¶ Description
Mono-product assignment solver using exact techniques.
¶ Parameters
¶ Parameters tab
Parameters:
- Each customers receive an ad about x products. x=
- Number of segments
- Total number of products
- Maximum number of offers
- Segment number
- Primary key customer
- Name product
- Margin product
- Min quantity product
- Max quantity product
- Informations de débogage en console
- Number of additional threads
¶ Description tab
Parameters:
- Script name
- Short description
- Revision
- Description
¶ Configuration tab
See dedicated page for more information.
¶ About
The ExactAssignmentSolver action solves the constrained product assignment problem using exact linear programming (LP) methods. It assigns a predefined number of products (nAds) to each customer in a population, while satisfying product quantity constraints and maximizing campaign ROI (based on per-product margins).
This operator is intended for smaller-scale problems with a manageable number of segments and products. For larger datasets or populations (e.g. >100K customers), consider using the meta-heuristic version of the assignment solver to avoid combinatorial explosion.
¶ Input Pins
¶ Pin 0 & Pin 1 — Segment Tables
Tables that describe customer segments, typically built using probability of purchase features via Stardust. All customers within a segment will receive the same product assignment.
¶ Pin 2 — Customer Table
Each row represents a customer, and includes:
- Probabilities of purchase for each product
- Segment number the customer belongs to
- Primary key (unique ID) of the customer
¶ Pin 3 — Product Constraints Table
Each row represents a product with the following structure:
- Product Name
- Margin (for ROI calculation)
- Minimum Quantity constraint
- Maximum Quantity constraint
¶ How It Works
The solver computes the optimal product-to-customer assignment by:
- Ensuring every customer receives exactly
nAdsproducts. - Respecting product-level constraints from Pin 3.
- Maximizing total campaign ROI, based on margins.
- Assigning identical product combinations to all customers within the same segment.
¶ Developer Configuration
-
Float/Key conversions:
- If negative:
abort - If > 2³²−3:
abort - If not integer:
set to lower integer
- If negative:
-
String/Float conversions:
- If not a number:
abort
- If not a number:
-
String/? to key: Similar rules as above
¶ Use Case Example
You want to run a marketing campaign that recommends 3 products to each customer. You’ve defined customer segments based on purchasing behavior, and you want to ensure:
- Product margins are considered for ROI
- Quantities don’t exceed warehouse limits
- Each segment gets a uniform product combo
Configure nAds=3, nSegments=500, and link your segment/customer/product tables to the appropriate pins. The action will return an optimized allocation table for deployment.
¶ Notes
- Avoid setting
nSegmentsornProductstoo high (e.g., >1000), as LP-based solving can quickly become intractable. - All customers within a segment must share the same product assignment.
- Best used after segmentation with
Stardust.
¶ Best Practices
- Use this operator for small to medium populations.
- Prefer meta-heuristic assignment solver for very large populations.
- Validate margins and constraints before execution to prevent infeasible LP solutions.
