Thinklytics

Stephens Inc. · Financial Services · New York, NY · 12 weeks

Deal pipeline data foundation built in 12 weeks

A boutique investment bank was wasting two full days each week compiling pipeline reports from six separate data sources. We created a standardized deal pipeline data system that cut report prep time to 90 minutes and saved $420,000 a year in analyst labor costs.

Challenge

The investment banking team tracked deal pipeline data in six separate places: CRM, Excel, email, SharePoint, deal room, and Bloomberg. Two senior analysts spent two full days each week manually compiling the report. Managing directors often challenged the report’s accuracy because the data was inconsistent and fragmented.

Approach

We created automated connectors for six data sources and developed a certified deal pipeline model in Snowflake. Then, we replaced the manual Excel report with a Power BI dashboard that allows drill-down by sector, stage, and managing director. We also set up a data entry governance process to stop disconnected data issues from happening again.

Outcome

We cut weekly pipeline report prep from two days to 90 minutes by automating data pulls and standardizing metrics. This freed up $420K in analyst time annually. After launch, the managing director stopped raising data accuracy concerns, zero questions in the first month, showing the improved reliability of our reporting.

We dug into email folders and hit the jackpot. That’s where the real data gold was for our analysis.

We stumbled on an email folder stuffed with deal updates that weren’t showing up anywhere else. Pretty goldmine stuff. So, we built a quick NLP pipeline to snag the deal stage changes from those emails and fed that straight into our model. It was a simple move, but it seriously boosted our results.

We laid down simple rules to stop the same issues from coming back.

For five years, deal data was scattered everywhere, and frankly, no one really took charge of it. So, we gathered all that info into one system. Then, we put rules in place to catch any entries made outside of it. That gave ops instant visibility and helped keep the data clean.

We missed our data accuracy goal in month one, here’s the story.

Every week, the managing directors kept grilling us about the pipeline numbers. So, we just went for it. We built a certified pipeline dashboard that cut through the data chaos and stopped all the second-guessing.

Results

  • $420K Annual analyst labor saved
  • 2 days to 90 min Pipeline report preparation
  • 6 to 1 Data sources unified
  • 0 Data accuracy questions in month 1

Before, we’d spend two full days a week just arguing over pipeline numbers. Thinklytics created a system where we all get the data from the same place. Since then, not once have we had a disagreement about data accuracy in any meeting.

Head of Investment Banking Operations, Boutique Investment Bank

In the client's terms

The question

Where does this number come from, and can the path be trusted?

The situation

Deal pipeline lived in six places: CRM, Excel, email, SharePoint, the deal room and Bloomberg. Two senior analysts spent two full days a week assembling the report, and managing directors regularly disputed it.

What we found

The disputes were the useful signal. The report was not trusted because six sources disagreed and the reconciliation happened invisibly inside a spreadsheet, so there was no way to settle an argument about a number.

What it produced

Weekly preparation fell from two days to 90 minutes, worth $420K a year in analyst time. Six sources became one. Data accuracy questions from managing directors in the first month after launch: zero.

The roadmap

Automated connectors to all six sources and a certified pipeline model in Snowflake, so the reconciliation is explicit and repeatable. The Excel report was replaced by a Power BI view with drill-down by sector, stage and managing director. Entry governance was added so the sources stop diverging again.

The work behind this result

Related case studies

Frequently asked questions

Why do managing directors dispute pipeline reports?

Because the reconciliation is invisible. Deal data at Stephens Inc. lived in six places including email, and two analysts assembled the weekly report by hand, so there was no way to settle an argument about a number. After a certified model, accuracy questions in the first month fell to zero.

How much time does manual pipeline reporting consume?

Two full days a week from two senior analysts here, worth $420K a year. Automated connectors and a certified model in Snowflake took that to 90 minutes.

What do you do about deal data that only exists in email?

Extract it rather than ignore it. Stage changes at Stephens were being recorded in email and nowhere else, so a short NLP pipeline pulls them from that mailbox into the model alongside the other five sources.