Performance Metrics

Track Nobi's business impact with key performance indicators, revenue attribution, and engagement metrics. Monitor how performance changes over time and identify opportunities for optimization.

Overview

The Performance Metrics page shows how Nobi contributes to your business across three main areas:

Business impact - Revenue, orders, and conversions driven by Nobi

Engagement - How customers interact with Nobi (searches, clicks, add-to-carts)

Trends over time - How metrics change day-to-day, week-to-week, or month-to-month

Key Metrics

Total Messages

Total number of customer interactions with Nobi during the selected time period. This includes:

Search queries

Conversational messages

Follow-up questions

Product inquiries

Period comparison shows percentage change versus the previous period (e.g., "+14.5% vs previous period").

Searchers & Messagers

An approximate count of the unique site visitors who interacted with Nobi during the time period.

One customer can send multiple messages, so this number is typically lower than Total Messages. Use this to understand reach rather than activity volume.

Revenue

Total revenue from orders attributed to Nobi interactions. Revenue is attributed when customers purchase within 24 hours of using Nobi. Nobi only attributes revenue to the line item(s) that Nobi recommended, not the entire order, and revenue is attributed on a pre-tax basis for each line item.

Transaction Metrics

Orders

Number of completed purchases attributed to Nobi interactions.

Conversion Rate

Percentage of messagers that made a purchase.

Average Order Value (AOV)

Average dollar value of orders attributed to Nobi.

Add to Cart Rate

Percentage of product views that result in add-to-cart actions. Measures how compelling products are after customers find them.

Items/Order

Average number of items per Nobi-attributed order.

Engagement Metrics

Views/Search

Average number of products customers view per search. Higher numbers indicate deeper engagement with search results.

Revenue/Searcher

Total attributable revenue divided by unique searchers.

Search Pageviews

Total number of product pages viewed following a search query.

Metrics Over Time

The time series chart visualizes how your selected metric changes over the time period.

Available metrics:

Message Volume

Searchers & Messagers

Revenue

Orders

Top Queries

See the most common search queries with detailed performance metrics.

Columns:

Query - The search term customers used

Count - Number of times this query was searched

CTR - Click-through rate (percentage of searches that resulted in product clicks)

CVR - Conversion rate (percentage of searches that resulted in purchases)

Orders - Number of purchases attributed to this query

Revenue - Total revenue attributed to this query

Actions - Additional actions you can take (like marking a query as spam)

Query Tabs

Filter queries by type:

All Queries - Every search and message

Short Queries - Brief search terms (typically 1-3 words)

Long Queries - Detailed searches or conversational messages

Spam - Queries identified as spam

Both the Top Queries and Queries by Channel sections can be exported to CSV files using the Export button in each section header.

Filtering and Time Ranges

Time Range Selection

Selects the time range for the data. Metrics automatically compare to the equivalent previous period.

Filter by Entry Point

View metrics for specific launch components:

All Entry Points (default) - Combined data from all components

Specific components:

Search Bar

Button

Toggle

Suggestion Pills

Programmatic API

Queries by Channel

See which traffic channels and sources drive search queries on your site, and which queries lead to revenue. This section appears at the bottom of the Performance page.

💡

This section only shows queries with a known traffic source (attributed queries). Counts may differ from the Top Queries table, which includes all queries regardless of attribution.

Two views are available via tabs:

By Channel — Expandable hierarchy: Channel → Source → Query. Columns: Channel, Attributed Count, CTR, CVR, Orders, Revenue. Start at the high level and drill down.

By Query — Server-side paginated table of all attributed queries. Columns: Query, Queries (count), CTR, CVR, Orders, Revenue. Expand any query to see which channels and sources sent it.

Channel Hierarchy

The By Channel view groups queries into three expandable levels:

Channels — Organic Search, Paid Search, Social, Email, Referral, Direct

Sources — Individual traffic sources within each channel (e.g., Google, Bing under Organic Search)

Queries — The most-searched queries plus every query that generated revenue, even if searched only once

Queries that drove revenue are shown in bolder text in the Orders and Revenue columns. If there are more than 10 queries, the rest are behind a "show more" link that tells you how many hidden queries have revenue.

Sorting

Click any column header to sort. In the By Channel tab, sorting cascades to all levels — channels, sources, and queries all reorder together. In the By Query tab, sorting is handled server-side and resets to the first page. Click the same column again to reverse direction.

Attributed Count vs. Count

In the By Channel tab, the "Attributed Count" column shows searches by visitors with a known traffic source. In the By Query tab, the equivalent column is labeled "Queries." Both may be lower than the "Count" in Top Queries, which includes all searches regardless of attribution. For example, if "linen" was searched 33 times total but only 15 visitors came from a known source, Top Queries shows 33 and Queries by Channel shows 15.

Traffic Channels

Traffic is classified based on UTM parameters and referrer data:

Organic Search — From search engines without paid ads

Paid Search — From paid search ads (utm_medium = cpc, ppc, or paid)

Social — From social media platforms

Email — From email campaigns (utm_medium = email)

Referral — From other websites linking to your store

Direct — Typed URL or bookmark

Paid Social — From paid social media ads (utm_medium = cpc, ppc, or paid on a social platform source)

Organic AI — From AI platforms like ChatGPT, Claude, Perplexity, Gemini, or Copilot

Paid AI — From paid placements on AI platforms (utm_medium = cpc, ppc, or paid on an AI platform source)

Display — From display or banner advertising (utm_medium = display, banner, or cpm)

Direct / Unknown — No referrer or UTM data available. The visitor typed the URL directly, used a bookmark, or their browser did not send referrer information. These queries are excluded from the Queries by Channel section.

Channel classification follows the same conventions as Google Analytics 4 (GA4). If you are comparing Nobi data to GA4, the channel definitions should align.

How Traffic Attribution Works

Nobi automatically tracks where your site visitors come from — which search engine, social platform, ad campaign, or referral link brought them to your store. This is called traffic attribution. Below is a detailed explanation of how it works.

Sessions

A session represents a single visit to your website. Nobi groups a visitor's activity into sessions to correctly attribute their queries to the traffic source that brought them to your site.

30-minute timeout — A session expires after 30 minutes of inactivity. If a visitor leaves your site and returns 45 minutes later, that counts as a new session with fresh attribution.

Per device and browser — Sessions are tracked in the visitor's browser using localStorage. The same person visiting from their phone and their laptop counts as two separate sessions.

Continuous browsing extends the session — As long as the visitor keeps interacting with your site (loading pages, searching, clicking), the 30-minute timer resets. A visitor who browses for 2 hours straight is still one session.

Session expiry clears attribution — When a session expires, all stored traffic source data (UTM parameters and referrer) is cleared. The next visit starts fresh and gets attributed to whatever source brought the visitor back.

UTM Parameters

UTM parameters are tags added to URLs to track where traffic comes from. When someone clicks a link with UTM tags (e.g., from a Google ad or an email campaign), those tags tell Nobi exactly which channel, source, and campaign brought the visitor.

Example URL with UTM parameters:


https://yourstore.com/products?utm_source=google&utm_medium=cpc&utm_campaign=spring_sale

Nobi tracks five standard UTM parameters:

utm_source — The platform or site that sent the traffic (e.g., google, facebook, newsletter)

utm_medium — The marketing medium (e.g., cpc, email, social, banner). This is the most important parameter for channel classification — it determines whether traffic is "paid" vs "organic."

utm_campaign — The specific campaign name (e.g., spring_sale, black_friday_2025)

utm_term — The paid search keyword (e.g., running+shoes)

utm_content — Used to differentiate ad variants or links within the same campaign

ℹ️

UTM parameters only exist on the landing page URL. Once a visitor navigates to a second page on your site, the UTM tags disappear from the browser's address bar. Nobi solves this by capturing UTM parameters on the first page load and storing them in the browser's localStorage for the duration of the session. Every subsequent query the visitor makes is still correctly attributed to the original traffic source.

Referrer Tracking

When a visitor arrives at your site by clicking a link on another website, their browser sends the URL of that other website as the referrer. Nobi uses this to identify traffic sources even when there are no UTM parameters.

How it works:

External vs. internal referrers — Only referrers from a different domain are tracked. When a visitor navigates from one page on your site to another, that internal navigation is ignored.

Original referrer is preserved — Nobi captures the external referrer from the landing page and stores it for the entire session. If a visitor arrives from Google, browses 10 pages, and then searches using Nobi, that query is still attributed to Google.

Domain extraction — Nobi extracts the domain name from the referrer URL. A referrer of "https://www.google.com/search?q=shoes" becomes the source "google". Regional variants like google.co.uk or google.com.au are normalized to "google".

Referrer as fallback — If UTM parameters are present, they take priority for determining the source. The referrer is used as a fallback when no UTMs exist (e.g., a visitor clicks an organic Google search result, which has no UTM tags but does send a referrer).

Attribution Model

Nobi uses a hybrid attribution model within each session:

Referrer: First-touch — The original external referrer captured on the landing page is never overwritten during a session. This preserves the true "entry point" to your site.

UTM Parameters: Last-touch — If new UTM parameters appear during a session (e.g., a visitor clicks a second ad campaign link while already on your site), the UTM values update to the newest ones. This ensures the most recent campaign gets credit.

This hybrid approach balances two goals: knowing how the visitor originally found your site (referrer) and knowing which specific campaign most recently influenced them (UTMs).

ℹ️

Attribution resets between sessions. If a visitor comes from Google on Monday and returns via a Facebook ad on Tuesday, the Tuesday visit is attributed to Facebook — each session gets its own attribution.

Channel Classification Rules

Nobi classifies traffic into channels using two inputs: the source (which platform) and the medium (paid vs. organic). The source comes from utm_source or the referrer domain. The medium comes from utm_medium.

Classification is evaluated in priority order. The first matching rule wins:

Paid AI — Source is an AI platform AND medium indicates paid (cpc, ppc, retargeting, or any value starting with "paid")

Organic AI — Source is an AI platform (no paid medium). AI platforms: chatgpt, openai, claude, anthropic, gemini, perplexity, copilot

Paid Search — Source is a search engine AND medium indicates paid. Search engines: google, bing, yahoo, duckduckgo, baidu, yandex

Organic Search — Source is a search engine (no paid medium)

Paid Social — Source is a social platform AND medium indicates paid. Social platforms: facebook, instagram, twitter/x, tiktok, pinterest, linkedin, snapchat, reddit, youtube, nextdoor

Organic Social — Source is a social platform (no paid medium)

Email — Medium is "email" (regardless of source)

Display — Medium is "display", "banner", or "cpm"

Referral — An external referrer exists but does not match any known search engine, social platform, or AI platform

Direct / Unknown — No UTM parameters and no external referrer. Not shown in the Queries by Channel section.

Source Normalization

People use different names and abbreviations for the same platform. Nobi normalizes common aliases so they appear under a single source name:

fb → facebook

ig → instagram

x or t → twitter

lnkd → linkedin

openai → chatgpt

anthropic → claude

Regional domain variants (e.g., google.co.uk, google.com.au) are also normalized to the base platform name.

Frequently Asked Questions

How does Nobi figure out the channel and source?

Two signals are used: (1) UTM parameters in the URL, and (2) the referrer — the website the visitor came from. UTM parameters take priority when present. If there are no UTMs, Nobi falls back to the referrer domain. For example, if a visitor clicks an organic Google search result, there are no UTM tags, but the browser sends google.com as the referrer, so the query is classified as Organic Search from Google.

What happens when a visitor navigates between pages?

UTM parameters disappear from the URL after the first page, and the referrer changes to your own domain (an internal referrer). Nobi handles this by capturing the original external referrer and UTM parameters on the landing page and storing them in the browser's localStorage. Every subsequent event during the same session uses this stored data, so attribution is preserved no matter how many pages the visitor browses.

Why do the numbers in Queries by Channel differ from Top Queries?

Top Queries shows all queries regardless of attribution. Queries by Channel only shows queries where a traffic source was identified. Visitors who type your URL directly or arrive without any referrer or UTM data are not attributed to any channel and are excluded from the channel breakdown. This means the total count in Queries by Channel is typically lower than in Top Queries.

How long does attribution last?

Attribution lasts for the duration of a session (until 30 minutes of inactivity). If a visitor leaves and comes back after the session expires, they start fresh. The new visit is attributed to whatever source brought them back — or not attributed at all if they typed the URL directly.

What if a visitor comes from Google, leaves, and returns from Facebook?

Each session gets its own attribution. The first visit's queries are attributed to Google (Organic Search). If the second visit is a new session (more than 30 minutes later), those queries are attributed to Facebook (Organic Social or Paid Social, depending on UTM parameters). If the second visit is within the same session, the original referrer stays as Google, but if the Facebook link had UTM parameters, those would update (since UTMs use last-touch attribution within a session).

Does Nobi track attribution across devices?

No. Attribution is stored in the browser's localStorage, which is per-device and per-browser. A visitor who clicks a Google ad on their phone and later visits your site from their laptop will show as two separate unrelated sessions.

Do I need to set up UTM parameters for this to work?

No setup is required for organic traffic. Nobi automatically detects referrers from search engines, social platforms, and AI tools. However, for paid campaigns (Google Ads, Facebook Ads, email campaigns, etc.), you should use UTM parameters on your ad and link URLs to get the most accurate attribution. Most ad platforms add UTM parameters automatically, but it's good practice to verify they are in place.

How does this compare to Google Analytics?

Nobi's channel definitions align with Google Analytics 4 (GA4) conventions. The same UTM parameters and channel groupings are used. One difference: Nobi only tracks queries made through the Nobi assistant, not all site traffic. So Nobi's numbers represent "searches and conversations through Nobi, broken down by how visitors arrived at the site" — not total site traffic by channel.

What counts as "Direct / Unknown" traffic?

Direct traffic means the visitor arrived without any tracking signals — no UTM parameters and no external referrer. Common scenarios include: typing the URL directly, clicking a bookmark, clicking a link in a native mobile app (which often strips the referrer), clicking a link in a PDF or document, or when the visitor's browser privacy settings block referrer information.

When does Nobi process channel classification?

Channel classification happens during a nightly data aggregation process, not in real time. Raw events are stored with their UTM parameters and referrer data, and the classification into channels (Organic Search, Paid Social, etc.) is applied when the daily metrics are computed. This means newly classified data appears the following day.

Performance Metrics Key Metrics Total Messages Searchers & Messagers Revenue Transaction Metrics Orders Conversion Rate Average Order Value (AOV)Add to Cart Rate Items/Order Engagement Metrics Views/Search Revenue/Searcher Search Pageviews Metrics Over Time Top Queries Query Tabs Filtering and Time Ranges Time Range Selection Filter by Entry Point Queries by Channel Channel Hierarchy Sorting Attributed Count vs. Count Traffic Channels How Traffic Attribution Works Sessions UTM Parameters Referrer Tracking Attribution Model Channel Classification Rules Source Normalization Frequently Asked Questions How does Nobi figure out the channel and source?What happens when a visitor navigates between pages?Why do the numbers in Queries by Channel differ from Top Queries?How long does attribution last?What if a visitor comes from Google, leaves, and returns from Facebook?Does Nobi track attribution across devices?Do I need to set up UTM parameters for this to work?How does this compare to Google Analytics?What counts as "Direct / Unknown" traffic?When does Nobi process channel classification?