Advanced TURF: hybrid optimization
Advanced TURF combines exhaustive search, swap optimization and reverse pruning to find better solutions than the greedy algorithm. It is computationally intensive but delivers provably better assortments.
Greedy TURF is fast but suboptimal. Advanced TURF methods attack the problem from multiple angles. Exhaustive search tests all possible combinations — perfect for small problem spaces but exponentially expensive. Swap optimization starts from the greedy solution and systematically tests whether swapping individual products improves the result. Reverse pruning starts with all products and removes the least contributing one at each step.
Reflect's hybrid optimization combines these methods. We use reverse pruning as an independent search path (it often finds different solutions than greedy), swap optimization as polishing of the best candidates, and exhaustive search as validation at feasible problem sizes. The result is that we consistently find better solutions than pure greedy.
In practical tests, hybrid optimization improves reach by 1-3 percentage points compared with greedy — and for volume optimization the difference can be even larger. It sounds small but for an assortment with hundreds of millions in revenue, every percentage point is significant.
Key takeaways
- Greedy TURF is fast but suboptimal
- Exhaustive search guarantees optimum but is computationally expensive
- Swap optimization systematically polishes the greedy solution
- Reverse pruning searches from the opposite direction and often finds new solutions
- The hybrid method combines all three for the best result
Example
When optimizing an assortment with 30 candidates and 8 slots, greedy found a solution with 76.2% reach. Hybrid optimization found an alternative combination with 78.1% reach — a difference corresponding to approximately 15,000 more households reached.
Discuss your assortment with us
Contact us