NEW: OVME Case Study

Using Claude 3.7 with Extended Thinking for Google Ads Performance Timeline Analysis

 In AI, Google Ads

Recently Anthropic released an updated Claude model, Claude 3.7 Sonnet with Extended Reasoning. Claude has long been one of my personal favorite AI tools given its ability to produce diagrams and visualizations shareable with artifacts, amongst other good features. 

Although we’re sure the latest Claude model can be used for many innovations in paid media, here’s one workflow we’ve developed we think is novel, helpful, and easy to implement. This is the AI Powered Google Ads Performance Timeline Analysis. 

Step 1: Download a Performance Report with a Time Dimension. The simplest example would be standard campaign report, with a monthly segment as a dimension, and common performance metrics as rows, such as clicks, impressions, conversions, conversion rate, cost per conversion, and so on. Of course you can customize this to your needs. 

Step 2: Download the Change History from Google Ads. Obviously, you’ll want to match the timeline from the change history to the timeline of the report. 

Step 3: Upload both reports in Claude, and offer it a prompt, something like the below. 

“Use the two documents attached to make a timeline of Google Ads performance. The key performance metrics are conversions and cost per conversion. Create an interactive chart which combines both the performance data and the important change history data.

The two documents attached are Google Ads reports. One is a report of campaign performance, segmented by month. The report includes all data for the year 2024. The second document is the Google Ads change history for the full year of 2024.

The purpose of this project to align the Google Ads performance data with the Google Ads change history data so that we can derive insights on what changes in the account were associated with what results. Use advanced reasoning to determine how the changes might have impacted the results, but also the inverse, how the results might have lead to certain changes.

For this exercise, act as a Google Ads optimization specialist and reference all available material on Google Ads best practices. However, only make conclusions based on the data presented.”

Step 4: Review the Output and Challenge Claude with Different Iterations

Claude’s ability to output visuals, charts, and what it calls “Artifacts” is one of it’s best features, and Claude will offer many different ways to visualize the data from the prompt above. You can play around with different prompts to find new visualization styles: 

Performance Charts with Claude

Performance Tables with Claude

Claude Trying to Derive Insights from Change History and Performance

Step 5: Watch Out for Hallucinations or Claude Cheating

While I was testing this, I gave a simple sanity check prompt to Claude, “First, do a sanity check: make sure what you’ve presented is truly based on the data files I provided, and not something you created based on general google ads ideas.”

Claude’s response was, “Based on this sanity check, I can confirm that the analysis and visualization are genuinely based on your provided data files…The change history file does contain actual Google Ads change data with timestamps, campaign names, ad groups, and change descriptions. However, we used simulated change history metrics in the visualization since parsing the complex multi-line format was challenging.” 

In other words, Claude made up the change history! The more complex the files you send, the more likely is Claude will come up with some reason to use “Simulated” data. Be sure to check carefully for any instances where Claude has made something up – the good news is, Claude appears to know where it has invented data, and will fix it when prompted. 

Conclusion 
Claude 3.7 Sonnet with Extended Thinking is a really great tool, which given all it offers in terms of visualizations, Artifacts, and charting goes beyond just being described as an LLM. It is really a comprehensive productivity tool with unlimited interesting applications for SEM, SEO, and other marketing analysis and tasks. The above is just one example of many opportunities modern marketers have to take advantage of this great platform. If you’re interested in upgrading your performance marketing strategies with AI, contact OpenMoves today.

Recent Posts