How a London Retailer Made AI Visibility Measurable at Scale

A London retailer faced a growing reporting problem. Its teams could track website traffic, conversions and trading performance in GA4, but they had no clear way to see how often the brand appeared in AI-generated answers — or whether competitors were being recommended instead.

To solve that, the retailer ran a pilot combining Quadrant with server-side GA4 collection, BigQuery and Looker. The goal was not to promise guaranteed rankings or citations. It was to build a repeatable way to monitor shopper-style prompts, analyse citations and share findings with marketing, e-commerce and analytics teams.

The commercial value was simple: AI visibility became measurable within the same reporting rhythm as other digital performance data.

Some customer-specific figures remain confidential, so those fields are marked clearly below. The workflow, reporting structure and measurement approach can still be applied more broadly.

The Reporting Gap the Team Couldn’t Ignore

The retailer had started to see more shoppers using AI assistants during product research, comparisons and category discovery. But the business had no joined-up view of what those assistants were actually saying about the brand.

Traditional reporting could answer questions like:

• How many users arrived from organic search?
• Which landing pages generated revenue?
• Which product categories converted?
• How did paid and owned channels perform?

It could not answer the questions that were becoming more important:

• Which shopper prompts mentioned the retailer?
• Which products were cited in the answer?
• Was the brand prominent, buried or absent?
• Which sources appeared to influence the answer?
• How often did competitors appear in the same response?
• Did content updates change the retailer’s presence over time?

This was not just a technical gap. It was an operating issue across teams.

• Brand managers needed a clearer view of positioning.
• Content teams needed evidence about which questions to answer.
• Analytics managers needed consistent definitions and exportable data.
• Senior stakeholders needed reporting that could sit alongside trading, acquisition and conversion metrics.

So the retailer treated AI visibility as a measurement workflow, not just another marketing score.

From GA4 to BigQuery to Quadrant

The pilot connected four layers, each with a specific role.

1. Server-side GA4 collection

The retailer used server-side event collection to create a controlled route for selected events and business signals. Google describes the GA4 Measurement Protocol as a way to send server-to-server and offline interactions to Analytics, while noting that it supplements rather than replaces standard tagging.

For this pilot, the important principle was clarity of responsibility. GA4 remained the source of truth for on-site behaviour and commercial events. Quadrant remained the source for monitored AI answers, prompt-level visibility and citation observations.

2. BigQuery as the analytical layer

The retailer’s GA4 property exported event data into BigQuery. Google’s GA4 export creates a property-specific dataset and daily event tables, with streaming exports available when enabled.

The analytics team then used BigQuery to combine approved GA4 dimensions with structured AI visibility data supplied by Quadrant. Typical fields included:

• Reporting date
• Market and language
• Prompt identifier
• Prompt category
• AI platform or model
• Brand or product detected
• Citation status
• Answer position or prominence classification
• Competitor presence
• Landing page or cited source
• Content owner
• Recommended action status

The aim was not to collapse everything into a single blended score. It was to give teams a reliable place to compare AI visibility observations with digital performance data while keeping each measure meaningful.

3. Quadrant monitoring and analysis

Quadrant was used to monitor the retailer across a defined set of shopper-style prompts. Its public positioning highlights prompt-level insight, citation monitoring, competitor benchmarking and export-friendly reporting for retail, FMCG and e-commerce teams.

The prompt set focused on real commercial questions rather than generic brand searches. For example:

• Which retailers are best for a specific product category in London?
• What should shoppers look for when comparing products in this category?
• Which brands offer a particular combination of price, quality and availability?
• Where can customers find products suitable for a stated need?
• Which retailers are known for a particular range or service?

This mattered because it separated brand familiarity from true discoverability. A retailer might show up often in branded prompts but still be largely absent from non-branded category questions.

4. Looker for shared reporting

Looker connected to the BigQuery dataset and gave marketing, analytics and leadership teams a governed reporting layer. Google documents Looker’s connection to BigQuery for SQL-based analysis, dashboarding and exploration.

The workflow was straightforward:

1. GA4 collected approved digital events.
2. BigQuery stored and transformed the data.
3. Quadrant supplied structured AI visibility observations and exports.
4. BigQuery combined the datasets using shared reporting dimensions.
5. Looker presented role-specific dashboards and scheduled reports.

What the Team Saw in the Dashboards

The reporting was designed around decisions, not platform features. Four views proved especially useful.

Prompt-cluster performance

This view grouped prompts by shopper intent, such as category comparison, product suitability, price sensitivity, local availability and service expectations.

It showed where the retailer appeared consistently and where visibility was limited to branded queries. That helped content teams prioritise category and product pages that could close genuine discovery gaps.

Brand citation share

Citation rate was reported separately from mention rate.

A brand mention meant the retailer appeared in an answer.
A citation meant the answer also referenced a source associated with the retailer or product.

Keeping those measures separate prevented a common reporting mistake: assuming that any mention meant the retailer’s own content had influenced the answer.

Competitor benchmarking

The dashboard showed the retailer’s presence alongside observed competitor presence for the same prompt set, market and timeframe.

That made discussions more useful in trading meetings. Teams could quickly see whether a shift reflected improvement for the retailer, a broader category change or both.

Category movement over time

A time-series view tracked visibility by category and prompt cluster. This supported a practical review cycle:

• Identify a visibility gap
• Review the sources and content associated with the answers
• Make an approved content or merchandising change
• Re-run the same prompt set
• Compare results against the baseline

Quadrant’s own positioning treats visibility, sentiment, citations and competitor gaps as separate signals. That distinction is valuable when teams need to turn AI observations into practical content and commercial actions.

Before-and-After Prompt Results

The pilot used a fixed prompt set, with a defined baseline period followed by an intervention period. Exact dates and values remain confidential until approved for publication.

Metric	Baseline period	Pilot period	How the team interpreted it
Monitored prompts	Confidential	Confidential	Confirmed that the same prompt set was used for like-for-like comparison
Prompt execution volume	Confidential	Confidential	Distinguished monitoring coverage from visibility performance
Brand mention rate	Confidential	Confidential	Measured how often the retailer appeared in relevant answers
Citation rate	Confidential	Confidential	Measured how often an answer referenced an approved retailer or product source
Prominent answer placement	Confidential	Confidential	Recorded whether the retailer appeared in a high-visibility position within the answer
Competitor share of observed mentions	Confidential	Confidential	Showed whether competitors appeared more frequently for the same prompt set
Reporting preparation time	Confidential	Confidential	Assessed the operational impact of replacing manual compilation with scheduled reporting

Methodology

The measurement approach relied on:

• A fixed prompt set
• Consistent market and language settings
• Documented monitored platforms
• Comparable reporting windows

Citation rate was calculated as the number of monitored answers containing an approved retailer or product citation divided by the number of valid answers returned for the relevant prompt set.

Prominent placement was defined in advance by the reporting team and applied consistently across both baseline and pilot periods.

These results should be interpreted as directional performance for the monitored prompts and platforms. They are not a direct estimate of total consumer exposure, traffic or revenue, and they should not be used to claim guaranteed rankings, citations or placement in AI answers.

Exports That Made Reporting Easier

One of the biggest practical benefits was that teams no longer had to rebuild the analysis manually every week.

A CSV export supported analyst-friendly tasks such as:

• Reviewing prompt-level changes in a spreadsheet
• Filtering results by category owner
• Adding commentary to a weekly trading pack
• Checking individual answers before making a content decision
• Sharing a concise view with non-technical stakeholders

A JSON export supported more structured workflows, including automated analysis:

{
  "prompt_id": "category_london_001",
  "market": "GB",
  "language": "en-GB",
  "captured_at": "YYYY-MM-DDTHH:MM:SSZ",
  "brand_mentioned": true,
  "citation_present": true,
  "answer_position": "prominent",
  "competitor_mentions": ["approved_competitor_reference"],
  "source_url": "approved_source_reference"
}

This example shows the sort of structure needed for downstream analysis. The exact export format should always be confirmed against the live implementation.

That allowed the team to use the same underlying observations in three ways:

1. Weekly trading updates
   A short Looker view showing movement by category and prompt cluster.

2. Marketing reviews
   Evidence to guide content, product information and source-quality decisions.

3. Deeper analytics
   BigQuery analysis joining AI visibility observations with approved GA4 dimensions.

This reduced the gap between insight and action. A visibility issue could be assigned to a content owner, reviewed against the relevant page or product information, and re-measured using the same prompt definition.

What Other UK Retail Teams Can Learn

Treat the prompt set as a measurement asset

A prompt list should reflect real shopper intent across discovery, comparison and purchase consideration. It should be version-controlled, reviewed and protected from unnecessary changes during a reporting period.

Keep mention, citation and placement separate

These metrics answer different questions:

• Mention rate measures presence
• Citation rate measures source association
• Placement measures prominence

Combining them too early makes reporting harder to interpret.

Give marketing and analytics different views of the same data

Marketing teams need prompt examples, category gaps and recommended actions. Analytics teams need definitions, timestamps, dimensions and reproducible exports.

A shared data layer can support both without forcing everyone into the same dashboard.

Use existing governance where possible

The retailer’s advantage came from fitting AI visibility reporting into existing GA4, BigQuery and Looker controls. That made ownership, permissions, refresh schedules and reporting standards easier to manage.

Measure content changes with discipline

The strongest workflow is not “publish and hope”. It is:

• establish a baseline
• make a change
• re-run the prompt set
• compare results

Every result should be reported with the prompt set, timeframe, platform coverage and success definition attached.

The Bigger Lesson

For UK retail, FMCG and e-commerce teams, the lesson is straightforward: an AI visibility tracker becomes commercially useful when it moves beyond an isolated score.

Connected to an established GA4, BigQuery and Looker reporting setup, it can become a repeatable process that helps teams decide which questions to answer, which content to improve and where competitors are appearing instead.

That is the difference between simply collecting AI search data and making AI visibility part of everyday digital decision-making.