Optimizing lead time and resource utilization for service enterprises

Abstract

Lead time is generally defined as the period of time for which customers have to wait before receiving completed products or services from enterprises. In the key public utility markets, such as telecommunications market, that are tightly regulated by government watchdogs, service enterprises are obliged to provide equivalent services to all their customers. In the context of service lead time, this means that "standard lead times" should be universally applied to all customers. Existing industrial practice in deriving "standard lead time" is manual and erroneous. Regardless of the variances of demand and resource across geographical areas and across time horizon, service enterprises are generally confined to single fixed "standard lead times". Ill-derived "standard lead times" lead to wrong promises made to customers, which not only forces service engineers to work overtime to deal with overflowing jobs but also incur huge compensation costs due to delayed or failed customer services. In this paper, we set out to tackle the "standard lead time" problem. In particular, we devised an automated approach to testing the existence of legal optimal allocation schemes based on actual customer demand, enterprise resource and assumed standard lead time. We first modelled this as a combinatorial optimisation problem; we then experimented a heuristic local search method, namely "single day shift", inspired by previous work on virtual telescope scheduling; further on, we improved the first method by introducing a new heuristics, namely "multiple day spread", which achieved a speed of search for optimal solutions, 1.29 times faster than the first method. A case study on comparisons between existing default "standard lead time" and the calculated one using our approach for a telecommunication service provider was conducted to demonstrate the necessity and effectiveness of this approach. Finally, we compared our approach with related work. We concluded that our approach is more suitable for the problem of optimizing lead time and resource utilisation, and the results from our work are widely applicable to various industry sectors that concern equivalent customer services, balanced work load and optimal lead times.

Variable neighborhood search for the workload balancing problem in service enterprises

“…For solving the WBP, Li and He [26] propose two greedy search algorithms named Single Day Shift (H1) and Multiple Day Spread (H2). Their experiments show that H2 outperforms H1 in run-time performance, but do not reveal any insight into the solution quality of the heuristics in terms of loss of optimality.…”

“…Similar to the approach in [26], the initial solution in Step 1 is generated deterministically by allocating each daily demand λ i across the time interval between the internal release date r i and the due date d i , as evenly as possible.…”

“…The CPU times for solving three test instances of class 4 by using their heuristics, H1 and H2, lie between 39 and 67 min (as indicated in [26] without any information regarding the solution quality and the used computer), whereas our maximum CPU time is 200 sec. The main reasons for this big gap are as follows.…”

“…The WBP was only once investigated by investigated by Li and He [26]. They proposed two greedy local search algorithms to tackle the problem.…”

“…This paper proposes several improvements to the previously published work by Li and He [26] as follows: Section 2 presents a mathematical model for the WBP, based on which an exact solution can be derived. In Section 3, besides an exact method, two new heuristics based on ILS are proposed to improve search speed and solution quality.…”