案例分析题

Retail World (RW) is a major international retail chain, selling groceries, clothing, electronic items, toiletries and homeware items. It has grown rapidly across a number of different countries, offering a broad product range to suit a wide range of customer segments. Growth has been through the expansion of existing stores, in addition to the opening of new stores.

The new finance director, whose background is in a non-retail environment, is keen to understand the sales trends of the organisation, as well as the industry, in order to help develop a strategy which can take advantage of these trends in the future. A business analyst has provided summarised internal sales data for this purpose. The company’s IT systems are fully integrated and associated controls are rigorous, allowing the data to be manipulated in many ways. The analyst has provided time series analysis (Figure 1) and regression analysis (Figures 2 and 3). He wanted to explore the possibility of identifying different causal relationships, carrying out least squares regression analysis linking sales to both time (Figure 2) and number of stores in operation (Figure 3).

The number of stores has grown annually and the analyst believes that this is a better indicator of the expected future revenue than simply the passage of time. The average number of stores expected to be in operation in 2017 is 3,700 rising to 4,000 in 2018.

Notes:

(1) Svar is the expected seasonal variation, calculated by averaging the variations of each quarter

(2) Sadj is the sales total, adjusted for average seasonal variation.

Figure 2 – Least squares regression analysis (time)

An extract of the values used in the calculation of this is shown below. This illustrates the basis of the regression analysis. All values were used in the calculation of the least squares regression formula.

Figure 3 – Least squares regression analysis (number of stores)

The values used in the calculation of this are shown below:

问答题

Analyse the data given above in the time series analysis and least squares regression analysis (Figures 1, 2 and 3) and evaluate the appropriateness of each technique in forecasting future sales and developing strategic plans.

【正确答案】

Time series analysis uses a moving average to define a trend. In Figure 1, the moving average is upwards in a year-on-year basis, with each quarterly result being greater than the equivalent quarter in the previous year. However, that trend does fluctuate on a seasonal basis within each year, with Q1 and Q4 showing positive seasonal fluctuations and Q2 and Q3 showing negative seasonal fluctuations. Given the seasonal nature of the business, it would seem appropriate to use this as a method of forecasting sales.

Although there is no single agreed method of extrapolating the trend, it could be suggested that it appears to be growing at about $2m per quarter over the last few quarters. If this were to continue, estimates for 2016 Q4 and 2017 Q1 would be as follows:

2016 quarter 4: 134·0 (2015 quarter 4 figure) + 8 (growth over the year) + 14·60 (seasonal adjustment) = $156·60m

2017 quarter 1: 136·0 (2016 quarter 1 figure) + 8 (growth over the year) + 25·02 (seasonal adjustment) = $169·02m

The relatively low residuals would suggest that the method of forecasting is reasonably accurate and that residual factors have little effect.

Although this method seems appropriate when accounting for the seasonal nature in forecasting, it does not take into account where the sales are growing. It is mentioned that business growth is through both expansion of existing stores and through the introduction of new stores. It would be worthwhile understanding which is having the greater effect. The analysis also does not take into account external trends such as overall industry growth as the data used is completely internal to the company.

The least squares regression has more varying results. Because of the highly seasonal nature of the data, the correlation between time and revenue does not seem particularly useful. For example, if the formula were to be used to predict 2016 Q4 revenue, it would suggest $138·08m ($110·93m + ($1·81m*15)) which is much lower than time series would suggest, which takes seasonal factors into account. The 2017 Q1 revenue would be predicted to be $139·89m using this method, again much lower than that predicted using the time series.

The correlation coefficient ‘r’, with a value of 0·33, suggests that the two variables are weakly connected. The coefficient of determination (r2) suggests that 11% of the variation in sales (y) is due to the passage of time (x). This would indicate a weak non-linear relationship, as suggested by the seasonal variations.

Therefore, this approach would not be considered an appropriate method of forecasting.

The least squares regression considering number of stores and revenue is more closely correlated. With a correlation coefficient of 0·94, there is a strongly positive relationship. Indeed, the coefficient of determination suggests that 88% of the variation in sales (y) is due to the variation in store numbers (x). This would seem to add further insight into the information provided in the time series analysis, suggesting the increase in store numbers is the greater driver for growth.

Using this method, the analyst could forecast average quarterly sales for 2017 and 2018 as $143·5m (69·50 + 0·02*3,700) and $149·5m (69·5 + 0·02*4,000) respectively, given the increase in store numbers predicted. However, this approach makes no attempt to take into account seasonal variations. Additionally, this method is using a very small data set, which does not provide trustworthy results.

Overall, therefore, given the seasonality of the industry, it would appear that time series would be the most appropriate approach for forecasting future sales. However, it may be that a combination of methods is used to extrapolate the general trend so as to maintain relevance for future periods.

There also remains the issue that the data presented for all three figures is somewhat limited in that it allows only a summary forecast of revenue for the entire organisation, rather than incorporating segmentation or external data, for example. Additionally, only limited, historical data has been used to determine relationships.

These may be partially resolved by the use of big data.

【答案解析】
问答题

Discuss how three ‘Vs’ of big data (volume, velocity and variety) could be used to enhance strategic development within RW.

【正确答案】

Big data is a generic term used to describe the exponential growth of data, provided from numerous sources, available to organisations. The data is not useful in itself, it is the analysis of such data which provides valuable insights to an organisation. The finance director is right to be interested in this, as it can lead to an in-depth insight into trends and the driving forces behind those trends.

The three Vs of big data, volume, velocity and variety, can be examined to determine their contribution towards strategic development.

Volume can enhance the understanding of customer requirements and behaviour. The more data available, the greater the reliability of the trends and relationships discovered. In the analysis provided, there was a limited volume of data, spanning less than three years and incorporating only two variables. The use of big data would allow multivariate analysis over a greater time period or a greater number of shorter time periods to understand purchasing patterns better. This could help RW to create better strategies to capitalise on discovered trends.

Velocity refers to the speed of use of real-time data. As the majority of business transactions are now carried out using technology, these transactions can be captured and processed in real time if sufficient processing capacity is available. This ensures that strategies can be continually updated, in order to deliver competitive advantage. For example, as a new product is trending on social media, RW may then ensure they stock this product and aggressively market it in order to capture greater market share. Similarly, when customers are shopping online, RW could analyse their transactions in real time and use current and historic customer information to make recommendations for further purchases.

Variety refers to the different sources from which data is provided. As sources take different forms and include those not in RW’s control, this is a challenging aspect of big data. However, if managed correctly, the variety provides the most detailed understanding of the market place, segmentation and individual customers. This could include competitor and industry information, sourced through key words online, to hashtags on social media and discussion forums.

There are many potential benefits which could be obtained through the analysis of big data. RW could use the results to determine where to locate their new stores. By accessing customers’ shopping habits from credit and debit card records, they could determine which competing stores are used, and in which locations. This could help in the strategic planning of store locations, especially as RW is intending to continue to grow store numbers, at least over the next two years. This could help maximise the additional revenue to be gained from new stores.

RW are clearly trying to identify trends and maximise the use of them. The use of big data will provide more reliable and robust trend analysis and could lead to the discovery of previously unsuspected trends, allowing RW to capitalise on these before its industry competitors have even recognised the trend.

Further revenue streams are also available through the selling of data. Given the industry RW is in, there will be a number of branded items on offer to customers. Manufacturers of these brands are keen to carry out their own analysis and will pay for information to help with this. RW could capitalise on this new revenue stream.

Overall therefore, it would seem that the finance director is right to consider the use of big data. Indeed, RW may well find itself at a competitive disadvantage if it fails to do so. However, as with all decisions, the cost-benefit implications would need to be considered before implementation.

【答案解析】