Customer segments

Second, data was analyzed from the customer segment perspective. From the da-ta, different customer segments A, B, C and D were selected. Then mean order and process lead times were calculated in the same way as for individual custom-ers Kisu, Misu, Sisu and Visu. Earlier shortening of the data according to the

cus-tomer segments indicated how the two production lines were actually divided to serve different customer segments. The descriptive analysis in Table 10 indicates that production line Alfa serves customer segments B and D, and Beta serves A and C. In the following analysis the customer segments and production lines are studied together. In this way, the study was able to compare not only the differ-ence between the customer segments, but also between customer segments that are produced in the same and in different production lines.

Table 11. Mean lead times indexed for different customer segments.

Customer segments B D A

Number of orders 178 88 88

Customer order lead time (COLT) 1.00 1.23 1.04

Std. Deviation of COLT 1.97 1.33 1.00

Total throughput time (TTPT) 1.00 1.14 1.15

Std. Deviation of TTPT 1.59 1.01 1.00

Production, packing and invoicing lead time (PPILT) 1.00 1.50 1.12

Std. Deviation of PPILT 1.00 1.64 1.11

Production lead time (PLT) 1.00 1.20 1.54

Std. Deviation of PLT 1.82 1.00 2.46

Production line Alfa Alfa Beta

As Table 11 shows, customer segment B had the fastest lead time means in all the areas measured. Having the smallest COLT index indicated that customer seg-ment B was offered the fastest lead time among all the customer segseg-ments meas-ured. Customer segment B also had the most orders delivered, with nearly 50 per-cent of the studied orders, and almost 67 perper-cent of all the orders for production line Alpha. These issues were also confirmed in the interviews, where it was claimed that shorter lead times were offered to high volume key customers. Here, the high volume was coming from a certain customer segment, but the expected outcome was the same as from the individual customer.

Surprisingly, the second fastest customer segment was not produced on produc-tion line Alpha. Instead, it was assembled on producproduc-tion line Beta for customer segment A. Customer segment A had the highest order volume on production line Beta. It had nearly 93 percent of the analyzed orders on Beta, and around 24 per-cent of overall orders. It was interesting that the second fastest customer segment A was indexed as notably slower on TTPT, PPILT and PLT than customer seg-ment B, but had only a slightly bigger COLT index value. All in all, the measured lead time indexes for customer segment A were ranked as the second fastest, ex-cept for PLT. PLT for customer segment A was the slowest of the three tested segments. In practice, this meant that products produced for customer segment A took the longest mean time to be assembled on the production line, indicating that

time-based flexibility was present in earlier processes like engineering and pro-curement.

Customer segment D was served with the longest COLT. Products for customer segment D were assembled on production line Alpha. The number of orders from customer segment D was equal to customer segment A. With 88 orders, customer segment D had slightly more than 33 percent of the overall volume produced on production line Alpha. From the three customer segments, customer segment D had the highest COLT index, indicating that it had the highest mean COLT. Cus-tomer segment D was also the slowest on PPILT, but slightly faster on TTPT and notable faster with PLT than customer segment A. PLT was indexed at 0.20 units slower than the fastest segment B, but 0.34 units faster than the second fastest customer segment A.

The difference between the two fastest mean COLT values for customer segments B and A was not that significant. Interestingly, the case firm representatives in the interview also argued that these two production lines were distinguished by a fast and slow category. This seemed to be the case. Production lead time (PLT) had a significantly higher mean COLT value for customer segment A assembled on line Beta than for customer segments B or D assembled on line Alfa. Despite the longer PLT, the mean COLT for customer segment A was nearly as fast as it was for customer segment B. This indicated that the time-based flexibility had been created in processes other than physical production such as engineering and pro-curement.

The ranking based on the mean COLT index values indicated that different cus-tomer segments were served with significantly diverging lead times. As for indi-vidual customers, the flexibility of offering different COLTs was tested with the standard deviation variation of the COLT values. These indexed standard devia-tion values in Table 11 indicate the same as in the analyses for individual custom-ers. Here, customers with the highest volumes were offered the most time-based flexibility for the COLTs. Similarly, customer segments with the highest volumes were offered the most time-based flexibility on COLTs. The indexed mean COLT variation was nearly double compared with customer segment A, which had the second smallest mean COLT index. Thus, it can be argued that the case firms delivered similar products with significantly diverging order delivery lead times for different customer segments.

Kruskal-Wallis test on customer segments

The Kruskal-Wallis rankings in Table 12 gave more weight to the interview based claim that high volume customers had more flexibility on COLTs. The highest volume customer segment B from production line Alfa had the smallest mean rank, indicating the fastest COLTs. The second highest volume was divided be-tween customer segments A and D. Customer segment A was produced on pro-duction line Beta and D was produced on the same propro-duction line as B. In Table 11, customer segment A was ranked with the second smallest mean rank value, indicating it to be the second fastest customer segment. Customer segment D was third, and C fourth.

The Kruskal-Wallis test statistics table indicates a significance level of 0.000, which means that there is a 0.0% chance of obtaining a rank-difference chi-square equal to or greater than that observed (30.575) by chance. This means that the ratings of the referendum issue do differ significantly by media.

Table 12. Nonparametric Kruskal-Wallis test results on selected customer segments.

Thus, this provides sufficient evidence to reject the null hypothesis on customer segment level and conclude that there is a statistically significant difference in distribution of the COLTs between different customer segments.