, NOSQL, Access, Excel and .net C++.
Once the data has been cleansed in the data warehouse there is the reporting function, which includes modelling, What If analysis and business intelligence.
Finally there are outcomes which put these reports into action. The process is then constantly reviewed.
Of course, now that you have all this cleansed and filtered data, the question is what to do with it? Nowadays the answer is that it is delivered to a myriad of devices – smartphones and tablets being the most common devices to view the massaged data on.
Data journeys are complex. It is beneficial to start out by labelling the outcomes and results you need on a level-by-level basis.
Figure 6: The Data Journey
As can be seen, different businesses will have different levels of data analysis. The Transport Case Study demonstrated how Level 1 data can lead to Level 2 data mining to achieve specific positive outcomes for the business.
The data journey may seem complex, but don’t let that deter you from jumping in and getting started.
Case Study 2 – The Retail House (a Pharmacy)
A retail house can be represented by the following model:
Figure 7: The Retail House
The key considerations of the model are:
• Location – where and when do I set up shop? Am I in the right location?
• Products and services – do my customers want the products and services I deliver?
• People – do I have the right people on board – with the knowledge and skills to complement my offering?
• Value – does the offering provide value to the customers, for which they are prepared to part with their dollars?
• Marketing – how am I going to market my products and services to my customers?
There is also an inner triangle supporting the retail house:
The particular retail house used in this case study is a pharmacy.
Like most businesses a pharmacy can have four stages in its life.
The aim of the work I do with pharmacies is to avoid them going into the fourth stage – death by customer disloyalty.
Pharmacies are the main place that people receive their pharmaceutical drugs, both with private scripts (where the government has not contributed to the cost of supply) and PBD scripts (where the costs are subsidised by the government).
They can be located in shopping centres, strips, regional towns, but they must be owned by pharmacists.
Their annual turnover ranges from $500,000 to $10,000M plus.
The range of products and services offered varies between pharmacies, depending on whether they operate under a brand, where they are located and the size of their turnover. The bulk of the profit made by pharmacies is in the dispensing of prescribed drugs, and often the front of the shop (representing up to 80% of the floor space) returns little or no profit. As a result of this they have been forced to move to a health solutions model of serving their customers.
Moving to a health solutions delivery model is not as simple as it might seem as there are numerous road blocks to overcome.
Figure 8: Road Blocks
The road blocks include:
• Changing from a historically successful model which is built on maximising the number of scripts for the minimum labour cost without sacrificing patient safety. Success under this model is usually defined by pharmacists as minimising patient-facing time – this is highlighted most when a pharmacy owner tells you how many scripts he can get done in a day
• As such the alternative model is in conflict as it revolves around increasing customer-facing time by Pharmacists albeit still with the business emphasis on increasing revenues
• Inherent in this shift is changing the common culture of the Pharmacist where the historical preference has been to not be customer-facing unless required.
• Those who seek to defend the historical model will often decry the lack of profitability in delivering services. But this view of course overlooks the fact that services income and the cost of delivery cannot be viewed in isolation. The real business benefits that flow from a services model is that improved patient outcomes usually come when a management and prevention component is connected to the dispensed medicine. Hence the combined solution involves retail products and increased customer satisfaction and therefore ongoing loyalty
• Of course dispensary and pharmacy layouts often work against a services delivery model as they have been designed to protect the pharmacist from customers so he can maximise dispensing time
• To change a model requires significant training and leadership from management to guarantee a successful transition
Changing behaviour is assisted and reinforced by focussing on the right metrics. All pharmacies should focus on:
• Sales per square metre
• Gross Profit dollars per script
• Script and customer growth
• Gross Profit dollar growth
• Wages as a percentage of Gross Profit dollars and Other Income. Note here that the inclusion of Other Income is important as pharmacists should increasingly be spending their time generating services-related income as well as product sales.
• Gross Profit per space, which is measured down to category level needs.
Figure 9: Mapping the Shelf
You can map the pharmacy layout based on GMROS – the return on space, influenced by space, the Gross Profit percentage, unit volume and stock mix. The results of this can be displayed as graphical data to enable the pharmacists to visualise the changes needed within the store.
Figure 10: The Pharmacy Layout
As much as future growth opportunities exist by developing expertise and solutions around specific conditions, there are actually simple opportunities that already exist in many pharmacies that could be converted immediately.
One example is shown in Figure 11 below for a particular pharmacy (Example Pharmacy in the graph). This compares their category performance from seven key health categories, compared with my firm’s client-base averages (JR Pharmacy in the graph). It is clear that the Example Pharmacy is lagging behind the averages in all categories, due to them being too aggressive in their pricing decisions, totalling approximately $30,000 per annum.
Figure 11: Departmental Sales Analysis Based on GP ($)
It was easy to fix the problem faced by the example pharmacy above, but underperformance is not solely a result of incorrect pricing decisions. Buying terms, merchandising and staff skill sets can all contribute but ultimately the data analysis will provide both the necessary insight and the call to action to chase the opportunities.
We capture the information to produce the final KPI analysis in a number of ways, including:
• Tillink – where we extract the data from the Point of Sale on a daily basis, which gives us not only Sales and Gross Profit dollar information, but qualitative data, such as scripts, customer numbers, items sold and hours opened.
• The payroll / roster system – measures the hours spent by staff in Dispensary, Front of Shop and any specialised areas
• Spacelink – information from the Dispense and Point of Sales systems overlaid in the store layout system, to identify potential growth areas, and potentially what may not be correct in the layout of the store
• Financials – we extract information from their financial system (whether it be the accountant-supplied online system, MYOB, QuickBooks etc.)
• Data warehouse – where it all comes together from the different systems and reports back to the owner
As can be seen from this case study, there are multiple points of information that can be gleaned from one business – not just financials.
It is our view that there is both short and long term opportunities existing in the majority of pharmacies across Australia, and we remain optimistic about the future, despite the impending loss of profit through price disclosure. Those with high debt and rent levels are most exposed. There is a need for all pharmacies, though, to address the ongoing industry changes, while also understanding the changing capabilities of bigger retailers, via data analytics, social media and the internet.
Achieving growth and optimising business performance for community pharmacies, and the majority of retailers for that matter, is ultimately about being able to engage the customer on something other than price. For pharmacies this is about pharmacists using the existing dispensary traffic to leverage professional knowledge into health categories. To do this efficiently, many pharmacies still need to address procedures and workflows within the engine room that is the dispensary.
Greater insight is available by simply using the data that already exists in the Point of Sale systems and financial statements. The challenge though is to convert this data into meaningful information which can drive actions and deliver meaningful customer outcomes.
Figure 12 provides an example of the sort of KPI data that can be collected to provide the following benchmarks:
• Information from the dispense system
• Information from the Point of Sale (POS)
• Financial data
• Payroll data – FTEs
There are 38 main KPIs we measure to design our advice to give the best outcome to the pharmacy in question.
Figure 12: Sample KPI Data
One major income-related issue is that every six months the pharmacy’s income from dispensing drugs is falling through agreements made with the government to reduce the PBS budget.
The data sets have been collected for long enough now to identify the key metrics that the store should be focussing on.
Case Study 3 – How Trendfinder Works
Trendfinder helps you see data clearly with no additional data entry work required.
Figure 13: A Model of Trendfinder
Trendfinder uses the data you already collect in POS systems, loyalty and other data. Your data is extracted from your computer system, transformed using the latest analytics software, and stored in your own personal and secure data warehouse in the cloud. Snapshots of your data are stored at points in time and then analysed, using the latest retail analysis techniques. Millions of individual pieces of information across a business are analysed and sorted to find the highest impact areas in the business. Results are presented in an easy-to-read visual form via a secure web connection.
Trendfinder uses traffic light symbols to highlight improvements and potential issues, and provides the ability to drill down to find causes.
Figures 14 and 15 show how you can compare stores against each other, to enable management to have a roadmap to improve the overall business standards.
Figure 15: Stores Can Measure Their Performance Against Others
By using customer mapping against promotional products it is easy to identify which catalogues are working and whether those catalogues are successful in bringing in new customers; thus measuring the results of marketing can lead to cost savings. Some of this data is shown in Figure 16.
Figure 16: Cost Savings in Marketing by Measuring Results
Departmental analysis compares monthly performance against previous years. In the pharmacy case study, drilling down identified a consistent downward trend in one department / branch. A competitor was poaching customers with attractive pricing on a small part of the product range. As a result of this the firm was able to reduce pricing in that area to increase competitiveness and they were able to approach the lost customers directly with the new pricing offer. This is demonstrated in Figure 17.
Figure 17: Cost Savings in Marketing by Measuring Results
The Quick Stock Review feature helps firms identify which ranges and products are overstocked and where they are missing out on sales opportunities by not having enough stock. The result is increased sales with less stock holdings. This analysis is shown in Figure 18.
Figure 18: How Much Stock to Hold
Where to Go From Here
Firstly research where your data points will come from. Surround yourself in knowledge around what others in your industry (and in others) have done with data. Identify the people you want to work with and what skills and knowledge they require.
Four key tips to those willing to go through the process of using the big data around you:
1. Put your team together
2. Don’t be afraid to change tack
3. Continually revert back to the strategy
4. Continually seek feedback from your customer