Product Assistant analytics

This page explains how to interpret Product Assistant performance so you can make better business decisions. It is written for merchants, not as a technical API reference.

Where to find this report

In Merzio Admin, open Product assistant analytics and choose a period:

• 7 days
• 30 days
• 90 days
• 365 days

For meaningful decisions, compare equivalent periods (for example last 30 days vs previous 30 days).

How to read the assistant funnel

Typical shopper flow:

1. Impression - assistant is shown.
2. Start - shopper begins the flow.
3. Result view - shopper reaches final results.
4. Product click / Add to cart - product interaction.
5. Purchase / Assisted purchase - purchase attributed to assistant influence.

If you lose users between impression and start, the issue is often first-step clarity or placement. If drop-off is between start and result view, question design or overly strict rules are usually the cause.

Meaning of overview cards

• Impressions - assistant visibility count.
• Starts - how many times the assistant flow started.
• Result views - how many times shoppers reached final results.
• Add to cart - add-to-cart actions from assistant context.
• Purchases - total purchases attributed to assistant channel.
• Assisted purchases - purchases where assistant contributed to the path.

Key ratios:

• Start rate = starts / impressions x 100
• Result view rate = result views / starts x 100

Start rate can be above 100% in specific cases. This can be expected for event-count based reporting.

Purchase vs assisted purchase

• Purchase: direct purchase attribution from assistant interaction for the product.
• Assisted purchase: assistant influenced the path, but was not the sole direct action.

How to interpret:

• high purchase suggests strong direct conversion influence,
• high assisted purchase suggests the assistant is valuable in discovery and consideration.

You should evaluate both together.

Events by type

Use the event distribution to detect where journey quality breaks:

• many starts, low result views -> weak question progression,
• many result views, low clicks/ATC -> result relevance or assortment mismatch,
• many clicks, low purchases -> friction likely happens after assistant (PDP/cart/checkout).

Top products

Top products help you optimize assistant quality:

• identify products naturally supported by guided selection,
• detect products with clicks but weak purchase follow-through,
• infer which question paths align with actual demand.

If top products repeatedly do not match your commercial priorities, adjust answer matching rules.

By assistant

If you run multiple assistants, compare:

• start quality (start rate),
• progression quality (result view rate),
• final business impact (purchases).

Do not judge performance by impressions alone.

Common interpretation mistakes

• “High impressions means the assistant is successful.”
  Visibility alone is not conversion quality.
• “Low purchases means assistant is bad.”
  Often the issue is outside assistant (offer, pricing, checkout friction).
• “One metric is enough.”
  Reliable interpretation requires the full funnel context.

When purchases stay low

Check:

• whether assistant results include products customers actually buy,
• whether matching rules are too restrictive,
• whether purchase flow stays in the same browser/session context,
• whether conversion friction happens after assistant interaction.

Recommended optimization loop

1. Change one assistant area (question or matching rule set).
2. Observe performance for at least 1-2 weeks.
3. Evaluate impact on start rate, result view rate, and purchases.
4. Apply the next meaningful change only after evaluation.

This prevents random edits and helps you identify what really improved performance.