Data-Driven Decisions: We Have the Dashboard. Now What?

For years, much of nonprofit data strategy has focused on getting data ready to use. We had information scattered across fundraising systems, email platforms, websites, advocacy tools, event systems, program databases, spreadsheets, and who knows what else. Before we could do much with any of it, we had to bring it together.

Of course, bringing it together was only the beginning. Once the data was in the same place, we had to normalize it, reconcile people who appeared differently across systems, decide which sources we trusted, and agree on what things actually meant. An active donor sounds straightforward until five people in a room offer six definitions. The same goes for retention, engagement, household, sustainer, member, or almost any other term we depend on to understand constituent relationships.

None of that work is finished, and I’m not sure it ever will be. But we’re much better at it than we used to be. More organizations have connected data, stronger analytical models, better identity resolution, shared definitions, and dashboards they trust. Increasingly, AI is making it possible for more people to explore that data without understanding the underlying structures or waiting for someone to build a report.

After years of work to get here, though, I think we’re beginning to see the next challenge. We can finally answer questions we couldn’t answer before, but that doesn’t necessarily mean we know which questions to ask. And even when we get an answer, we don’t always know what to ask next to turn that answer into something we can act on.

The Question Still Comes Before the Dashboard

A few posts ago, I wrote that the question should always come before the dashboard. At the time, I was thinking about our tendency to begin analytics work with the data we have or the things we know how to measure. We build a dashboard full of retention rates, response rates, average gifts, sustainer performance, acquisition results, engagement measures, and dozens of other metrics, and then sit down to see what they tell us.

I still think we have that backward. But the more I’ve thought about it, the more I think the issue goes beyond dashboard design. Knowing what question to ask is becoming an increasingly important part of using data well.

Take donor retention. We can build a perfectly good dashboard that tells us retention is down. We trust the data, agree on the definition, look at the trend over time, and can reasonably be confident the number on the screen is correct. That’s valuable information, but it isn’t particularly actionable on its own. “Retention is down” describes what happened. It doesn’t tell us why it happened, whether we should be concerned about it, or what we might do differently.

That’s when the questions start to matter. Is the decline concentrated among first-year donors or happening across the file? Is it associated with a particular acquisition source? Has the donor population composition changed? Are sustainers behaving differently? If retention is lower among donors acquired through one channel, are those donors producing enough additional revenue or long-term value that the lower retention is actually an acceptable tradeoff?

Each answer gives us a little more understanding, but it also gives us a better question. Good analytics isn’t simply about getting to the answer. It’s about knowing how to keep following the answer until we understand enough to decide.

The Bottleneck Is Moving

This becomes especially interesting as AI begins to change how we interact with data. As Karen Taggart recently wrote in AI Requires a Strong Data Foundation, connected data, identity resolution, shared definitions, semantics, and consistent analytical structures have to exist underneath AI before we can rely on what it tells us. If the underlying data is ambiguous, giving someone a natural-language interface simply lets them get an ambiguous answer faster.

But imagine we’ve done that foundational work reasonably well. Something that once required a reporting request, knowledge of the underlying data model, and perhaps several rounds of analysis can become a conversation. Someone asks why retention declined, looks at the answer, narrows the population, compares acquisition sources, changes the timeframe, explores a cohort, and follows an unexpected result without starting a new reporting process every time.

That’s an enormous opportunity, but it doesn’t eliminate the need for analytical thinking. In some ways, it makes it more important. If the friction involved in asking a question approaches zero, the scarce resource is no longer access to the data. It’s knowing what is worth asking and recognizing when an answer should lead us somewhere else.

For years, one of the biggest bottlenecks in nonprofit analytics was getting the information. Increasingly, the bottleneck may be knowing what to do with it.

Sometimes Our Own Data Isn’t Enough

Even when we ask good questions, our data can’t tell us everything. For example, we might know that donor retention has fallen from 65 percent to 62 percent. We can compare that with our own history, identify which audiences are driving the change, and investigate whether something about our acquisition or cultivation strategy has shifted. Our own data can’t tell us whether 62 percent is unusual.

This is where benchmarking can add a different kind of context. If comparable organizations are retaining donors at 70 percent, we suddenly know something we couldn’t know by looking only at ourselves. The difference may deserve more attention than we originally thought, but the benchmark hasn’t told us what to do. In fact, it has mostly given us more questions.

Perhaps our acquisition mix is different. Maybe we’re acquiring more aggressively through a channel that produces lower first-year retention but greater long-term value. Maybe our acknowledgment or cultivation strategy needs attention. Maybe our donor population is changing. Perhaps our peers are doing something we should understand. Or maybe the organizations we’re comparing ourselves with aren’t as comparable as we assumed.

This is why I don’t think the goal of benchmarking is necessarily to get closer to the benchmark. If we’re at 62 percent and the benchmark is 70 percent, it’s tempting to assume success means moving toward 70. But 62 percent could be the rational result of a deliberate strategy that’s producing better outcomes somewhere else.

The benchmark hasn’t diagnosed the problem. It has given us context and helped us identify a better question. The real value is understanding why we’re different, deciding whether that difference matters, and determining what, if anything, we should do about it.

The Answer Is Part of a Larger Conversation

Maybe this is where our mental model of analytics needs to change. We tend to think of the answer as the destination. Someone asks a question, we build the report, the dashboard shows the result, or the benchmark provides the comparison. The analytical work feels complete because we got the number.

But the decisions nonprofits are trying to make rarely work that way. An answer gives us something to observe. We add context to determine whether it deserves attention. That leads to more investigation, which eventually allows us to form a hypothesis about what’s happening. Only then are we in a position to decide whether we should change something, and even then it isn’t the end, because we need to measure what happened after we acted.

Did the change work? Did the metric move? Did something else move that we didn’t expect? Was our original hypothesis correct? Whatever we learn becomes context for the next question.

Question → Answer → Context → Better Question → Investigation → Hypothesis → Action → Measurement → Learning → Next Question

I don’t mean every interesting metric needs to go through a ten-step methodology. Sometimes investigation tells us a difference doesn’t matter. Sometimes we don’t have enough evidence yet and need to keep watching. And, as I’ve written before, sometimes the next best action really is no action at all.

But no action can still be a decision informed by what we’ve learned. That’s very different from putting a chart on the screen, agreeing that it’s interesting, and moving on to the next slide.

More Answers Won’t Automatically Create Better Decisions

I suspect this becomes more important, not less, as our analytical capabilities improve. AI will make it easier to explore data. Connected systems will let us see more dimensions of constituent relationships. Better dashboards will make information available to more people. Benchmarking can provide external context that was previously difficult or impossible to obtain.

We’re going to have a lot more answers. But an organization can drown in insights just as easily as it can drown in data. If every meeting produces five interesting observations and none leads to another question, a hypothesis, a decision, an experiment, or even a conscious choice to keep watching, then better analytics haven’t necessarily made us more data-driven. They’ve just given us more sophisticated things to look at.

That’s why I think the next stage of nonprofit data maturity may be less about another piece of technology and more about developing the habits required to use the technology we’ve built. We’ve invested in collecting the data, bringing it together, resolving identities, agreeing on definitions, and building analytical structures that let us trust what we’re seeing. Now we need to become equally deliberate about asking questions, following where the answers lead, deciding when we know enough to act, and learning from what happens afterward.

Maybe We Should Design Analytics Backward From the Decision

That changes the way I think about the dashboard itself. Instead of starting with What metrics should we show?, perhaps we should start with What are we trying to understand, and what decisions could that understanding help us make?

If we’re tracking donor retention, what would we want to understand when it moves? If we’re looking at sustainer performance, what are we prepared to investigate when something changes? If we’re measuring engagement across channels, what would cause us to rethink our assumptions about a constituent relationship? If we’re benchmarking ourselves against similar organizations, which differences would actually be meaningful enough to investigate?

We don’t have to know in advance what the data will tell us. One of the great values of analytics is discovering something we didn’t expect. But we can be more intentional about what happens when we do. Who asks the next question? What additional context would help us understand the result? What would give us enough confidence to try something different? If we do, what will we measure afterward?

Those aren’t really dashboard questions. They’re questions about how an organization learns.

The Dashboard Was Never the Destination

For years, many of us working in nonprofit technology have imagined a future in which organizations could connect their data, trust their reporting, understand constituents more completely, compare their performance meaningfully, and get answers without waiting days or weeks for somebody to assemble them. We’re getting closer to that future, and rather than making the human part of the work less important, I think we’re making it more visible.

When the data is fragmented, we spend our energy bringing it together. When definitions don’t match, we spend our energy reconciling them. When reporting is difficult, we spend our energy getting the report. As those problems become easier to solve, more of our attention can finally move to the part technology was supposed to help us with all along: figuring out what we’re trying to understand, asking better questions, deciding what matters, trying something, and learning from what happens next.

A few posts ago, I argued that the question should come before the dashboard. I’m increasingly convinced that the most important thing that comes after the dashboard may be another question. The two ideas are really part of the same loop. What we learn should make us better at asking what comes next.

For years, nonprofit data strategy has focused on building the foundation that helps us get better answers. Maybe the next stage is developing the organizational muscle to ask better questions of those answers and, eventually, turn what we learn into better decisions.

We have the dashboard. Now what?

Are you wondering what should come next? Let’s Talk!