How we compute product alternatives

Overview

Every product on Liners can have an alternatives section showing similar products a user may want to consider. Alternatives are computed based on multiple signals such as the product categories, features, technologies, countries, platform signals, and product positioning.

The goal is not to declare one product "better" than another. It is to help users find nearby options that solve a similar problem.

How Matching Works

When we compute alternatives for a product, we compare it with other published products and look for meaningful overlap. The strongest signals are usually shared product features, category fit, and whether products serve similar markets or customers.

Signals can include:

• Product categories.
• Product feature tags.
• Technology and platform signals.
• Country and regional availability.
• Business model and audience fit.

Products that share a broad category but do not solve a similar problem will not appear as strong alternatives. For example, a payment gateway and a savings app may both sit under fintech, but they are not automatically substitutes.

Market Context

Country and region signals help keep alternatives useful. Products available in the same market are generally more comparable than products that serve completely different geographies.

Pan-African products are treated as broader options because they may serve users across multiple markets.

Quality Review

DD Dave helps compute alternatives, and QA Quinn helps monitor whether the results remain sensible as the taxonomy changes.

If a product's category, tags, or availability data changes, alternatives may be recomputed so the recommendation stays aligned with the current listing.

What Does Not Affect Alternatives

Verified status does not mean a product is a better alternative. Paid advertising does not buy placement in organic alternatives. Alternatives are based on similarity and usefulness, not sponsorship.

When Alternatives Are Computed

We compute alternatives at several points:

• When a product is published/updated, we automatically calculate alternatives.
• On request, alternatives can be regenerated when product data changes.
• When metadata changes, significant category, tag, country, or product-positioning changes can trigger recomputation for related products.

How Alternatives Are Stored and Displayed

Alternatives are pre-computed and stored so product cards, product pages, and alternatives pages can use consistent results. This also keeps public pages fast because they do not need to recompute similarity on every request.

Summary

Alternatives are designed to show nearby products a user may reasonably compare. The system uses product metadata and market context, while keeping sensitive scoring details private to protect result quality.