This paper reviews the literature within  the use of Adaptive Large Neighborhood Search (ALNS) for the Split Delivery Vehicle Routing Problem (SDVRP) and related Vehicle Routing Problem (VRP) cases. Across many settings such as time windows, split deliveries, multi-depot, two-echelon networks, EV charging, parcel lockers, zone pricing, synchronized visits, and loading rules, ALNS usually finds solutions close to the best-known and sometimes new best solutions. It also runs in a short time and works well on large, realistic instances. The papers were grouped into five areas: core SDVRP, network extensions, practical innovations, green goals, and ALNS design. Notably, the right destroy, and repair operators make the search faster and more stable. Still, there are clear gaps. Most studies test only on benchmarks instead of real retail data. Few models look at cost, service, and emissions together. It is not always clear which operators help. Only a small number of papers deal with quick replanning when plans change. It is worth noting that it is recommended to use real cases, testing operators, sharing code with common KPIs, and building models that combine cost, service, and emissions. In general, for retail replenishment and last-mile delivery, ALNS is practical and strong.