Not all grants pan out perfectly—but our failures are opportunities to learn.

As grantmakers, we love a good success story.

When we sit down with nonprofits to discuss a grant, we work together to identify a project that can deliver impact, ideally at scale, and that furthers both the nonprofit’s mission and our own as a funder. (At Tableau Foundation, we —of course!—invest in projects that effectively use data for solving problems.) Once we’ve landed on the project and its scope, we’re eager to see signs of success—key performance indicators (KPIs) being met, impact being made on the ground.

Sounds simple, right? Work with a nonprofit to identify a project to fund, and watch that funding do what everyone is hoping it will do.

In reality, we all know grantmaking is much more complicated than that. So, I’m going to share some key learnings on a grant we awarded that did not pan out the way we envisioned.

Back in 2015, when the Tableau Foundation was just getting started with grantmaking, we connected with Feeding America?—the largest domestic hunger-relief organization in the U.S.—about supporting their idea to transform the way it used data. The Feeding America national office is based in Chicago, but it’s comprised of a network of 200 independently-run member food banks across the U.S. The organization's goal was to build a culture of data use and reporting both within the national office and across all 200 food banks. They’d been successful in working with member food banks on creating 12 metrics that they each would measure and report. But they also needed a way for the food banks to track the KPIs. They asked Tableau Foundation for help.

Feeding America had the vision to roll out Tableau to all 200 food banks, and then train their staff to use the platform to analyze data. The idea was that as the food banks grew their data capacity, they could feed their expertise back into the network—so if a food bank in New Mexico designed a dashboard that was really helping its performance, Feeding America’s national office could adapt that tool and send it out for the rest of the network to use.

It was a big, expansive vision for organizational change that had as its end goal the Feeding America network of food banks being able to serve their communities better.

But as funders, we did not initially appreciate the breadth of this project, and what it meant for our commitment to accomplishing its goals. In short, we didn’t fund this project well enough.

We now know, after more than five years of grantmaking, that massive, transformational projects like this require significant, multi-year, flexible funding and ongoing support. For this project, we granted $50,000 in financial support for one year, along with around $1 million worth of software. We also established a “data fellowship” program within Feeding America, which funded 15 people from the various food banks to learn Tableau and build out dashboards that, ideally, could be used by the network. As we started to work with the Feeding America network, we recognized that each food bank was very diverse—but we didn’t quite realize what a challenge that would pose for implementing the project. From geographic to socioeconomic to their internal capacity, there were so many differences between each food bank that the expertise of a handful of data fellows wasn’t able to permeate through the entire network.

Many of the food banks within the Feeding America network found success using Tableau, and are still using it to track and report their metrics. But in order to have been successful in scaling the use of Tableau and advanced analytics organically throughout the network, we would have had to fund it very differently. In short, we would have had to fund it like a partnership, not like an experiment.

Let me explain a little more. When thinking about grantmaking now, it’s helpful for me to think along three different levels.

An experiment is something of an exploratory grant. It’s basically—for Tableau Foundation—giving a nonprofit a certain amount of software and support and letting them experiment with how it might transform their organization and their work. It’s pretty light touch, and designed to see if there’s potential for a more involved project out of it (at Tableau Foundation, we now offer nonprofits a “package grant”—a set amount of software and training —as a first step and potential gateway to a more in-depth relationship).

A more involved project might be a pilot: a more structured, clearly delineated project with slightly longer-term funding and more involvement and support from funders. While with experiments, the aim of the grant might not be anything more specific than just “see what happens,” pilots have clearly defined aims.

Out of many experiments or pilots, a few evolve into partnerships, which I think of as the most expansive kind of grantmaking we do. While many of our partnerships emerge out of pilots, some recent ones—like our partnership with Bridges to Prosperity, an NGO that constructs footbridges to connect rural communities isolated by impassable rivers to economic centers—have grown directly from experiments. What it takes is a strong understanding within the partner organization of how impactful data can be and a clear vision for how they’ll implement it across their work. With our partnerships, we aim to support nonprofits taking on projects that are wide in scope and transformative in what they’re aiming to accomplish. We know, going in, that they won’t see results overnight, but with the nonprofit, we set out a path to success and commit to supporting them for the full journey. That’s what Feeding America’s vision required, but we didn’t deliver on it.

As grantmakers, we often think that if we’re not funding something that sounds like a partnership, we’re not doing enough. And that creates a cycle where nonprofits feel that they can only approach funders if they have an idea for an enormous, transformative project. Even if the first step may really be something at an experiment level–a nonprofit needing to figure out what dashboard works best for their needs, for example—they feel like they need to bypass the more incremental, small-scope phases in order to capture the attention of grantmakers with the next big idea.

Breaking our funding down into these three levels has—I hope—created more space for nonprofits to move incrementally when taking on something as transformative as building a network-wide culture of data. We’ve even had a handful of grantees, like Splash, start off with a package grant and eventually work up to partnership-level because they’ve been able to make incremental progress, proving to themselves and us that data could indeed be transformative.

We didn’t have this framework in mind when we started working with Feeding America—but it’s dramatically reshaped the way we think about grantmaking now. This year, we are renewing our grant to Feeding America, and leveraging our key learnings so we can enable Feeding America’s network of food banks to build a workable data culture from the ground up. We’ll be funding Feeding America’s central office more robustly so they can support all of the food banks to collect data more consistently and analyze it in real time to improve their operations, and continually reassessing progress for future funding. This grant, in contrast to the first one, will focus on building Feeding America’s data capacity from the inside out and ultimately create a strong culture of data across the entire organization.

Instead of pointing to the nonprofit when a grant doesn’t pan out as expected, I’d recommend first looking to yourself. Was what you as a funder offered well matched to the project? Did you fund an experiment but expect a partnership? Were expectations reasonably set from the get-go? Was the necessary support sustained throughout the duration of the grant, and was the grant long enough to actually bring the project to life? These are all important questions we ask ourselves, and if you find yourself answering “no” to some of them, it’s not always a cause to turn your back on the grantee. Rather, it can be a signal to reopen the conversation, incorporate what you’ve learned, and try again.