PROPOSED PROJECT
Product Bundle Recommendations
Suggest products that work well together, using your catalog, available stock, and rules set by your retail team. Explain why each bundle makes sense.
A set that fits. A total you can check.
Desk + light + storage
Item budget: $275
One useful desk setup.
$25 below the $275 item budget.
No bundle discount applied.
The project in practice.
A homeware retailer wants a desk-setup bundle assistant. The catalog includes desks, lamps, organizers, and add-ons that do not fit every model. The brief is to recommend a useful combination under the shopper’s stated budget while respecting stock and compatibility.
WHAT IS USUALLY DONE
Related-product rows can pair items that duplicate a function, do not fit together, or are unavailable. A free-form AI assistant may invent a bundle price or ignore a compatibility rule.
WHAT SOVRUN CAN DO FOR YOU
Sovrun would scope a bundle assistant to one shopping occasion. Explicit rules filter compatibility and stock; a ranking model proposes eligible combinations, and AI explains why the items work together using catalog attributes. Prices are calculated from the commerce data.
The shopper asks
for a useful setup.
The request is a desk, task light, and organizer within a 1,250 × 650 mm footprint and a $275 item budget. The desk must fit the space, the lamp clamp must fit the top, and the organizer must attach to the right rail.
The desk fits the limit.
- Desk footprint
- 1,200 × 600 mm
- Maximum footprint
- 1,250 × 650 mm
- Difference
- 50 mm on each dimension
The clamp fits the top.
- Top thickness
- 30 mm
- Clamp range
- 10–40 mm
- Fit source
- Approved catalog rule
The mount matches.
- Desk rail
- R1
- Organizer mount
- R1
- Fit source
- Merchant compatibility list
The desk meets the supplied footprint. The lamp adds adjustable task lighting without using tabletop space. The organizer attaches under the front rail to keep small items off the surface. AI can turn those confirmed attributes into a useful explanation; it cannot invent a fitting or discount.
A cheaper option
can still be the wrong one.
O2 costs $20 but uses rail R2, so it cannot replace the R1 organizer. L50 costs $40 but has zero available stock. Both stay out of this recommendation even though their prices would fit the budget.
Wrong rail.
The desk’s R1 rail does not match this organizer’s R2 mount.
Not available.
A compatible clamp does not make an out-of-stock lamp eligible.
The displayed stock is a snapshot, not a reservation. Recheck stock, prices, and the applicable rules before placing an order. If a required attribute disappears, ask for the missing record rather than guessing.
A recommendation
with reasons behind it.
The proposed first project combines catalog rules, available stock, and an explanation of the product roles. Merchants can inspect the choices and test changes before launch.
One useful desk setup.
$25 below the $275 item budget.
No bundle discount applied.
5 complete setsOne of each item per set. The organizer limits availability.
Every choice has a reason.
- Selected
- D120 + L40 + O1
- Checks
- Space, fit, stock, budget
- O2 excluded
- Wrong rail
- L50 excluded
- No stock
- Conversion result
- Not measured
How we’d make it happen.
Constrain
Define combinations that are allowed, unavailable, or incompatible.
Recommend
Rank eligible sets against the stated occasion and explain the product roles.
Validate
Test budget limits, stock changes, missing attributes, and contradictory requests.
Where AI helps
AI interprets the shopping brief and drafts the explanation. Eligibility, availability, discounts, and arithmetic are handled by explicit business logic.
What people approve
Merchants approve compatible product groups and review sample bundles. The pilot does not change prices or create promotions autonomously.
What we’d need to start
- Catalog with compatibility attributes
- Stock and pricing source
- Approved bundle and promotion rules
How we’d test it
Check compatibility, accurate totals, availability, and explanation quality before measuring conversion in a separate live experiment.
Prepared for Retail & e-commerce. We would agree on the work after reviewing your brief, materials, and requirements.
This is a proposed project, not completed client work. The visuals and numbers show example outputs, not measured results.
