Getting sourcing decisions right first time every time is not a luxury but a necessity in the unpredictable, fast moving retail industry. But it is incredibly hard to make the right buying decisions, because there are so many variables – cost, lead-time, response to consumer trends, likelihood of mark-down, variability of demand and forecast accuracy. Each of these are difficult to analyse, and there is an element of chance in changing consumer tastes and factors like the weather. It can seem as though the odds are stacked up against the retailer.
At the same time, there are many buying options for retailers, especially given the changes to quotas on clothing and other products from China. Quotas were lifted at the start of the year, but then re-imposed, stopping product from entering Europe recently. Even though this trade dispute was resolved quickly, it highlighted the risks involved in long-distance sourcing. Retailers therefore often need to weigh up the higher cost of local suppliers against the lower cost but longer lead-time of Far East sourcing.
These decisions have a huge impact on retailers’ success. Take fashion retailers like Zara, who have based their entire business model on a fast and reactive supply chain, where product is manufactured in Europe. Other retailers, such as Marks & Spencer, have shifted much sourcing to the Far East, enabling them to offer more competitively priced products while maintaining or raising margin, and hence improving their financial results. But others have bucked this trend. For example, value-clothing retailer Peacock Group halved lead times and boosted results by moving most product sourcing from the Far East to Europe and the Near East.
So how can buyers make the best sourcing decisions to beat the odds? Buyers need to complement their own experience and expertise by using effective decision support tools. These should not dominate sourcing decisions, but can provide insight and understanding to improve them.
PA Consulting Group has developed a sophisticated model, which evaluates the profitability of different sourcing decisions. It incorporates factors such as forecast accuracy, sales volatility, and mark-down policies, as well as the buying fundamentals of product cost and lead time. The model simulates the selling season for a product, and by running thousands of scenarios, which incorporate random sales variations, it shows the range of outcomes that would occur, depending on how sales evolve.
The model reveals surprising insights for some scenarios. For example, the cost premium worth paying for local sourcing is less than 10 per cent, even for products with unpredictable sales. In other words, even if local sourcing is only 10 per cent more costly than Far East supply, it may still not be justified. For some sales patterns, the optimal initial order should be as much as 80 per cent to give the highest gross profit, even though this represents a large irreversible commitment against an uncertain forecast.
The results of other sourcing decisions match buying experience and expectations. For example, short lead times for re-orders allow the retailer to react far more effectively to higher or lower sales than forecast, maximising margin and sell-through, while minimising stock-outs or over-stocks. What the model shows, however, is just how much this greater responsiveness is worth, which means buyers can decide between different product prices and lead times more accurately.
The retail model has many variables that can be changed and so there are literally thousands of different product and sourcing scenarios that can be tested. In the examples that follow, we have kept many of the variables unchanged, so that we can focus on one or two factors at a time. The examples look at a 12-week selling season – relevant to product categories where new lines or styles are introduced several times per year, such as electricals, fast-moving consumer goods and seasonal home and garden products.
We look at two key results for each scenario: gross margin and the spread of outcomes. All the examples use the same base forecast, to sell 100,000 units in 12 weeks, with a full selling price of e20 and a unit cost of e10. So if that number of units is sold with no markdown, the gross margin will be e1 million.
The spread of outcomes is an important indicator of risk. If two buying strategies generate the same expected gross margin on average, but one can range from great success to total failure, whereas the other results in more consistent sales and margin, most retailers would find the latter easier to manage and better for their business. We have taken the result for the 10th and 90th percentiles to indicate the spread of outcomes. In other words, in one in ten outcomes, the sales and margin results will be equal to or worse than the lower figure, and in one in ten outcomes, the results will be the higher figure or greater. We have therefore excluded the really exceptional outcomes.
Our conservative base case has a four-week lead time to re-order product, giving the retailer the ability to react to the first few weeks of sales in deciding the reorder needed later in the season. The average gross profit achieved is e838,000 – versus the ‘perfect’ result of e1 million.
The spread of outcomes is surprisingly large, even though the sales variability is set to be low. In one in ten cases, gross profit is e672,000 or less, and in an equal number of cases, it is above e1 million. In other words, even excluding the one in five more extreme outcomes, the results range from 20 per cent below to 20 per cent above the average outcome (referred to as 40 per cent spread), and an even greater difference when compared with the original forecast gross profit.
The full distribution of outcomes is shown on the chart above.
The benefits of a world class supply chain is well recognised and we tested the concept further by comparing scenarios where lead times for re-ordering product varied from two weeks to seven weeks.
Reducing the lead-time in the base case, from four weeks to two weeks increased the average gross profit by only two per cent. This relatively small improvement reflected the fact that a four-week lead-time already allowed the re-order to react to the first few weeks of sales, and so reducing the lead-time further only gave marginal extra accuracy through two more weeks of sales data.
A balanced view
On the other hand, if the lead-time increased to seven weeks, the gross profit fell by almost 10 per cent. In this case, the lead-time became too long to allow the re-order to use any actual sales data. However, the fall in profit means that the product cost does not have to be massively lower to justify Far East sourcing. The lower cost of goods outweighs the potential cost of lost sales or surplus stock. Other factors also need to be considered when selecting suppliers, such as product quality and supplier reliability.
There is a much greater contrast when we look at the spread of outcomes. As noted earlier, the gross profit for the base case resulted in a 40 per cent spread. When the lead-time is reduced to two weeks, the spread of outcomes is still 40 per cent. But if the lead-time increases to seven weeks, the results can vary much more, with a 70 per cent spread.
We have looked at the profit impact of different reorder lead times for a product with a moderate level of sales variability, indicated by the forecast being at least 80 per cent accurate nine times out of ten. How do the results vary for volatile or predictable products, which are harder or easier to forecast accurately?
The more volatile product achieved five per cent less gross profit than the base case, and its spread of outcomes was much higher, at 58 per cent compared with 40 per cent. In contrast, the more predictable product generated 10 per cent more profit, and had half the spread of outcomes, at only 22 per cent. These results show that improvements in forecasting ability can significantly increase profits. While some aspects of product sales can never be predicted 100 per cent accurately, retailers need to consider every possible indicator of future sales if it can add just a few percent to forecast accuracy.
We also looked at how a longer re-order lead-time coped with changes in sales variability. For the more volatile product, a seven-week lead-time meant the gross profit was eight per cent less than for the fourweek lead-time. But the spread of outcomes was much higher, at 88 per cent. For the more predictable product, a longer lead-time reduced profit by six per cent, and increased the spread of outcomes to 36 per cent, versus the much tighter spread of 22 per cent for the 4-week lead time. These results illustrate again how a longer lead-time not only reduces profits, but also makes the management of sales and inventory far harder, due to a much bigger spread of outcomes.
The four scenarios are summarised, and the graphs show visually the distribution of results. (Many outcomes for the 7-week lead time show gross profit of e1 million because the re-order is committed before sales are known. Therefore in any outcome where sales would have been greater than the total forecast, the maximum units sold are the original forecast.)
The results above compare just a few sourcing strategies for one or two types of product. There are many other fundamental questions to test, for example: What strategies are best for high-margin products versus low-margin? What is the optimal size for the initial commitment? How can air freight best be used to achieve low product cost, while enabling reorders to react to occasional big sales fluctuations?
What impact do all these factors have for shorter or longer selling seasons?
We found surprising answers to some of these questions. For example, for some sales patterns, an initial order of 80 per cent of the season’s forecast gives five per cent more gross profit than an initial order of 70 per cent, even though this represents a bigger irreversible commitment. By making a bigger initial order, there is sufficient stock to allow the first week or two of sales to be measured, before placing the re-order. Although sometimes this could leave surplus stock even if no re-order is placed, on balance the value of knowing the first few weeks of sales is greater.
The results from PA’s model demonstrate the value of complementing buyers’ own experience and expertise with the use of effective decision support tools. These should not dominate sourcing decisions, but can provide insight and understanding to improve them.
We suggest that retailers initially use this sort of model as a training tool, enabling buyers and merchandisers to compare their experience and instinct with the results of thousands of runs. Seeing rapid results – within one minute – is rewarding, and prompts more questions to test other possible buying and re-ordering strategies.
It is also possible to incorporate real sales and sourcing data, to review current and planned sourcing decisions, and to tailor the model to the specific business environment for a retailer, or indeed for buying organisations in other sectors. At first, this can be done for historical data from past seasons, to confirm that the model matches the retailer’s actual results in known situations. Having seen the benefits of the model, buyers are then likely to want to use it as input into decisions for the current and future seasons.
Buyers and merchandisers are strongly motivated by a passion for their products, the excitement of making deals and the goal to maximise sales and margin. Even the most analytical minds find it hard to weigh up multiple factors such as cost, lead-time, consumer trends, mark-down, sales variability and forecast accuracy. The right support can improve the performance of these teams, while maintaining their motivation and passion for retail. n Alastair Charatan is a consultant in the Retail & Leisure Practice at PA Consulting Group
- The cost premium worth paying for a short lead-time may be less than expected – in other words, the risk of surplus stock or lost sales for products from the Far East is often outweighed by the cost savings
- Consider risk and uncertainty, as well as cost, which make managing sales and inventory far harder – the spread of outcomes is large, even for short lead-times and predictable products, and can become very high in other situations
- Reducing re-order lead-times to be as short as possible only gives a marginal benefit – most of the benefit comes as long as you can react to the data from the first week or two of sales
- More predictable products have significantly higher contribution and less risk – hence consider every possible indicator of future sales if it can add just a few percent to forecast accuracy Increasing initial orders can have the same effect as reducing lead-times, and raises gross profit – because this allows time to react to the first few weeks of sales