Website Anomaly Diagnosis with Google Analytics / Ptengine
Data Analyst, Marketing Manager, E-commerce Director, Head of Growth, CRO Specialist
Before
You get a notification or notice in a report: "Overall revenue dropped 30% yesterday." The panic sets in. To find the cause, you begin a painful, manual "data-diving" expedition in Google Analytics. You open dozens of tabs, trying to cross-reference dimensions. "Is it mobile traffic? No, that looks stable. Is it a specific country? Let me filter for the US. Okay, US traffic is down. Which channel? Let me add a secondary dimension. Organic search is down. Okay, which landing pages for organic search in the US are down?" This process is a frustrating rabbit hole that is slow, inefficient, and you can easily miss the real root cause.
After
Kocoro acts as your automated data detective. You start by stating the problem in plain language: "Our site revenue fell sharply yesterday. Investigate the cause." Instead of you hunting for the answer, the AI agent begins a systematic investigation. It automatically segments the data by the most likely dimensions (channel, device, location, landing page), identifies the most significant deviation, and presents a clear hypothesis. You can then ask follow-up questions in a natural conversation to drill down further, turning a multi-hour investigation into a focused, 5-minute dialogue.
Use Case Scenarios
1. Diagnosing a Sudden Drop in Sales
Scenario: It's Tuesday morning, and you see that Monday's sales were 40% below the daily average. You need to know if this was due to a technical issue, a marketing channel failure, or something else entirely before it impacts the whole week.
How Kocoro Helps: You prompt your agent: "Revenue dropped significantly yesterday. What was the primary cause?" The agent might respond: "The drop in revenue correlates with a 70% decrease in traffic from your 'email' channel." You can then drill down: "Okay, for the email channel, which campaign and landing page saw the biggest drop?" The agent can pinpoint the problem: "The 'Weekly Newsletter' campaign traffic to the '/new-arrivals' page was nearly zero yesterday. This suggests a broken link in the newsletter." You've found the root cause in two questions.
Investigating an Unexpected Traffic Spike
Scenario: Your website traffic suddenly doubles on a random Wednesday afternoon. This could be a viral blog post (great!) or a bot attack (bad!). You need to quickly identify the source to know whether to celebrate or escalate to IT.
How Kocoro Helps: You can ask Kocoro: "My website traffic spiked by 100% between 2 PM and 4 PM today. Where did this traffic come from?" Kocoro analyzes the data and reports: "The spike is 95% 'Direct' traffic from a single city ('Mountain View, California') and has a 99% bounce rate and 1-second session duration." This pattern is a clear indicator of bot traffic, allowing you to immediately apply a filter or report the issue, rather than misinterpreting it as genuine user interest.
Finding a Break in a Critical Conversion Funnel
Scenario: Your "Free Trial Signup" is your most important conversion goal. You notice that while plenty of people are visiting the pricing page, almost no one is reaching the "thank you" page today.
How Kocoro Helps: You can give Kocoro a specific diagnostic task: "Analyze the 'Free Trial' funnel for today's data. Segment the drop-off rate between the '/signup' page and the '/thank-you' page by browser." Kocoro quickly processes the data and reveals the issue: "The funnel completion rate for users on Chrome and Firefox is normal, but it is 0% for users on Safari. The drop-off rate for Safari users is 100% on the '/signup' page." This tells you there is likely a technical bug specifically affecting the Safari browser, allowing you to create a highly specific and actionable ticket for your development team.
How to Create
Connect Your Data Source: Ensure your Google Analytics (GA4) account is securely connected.
Know Your KPIs: Have a basic understanding of your website's normal performance metrics (e.g., average daily users, normal conversion rate). This is how you'll spot the "anomaly" to begin with.
Use a Dedicated Agent: Create an agent named
Data Detectiveand grant it access to the Google Analytics tool and Web Search.Start High-Level, Then Drill Down: Start your conversation by stating the high-level anomaly you observed. Then, use follow-up questions to dig deeper into the data, guided by the agent's initial findings.
Best Practice
Recommended User Prompt
For a high-level starting point:
For drilling down into a specific channel:
For drilling down into a specific page:
For combining internal and external data:
For a proactive check:
Recommended Agent System Prompt:
Instruction：
Model:
The Claude series is highly recommended for its strong logical reasoning and ability to follow a chain-of-thought, which is perfect for a drill-down investigation.
Tools:
Artifacts: For crafting a full dashboard.
Chart Generator: To create charts that visually confirm a sudden drop or spike.
Knowledge Base:
Upload documents like a Marketing Campaign Calendar, Promotion Schedule, or Website Maintenance Log. This gives your agent crucial context to correlate data changes with business activities (e.g., "The traffic drop coincides with the end of the 'Summer Sale' promotion.").
Data Source:
Google Analytics
Ptengine
Meet a new era of work,
with AI made easy for everyone.
Effortlessly uncover deep customer insights and free your team from routine tasks, so they can focus on what they do best: creative, high-impact work that drives results.
Kocoro
English
Kocoro. All rights reserved. © 2025