South African fast-moving consumer goods (FMCG) companies operate in an increasingly competitive environment (KPMG 2016a:1; Pricewaterhouse Coopers 2012:2). Furthermore, inaccurate demand forecasting is regarded as a main cause of uncertainty for FMCG companies operating in South Africa (Sanchez Rodrigues & Potter 2013).

Demand planning is an operational supply chain management (SCM) process comprising several steps in developing reliable forecasts (Cassettari et al. 2017:49–50). Effective demand planning contributes to the accurate alignment of inventory levels with peaks and dips in demand (balancing supply and demand), meeting customer expectations, remaining competitive and thus improving profitability for a specific channel or product item (Metcalfe 2012:58; Rouse 2010; Supply Chain Resource Cooperative 2004).

Demand management (i.e. balancing supply and demand) is a key factor in improving the efficiency of supply chain operations (Croxton et al. 2002). There is also growing proof that companies that have implemented demand-driven supply chains experience an increase in sales, a reduction in operating expenses and an improvement in working capital (KPMG 2016b:2). The supply chain planning process is often the one with the largest performance gap; it is also the one process most often misunderstood by supply chain and business executives. The process also plays a fundamental role in overall supply chain success and is often a key source of competitive advantage (Bursa 2014; Cecere 2014; Chase 2016).

Balancing supply and demand is a common problem in manufacturing organisations across all industries; it is becoming increasingly important to achieve this balance (Coker & Helo 2016:564; Supply Chain Market 2018). This problem is exacerbated in the context of FMCG organisations because of the high demand for product availability, which characterises FMCG. Therefore, a push-based manufacturing model is required (Bellini 2015; Coyle, Bardi & Langley 2003:224).

Company A is an established FMCG manufacturer, with different business units servicing different FMCG market needs in Africa. When interviewed on 17 May 2016, the supply chain executive of the company disclosed that following a supply chain performance metrics review, it was concluded by senior management of Company A that the company had an underperforming supply chain. The internal performance metrics used to determine supply chain performance in the company related mainly to customer service levels and logistics costs as a percentage of sales. The underperformance was ascribed to inaccurate demand planning; consequently, an intervention to improve demand planning accuracy in the affected supply chain was instituted in October 2013.

From the interview with the company executive, it also transpired that at the time of the commencement of this study in 2016, the company had conducted a baseline review of the impact of its demand planning intervention, confirming the improvement of the respective supply chain in terms of the internal benchmarks used (customer service levels improved by 9% and logistics costs as a percentage of sales decreased by 1.8%). However, the company had not formalised the process changes in detail for future reference, nor had demand planning performance been assessed in detail. Furthermore, a limited number of case studies around demand planning activities within the context of the local South African FMCG sector exist. Therefore, this research attempts to close this gap by providing further insight into the relationship between demand planning processes and supply chain performance at an FMCG company.

The primary research objective of this study was to determine the effectiveness of the demand planning intervention of Company A on the supply chain’s performance, from a comparison of pre- and post-intervention demand planning activities. In support of this primary objective, the secondary objectives were to determine and compare the pre- and post-intervention demand planning activities and performance of Company A.

This study followed a case study research approach that is aligned with a pragmatic worldview approach (Creswell 2009:13). This was an appropriate methodology for this study, as it took place over a defined period of time. Further supporting this choice of methodology is the nature of data collection, which was done from multiple sources using more than one method. The environment under analysis presented rich context-specific data related to the research questions. The focal organisation in this case study is referred to as Company A, as the company gave the go-ahead for this research, provided that it remained anonymous. A confidentiality agreement was signed in this respect.

For Company A, a detailed analysis of quantitative secondary data relevant to 443 SKUs was undertaken to assess forecast accuracy. Stock keeping units were analysed with regard to actual sales, forecast errors or accuracy and their Pareto classifications (i.e. ABC analysis) (Fritsch 2014; Wild 2002:36–40). The analysis considered SKUs individually and holistically, to draw suitable inferences. Stock keeping units were first assessed quantitatively using MAPE.

To gain insight into the demand planning process, pre- and post-intervention primary data were collected by means of interviews with four key personnel from Company A. The supply chain manager was initially interviewed, and supplied the secondary data used to assess demand planning accuracy, pre- and post-intervention. The marketing manager, demand planner and customer manager were interviewed to understand the interventions and how these directly impacted their particular function within the supply chain under investigation. The data were used to assess the pre- and post-intervention performances of the supply chain under investigation, and to determine whether or not the inefficiencies and gaps had been resolved through these interventions as per the study objectives. The only changes made to the supply chain during this period were those related to the changes or interventions detailed within the context of this study.

The results were analysed and categorised into two distinctive phases, namely, the pre- and post-intervention phases. The intervention was implemented in October 2013; therefore, the impact of the intervention is assessed from November 2013 – the post-intervention phase. The number of periods falling into the pre- and post-intervention phases differs, as pre-intervention data were only available for 13 periods (months), while post-intervention data are available for 33 periods (months). Consequently, the analysis includes a direct comparison of the 13 months’ pre-intervention versus the 13 months’ post-intervention. For the purposes of analysis, the full period under investigation is from October 2012 to July 2016 (46 months or periods, in total). The pre-intervention period is from 01 October 2012 to 31 October 2013, while the post-intervention period falls between 01 November 2013 and the end of July 2016, as per the intervention timeline detailed in Figure 2.

All 443 SKUs with a corresponding set of data in the supply chain under investigation were classified using ABC analysis, based on actual sales volume per item for the period July 2015 to July 2016. Based on the Pareto principle (80/20 rule), items with annual sales exceeding 130 000 units were classified as A items; those with a demand of between 130 000 and 60 000 units, as B items and those with a demand below 60 000 units, as C items. Of the 443 SKUs, 45 SKUs were classified as A items, 39 SKUs as B items and 359 SKUs as C items. Such a classification contributes to enhanced insight into the supply chain to assist in more accurately assessing the results of the interventions taken by Company A to improve forecast accuracy.

The MAPE was used as the primary metric for forecast accuracy assessment, as it is ideal for assessing large volumes of data because of its scale sensitivity, and associated ease of comparison thereof. From Figure 2 and Table 1, it follows that the intervention in the demand planning process resulted in an improvement in forecast accuracy for all SKUs, from the pre-intervention period to the post-intervention period, of 2.7% over 13 periods and 7.18% over 33 periods, reducing the MAPE to 14.71%. The improvement was more pronounced at a total SKU level, after the 13-month period, than in the initial stages of post-intervention.

For the A-classified SKUs, the MAPE dropped by 4.9% to 16.1% (Table 2). For B-classified items (Table 2), the forecast accuracy improved, by 10.4% in the first 13 months, post-intervention; at a total level, a 12.5% improvement in forecast accuracy was presented. For C-classified items (Table 2), there was an improvement of 6.3% in forecast accuracy within the first 13 months, post-intervention, and a further improvement of 5.1% thereafter, resulting in a total improvement of 11.4% in demand forecast accuracy since the initial intervention.

The research problem underpinning this study centred around the lack of information available regarding the impact of demand planning interventions on demand planning accuracy for Company A. The primary research objective was to establish whether or not the demand planning interventions of Company A had a positive impact on the performance of the supply chain under investigation. In support of this primary objective, the secondary objectives were to determine and compare the pre- and post-intervention demand planning activities and performance of Company A.

From a MAPE analysis of Company A’s pre- and post-intervention forecasting accuracy, a definite improvement in the accuracy of the demand planning process was recorded at Company A after the intervention process. It is further evident that the degree of improvement differed per SKU classification. B-classified SKUs showed the greatest improvement in MAPE over the 13 periods post-analysis and continued for the full post-period analysis. The C-classified SKUs recorded a similar improvement over the 33 periods. However, the high-volume A-classified SKUs, critical to the business, recorded the lowest percentage in forecast accuracy improvement. Nevertheless, as all three SKU categories recorded a forecast improvement, it is possible to infer that the interventions introduced had a positive impact on the supply chain because of improved demand forecast accuracy.

It follows that the changes in the demand planning process have resulted in an improvement in overall supply chain performance, as evident from a review of the company metrics:

During the research period, no other changes or activities related to the supply chain were instituted. From the comparison of pre- and post-intervention results, it is reasonable to conclude that a relationship exists between the improvement in the demand planning process and supply chain performance, as measured by customer service levels, sales as a percentage of logistics cost and internal supply chain ranking, across the total business.

In support of this evidence, the customer manager confirmed that the demand planning process had, indeed, improved. Similarly, and in even stronger terms, the marketing manager attested to this improvement. The increased visibility and engagement that occurred because of the new demand planning process prompted the marketing team to focus on further improving forecast accuracy.

In conclusion, Company A complied with Sonawane’s (2011) requirements for achieving a best-in-class demand planning environment. These interventions include the following:

For Company A to improve forecast accuracy even further, Myerholtz and Caffrey’s (2014) recommendation could be followed. They suggested the incorporation of statistically driven models that require little to no manual effort. Although this is not Company A’s current practice, in interviews with key personnel they indicated that a move to Demantra (Oracle demand planning software) was imminent in the demand planning process. Demantra is a software tool which offers ‘multi-dimensional data modelling functionality, flexible data hierarchies for up-to-date demand visibility and one number planning’ (Oracle 2015:1). The implementation of Demantra would aid in overcoming the challenges associated with the use of the oversized Excel document.

Insufficient resources remain a challenge for the demand planning team, as only one team member is responsible for conducting all the planning activities for all the SKUs. As a result, pre-demand reviews, which should include the customer team to further improve demand understanding and associated forecast accuracy, are still limited to the marketing team. However, the collection and use of data from several sources are a best-in-class practice that Company A has adopted, although neither order pattern prediction nor a learning model framework is in place yet. The implementation of such measures should be considered in the planning for future improvements in the demand planning process at Company A.

Based on the findings of this research, management needs to ensure that the demand planning team is appropriately resourced and that teams are collaborating to develop forecasts that are indicative of the broader organisation and requirements. Failure to ignore broader market indicators and to have the right software in place can result in adding more pressure to the organisation.

This study has contributed some insights into the relationship between the demand planning process and supply chain performance at an FMCG company. It may be inferred that, because of the demand planning interventions, overall supply chain performance has improved. This causal relationship can be assumed with near-absolute certainty because no other major interventions or measures were put in place during the research period for the specific supply chain under investigation. As such, pre- and post-intervention demand planning activities have been carefully explored and compared.