Boosting and merchandising for ecommerce search

Good ecommerce search is not only about relevance. The top results should reflect both what the shopper meant and what the store wants to move. Indx gives you two levers for this: field weights and boosts. Neither requires turning every query into a tuning project.

Relevance gets shoppers to the right shelf. Merchandising decides what sits at eye level.

Weight the fields that matter

Every searchable field has a weight, a float that only matters relative to the other searchable fields. If a match in the product name should count for more than a match buried in the description, give the name a higher weight. Set it once at configuration time and every query benefits.

engine.GetField("name")!.Weight        = 2.0f;
engine.GetField("description")!.Weight = 1.0f;

Boosts for merchandising

Boosts push documents matching certain criteria up the results without excluding anything else. A boost is built from the same filters you already use, such as a value filter on a category or a range filter on a year, combined freely and applied at one of three strengths.

var inSeason = engine.CreateValueFilter("season", "winter", out _);
var boost    = engine.CreateBoost(inSeason, BoostStrength.Med);

query.Boosts = new[] { boost };
query.EnableBoost = true;

Because the underlying results stay relevance-ranked, promoted products never feel random. They rise, they do not take over.

Personalized boosting

The same mechanism scales to per-user merchandising. Combine value filters for a user's frequently bought items into one filter with OR, and boost it. This handles hundreds of thousands of items without significant performance impact, so "your usual products first" is a few lines of code rather than a separate system.

Where to start

Begin with field weights, confirm the baseline relevance feels right, then layer boosts on top. Small adjustments compound across thousands of queries a day.

• Boosts in the C# API
• Personalized boosting how-to