1) maybe this is by design, but the the conversion rates used in 1 and 2 vs 3 are different. 1 and 2 use creation date as the denominator and 3 uses close date. I find using different definitions like this can confuse business leaders.
2) have you thought about fully automating it where there is no need to run it weekly but rather the agent does it directly by connecting to source systems and then sending emails to stakeholders once the analysis is complete?
3) How are you doing trending to see how 1-8 are performing over time? Are snapshots being saved anywhere?
4) I always wonder if using AI like this or if modern BI tools like Snowflake and Tableau are both more reliable and efficient (e.g for snapshot trend analysis, declarative logic accuracy, and no need to have human intervention to routinely run). What is built with Claude Code is cool, but wondering how much time in total was spend learning, iterating to get right, and maintaining versus doing it with modern BI tools?
Thanks for reading and for the comments. Here are some thoughts back on your questions:
1 - it's absolutely by design. Pipeline Conversion (Based on Created Quarter) When you look at conversion based on the Created Quarter, you are performing a cohort analysis. This groups opportunities by their "birth date," allowing you to isolate the quality of the inputs—such as specific marketing campaigns, SDR outreach strategies, or market conditions—at that specific point in time. If you measured conversion based on close date, you would be mixing leads generated yesterday with leads generated two years ago, making it impossible to see if your current lead generation engine is improving or declining. By anchoring to the creation date, you can track the true throughput of a specific vintage of leads to answer the question: "Of the opportunities we generated in Q1, what percentage eventually turned into revenue?"
Win/Loss Rates (Based on Fiscal Quarter) Conversely, Win/Loss Rates are typically tied to the Fiscal Quarter because they measure sales execution and financial performance within a specific accounting period. Sales teams live and die by their quotas, which are set on fiscal timelines; therefore, the win rate should reflect the outcome of the deals that were "on the table" to be closed during that period. This aligns your metrics with revenue recognition and answers the question: "Of the decisions made by customers this quarter, how many did we win?" This approach ignores when the lead was born and focuses entirely on the closing ability of the sales team during that financial window.
2 - yes - I'm looking at either direct API access, login, or using an MCP. It's on my to-do list - just not quite at a high enough priority but definitely something I plan to get to.
3 - Once I have a baseline of these 8 quarters, I would just have the agent drop the new quarterly results into the master spreadsheet. I'd change the flow a bit in the future now that I have the baseline. One thing it also does it grabs the full sales snapshot and keeps it so we also have the record of what it is analyzing
4 - The one difference I see in doing it this way is that I now have an intelligent reasoning engine that was looking at all of the calculations going through and creating an executive summary for me. It certainly doesn't replace the final thinking, but gives me a good place to start exploring. I don't know the other tools well enough to know if they can do that (I don't think they can, but since I don't use them regularly, I don't want to completely rule it out). The other advantage of using AI is I started by tallking to it and telling it what I wanted it to do to. I did very little typing. Through iteration it built 90%+ of that instructions file based on me describing tasks.
This is great. A couple thoughts / questions.
1) maybe this is by design, but the the conversion rates used in 1 and 2 vs 3 are different. 1 and 2 use creation date as the denominator and 3 uses close date. I find using different definitions like this can confuse business leaders.
2) have you thought about fully automating it where there is no need to run it weekly but rather the agent does it directly by connecting to source systems and then sending emails to stakeholders once the analysis is complete?
3) How are you doing trending to see how 1-8 are performing over time? Are snapshots being saved anywhere?
4) I always wonder if using AI like this or if modern BI tools like Snowflake and Tableau are both more reliable and efficient (e.g for snapshot trend analysis, declarative logic accuracy, and no need to have human intervention to routinely run). What is built with Claude Code is cool, but wondering how much time in total was spend learning, iterating to get right, and maintaining versus doing it with modern BI tools?
Hi Kory,
Thanks for reading and for the comments. Here are some thoughts back on your questions:
1 - it's absolutely by design. Pipeline Conversion (Based on Created Quarter) When you look at conversion based on the Created Quarter, you are performing a cohort analysis. This groups opportunities by their "birth date," allowing you to isolate the quality of the inputs—such as specific marketing campaigns, SDR outreach strategies, or market conditions—at that specific point in time. If you measured conversion based on close date, you would be mixing leads generated yesterday with leads generated two years ago, making it impossible to see if your current lead generation engine is improving or declining. By anchoring to the creation date, you can track the true throughput of a specific vintage of leads to answer the question: "Of the opportunities we generated in Q1, what percentage eventually turned into revenue?"
Win/Loss Rates (Based on Fiscal Quarter) Conversely, Win/Loss Rates are typically tied to the Fiscal Quarter because they measure sales execution and financial performance within a specific accounting period. Sales teams live and die by their quotas, which are set on fiscal timelines; therefore, the win rate should reflect the outcome of the deals that were "on the table" to be closed during that period. This aligns your metrics with revenue recognition and answers the question: "Of the decisions made by customers this quarter, how many did we win?" This approach ignores when the lead was born and focuses entirely on the closing ability of the sales team during that financial window.
2 - yes - I'm looking at either direct API access, login, or using an MCP. It's on my to-do list - just not quite at a high enough priority but definitely something I plan to get to.
3 - Once I have a baseline of these 8 quarters, I would just have the agent drop the new quarterly results into the master spreadsheet. I'd change the flow a bit in the future now that I have the baseline. One thing it also does it grabs the full sales snapshot and keeps it so we also have the record of what it is analyzing
4 - The one difference I see in doing it this way is that I now have an intelligent reasoning engine that was looking at all of the calculations going through and creating an executive summary for me. It certainly doesn't replace the final thinking, but gives me a good place to start exploring. I don't know the other tools well enough to know if they can do that (I don't think they can, but since I don't use them regularly, I don't want to completely rule it out). The other advantage of using AI is I started by tallking to it and telling it what I wanted it to do to. I did very little typing. Through iteration it built 90%+ of that instructions file based on me describing tasks.