I watched an analyst spend an entire morning copying numbers from one spreadsheet to another. That's when I knew what we needed to build.
- Cut analyst time per deal by 75% (5 hrs → 1.25 hrs)
- Supported $1.8B in annual deal volume
- Achieved 100% adoption through training and iteration
- Reduced support tickets by 10%
The Backstory
PFM's Capital Markets team advised clients on complex financial transactions. Municipal bonds, debt restructuring, capital planning. High-stakes work where the analysis actually mattered.
But the team had a problem nobody was talking about. The analysts, some of the smartest people I've worked with, were spending most of their time on tasks that had nothing to do with analysis.
I was an Associate Product Manager embedded with the team. My job was to figure out how to make them more effective. So I started by shutting up and watching.
A Day in the Life
Here's what a typical deal looked like:
One analyst put it bluntly: "I have a finance degree and I spend 80% of my time doing data entry."
The Real Problem
It would have been easy to say "they need better tools" and start building. But I wanted to understand the deeper issue.
Why had it gotten this bad? The team wasn't incompetent. They were busy. Every analyst had developed their own shortcuts and workarounds over the years. Those personal systems worked well enough that nobody stopped to fix the underlying mess.
And the mess compounded. Historical deal data was trapped in individual folders. When an analyst left, their knowledge walked out the door with them. There was no institutional memory, just a collection of personal filing systems.
The team also couldn't scale. Taking on more deal volume meant hiring more analysts to do the same manual work. The bottleneck wasn't thinking. It was data janitoring.
What Would Actually Help
I spent three months on discovery before proposing anything. Interviewed every analyst on the team. Mapped their workflows in embarrassing detail. Documented every spreadsheet, every database query, every manual step.
Don't try to change how they think
The analysts were good at analysis. They didn't need AI to tell them what a deal should look like. They needed the grunt work taken off their plate so they could focus on judgment.
Standardize the boring stuff
Calculations and report formats didn't need to vary analyst to analyst. Pick a methodology, document it, and make the system enforce it.
Make historical data accessible
Every deal the team had ever worked on should be searchable, comparable, and useful. Not buried in someone's personal folder.
Don't build for power users
Build for the analyst who just joined the team and doesn't know where anything is. If it works for them, it works for everyone.
What We Built
Automated Data Ingestion
The platform connects to the databases and pulls what you need. No more scavenger hunts. No more copying and pasting. Type in the deal parameters and the data appears.
Calculation Engine
One methodology, documented and consistent. The system does the math the same way every time. Analysts can focus on interpreting results instead of double-checking formulas.
One-Click Reports
Standard templates that populate automatically. The formatting is done. The sections are consistent. An analyst can generate a report in the time it used to take to set up the header.
Historical Deal Database
Every transaction the team has worked on, searchable and comparable. Want to see how this deal stacks up against similar ones from the last five years? That used to be a week-long research project. Now it's a query.
Getting People to Use It
This is where most internal tools fail. They get built, announced, and ignored. People go back to their spreadsheets because that's what they know.
I refused to let that happen.
I built the training program myself. Not a generic onboarding deck, but role-specific workshops where analysts learned the platform by doing their actual work. I recorded video tutorials for every major workflow. I wrote documentation that answered real questions, not imaginary ones.
Then I set up support channels and actually monitored them. Every question was a signal. Every repeated question meant something was confusing. I used support ticket patterns to prioritize improvements.
Three months after launch, we hit 100% adoption. Not because people were forced to use it. Because it was genuinely better than the alternative.
The Results
Analysts got four hours back per deal. That time went into actual analysis, not data entry. The team supported $1.8B in annual deal volume without adding headcount.
The 10% reduction in support tickets might seem small, but it meant something important. The platform was getting easier to use over time, not harder. We were actually learning from our users and improving.
Why This One Stuck With Me
Internal tools are unglamorous. Nobody writes blog posts about the analytics platform that made a finance team 75% more efficient. There's no TechCrunch headline.
But I think about that analyst I watched on my first week. Spending her morning copying numbers between spreadsheets. Smart, capable, stuck doing work that didn't use any of her skills.
CMAP gave her those hours back. Multiplied across the whole team, across hundreds of deals, that adds up to something real. Better analysis. Better client outcomes. People doing work that actually matters.
That's the kind of product I want to build.
My Role
- Led end-to-end product development as Associate Product Manager
- Spent three months on discovery before writing a single requirement
- Translated workflow analysis into product strategy and prioritized roadmap
- Worked with engineers on automated data ingestion and reporting
- Built training program that drove 100% adoption
- Iterated based on support patterns, reducing tickets by 10%