How do you make more accurate predictions? AngelList looked beyond spreadsheets

“I remember this one time, I was at a friend’s wedding. Everyone was out by the pool,” Derek said. “I could hear my friends laughing and enjoying the sun, and there I was…working on my spreadsheet.”

Derek Ou-Ponticelli heads Strategic Finance at AngelList, a software platform that currently supports $124B assets on platform and over 85K investors. AngelList's mission is bold—to build the infrastructure that powers the startup economy. That involves building software that allows fund managers to seamlessly launch or scale a fund without the back office headache, streamlines confusing paperwork and logistics for investors, and provides founders with a single place to raise capital and manage their equity.

AngelList’s scale, coupled with its broad exposure to private markets, comes with a ton of data. And Derek’s team sits at the heart of it all: tracking financials, modifying financial strategy based on current metrics and macroeconomic trends, and shaping AngelList’s strategic financial future. They constantly look ahead to make forecasts, while looking back to see which of their previous assumptions were correct.

To do that more efficiently, what they needed was to build increasingly powerful models–ones that would predict with better accuracy. Instead, they found themselves validating data and battling bulky spreadsheets.

The sheer volume of data

Derek’s team tracked a huge volume of data, pooled into their database by different teams. While all the data they needed was right there, it lived across multiple platforms. Finding the right data query to answer a specific question could be a challenge. “Say you’re looking for the number of funds started in a year. You’d search for all the data queries people had ever written. You’d get all these queries from different people written at different points in time, making it hard to know which data source to use..”

Pulling fresh data was a monthly task. Derek would run his data queries every month, then paste all of the data into a spreadsheet. He would then go through it all and validate, update, double-check formulas and results. That’s how he got the next month’s forecast.

For Derek’s team, looking back was a huge part of forecasting–it helped them identify what had changed between the previous and the current forecasts. They found themselves juggling countless spreadsheet versions to keep track of everything. “I'd have 18 billion versions saved and everything all hard-coded everywhere,” Derek said. “You know what it's like when a spreadsheet starts out really clean and then, over time, you add some stuff onto it, and then you add more stuff onto it, until it turns into Frankenstein’s monster. Eventually, I had to spend a month cleaning it all up.”

On top of that, AngelList's culture of rapid development brought new products to market frequently. Whenever that happened, the multiple spreadsheet versions they were juggling had to be updated to include the new product lines.

Finding just what he needed

At a product demo, Derek found out he could use Runway to consolidate data and get instant insights. “When I saw the different scenarios you could make and keep track of, I knew Runway could help us,” Derek said. Earlier, when he had to go copy-pasting formulas around in spreadsheets, he’d be the only one to understand the model. But not anymore. “Runway works with all of the core elements of every finance person’s forecast: the raw data, the actuals, forecast methodologies, and visualizations. But it's much more efficient than doing it on spreadsheets.”

Derek’s team moved away from the constant back-and-forth of pulling and updating data. Instead of juggling countless spreadsheet versions, they leveraged Runway to evaluate the financial implications of their strategic initiatives in real-time. “With Runway, we now have a sandbox where we can play around with different what-if scenarios.”

“Even the process of putting together the business review–it used to take hours, with all of the data and the charts, and now it takes a fraction of the time.” In addition to making sure the numbers were updated and accurate ahead of meetings, Derek found more time to interrogate the assumptions and imagine different upside and downside cases.

Transitioning to Runway was super easy too. “I feel like we've been spoiled with support. We've been meeting weekly for months now, and the Runway team has invested a lot in setting up the latest and the most efficient way of looking at data for us. There’s been a lot of custom building,” Derek said, referring to the dynamic features the Runway team built, which enabled AngelList to leverage their existing cohort-based models.

Creating new ways to grow

Using Runway, Derek and his team could more easily see how capital flowed through AngelList, and how key metrics were affected by the moving parts of the business–like the number of startups signing up or the amount of capital closed by fund managers. Next, they leveraged the power of predictive cohort-based modeling. “Once we realized we could do things more efficiently in Runway than on spreadsheets, we invested time to build cohort modeling, which was a pain to maintain in the past. Now we save time, and get deeper insights.”

What kind of insights? There could be a long period of time between when a fund launches compared to when they complete fundraising with investors. With the new cohort models that they built in Runway, Derek’s team could make more accurate predictions about when funds would hold their first close.

“Things were generally going fine,” Derek said, “but there are new points of growth added because of Runway that might have taken us longer to identify otherwise.”

“It’s about more than just saving time. With Runway, spotting trends is easier—now we can more efficiently tell stakeholders what number a metric’s going to be and why.” Runway has helped AngelList’s finance team gain a deeper understanding of their business and make more confident predictions.