When the Consultant Leaves: Making Data Systems Sustainable

In this series, I’ve argued that data should be a flashlight, not a hammer. I’ve also talked about building a data culture as a way to bridge the gap between leadership and the people doing the work every day.

However, there is a practical question that often gets overlooked: What happens after the consultant leaves?

This is where many organizations struggle. The dashboard looks great while the consultant is involved. Then six months later, it stops refreshing. The person who understood the formulas is gone. Staff start creating manual workarounds again.

The pipeline worked. It just wasn’t built to last.

A consultant shouldn’t leave behind a mystery machine. They should leave behind capacity. A toolbox, not a tollbooth.

For most nonprofits, this does not mean buying an expensive new system. It may mean making better use of tools the organization already has, such as SharePoint, Microsoft Lists, Excel, Power Query, Power BI, and Power Automate.

A few things can make a big difference.

Start with the decision, not the dashboard.

Before asking what you can measure, ask what you are trying to improve. Who needs the information? How often do they need it? What are they actually going to do with it?

Draw the pipeline on one page.

Show the path from intake to outcome, from outcome to report, and from report to decision. If staff cannot explain how the data moves through the system, the process may be more fragile than it looks.

Clean the data when it is entered.

Use dropdowns instead of relying on free text. Standardize dates. Create a unique client ID. Agree on what key terms actually mean.

A dashboard built on unclear definitions is not insight. It is decoration.

Make the transformations explainable.

People should be able to understand how a number was calculated and where it came from. If an organization cannot explain how an important number was produced, that number probably should not be driving an important decision.

Build one useful report first.

It is tempting to build everything at once. Resist that urge. Start with something staff will actually use.

Then test it. Can staff understand it? Can someone inside the organization refresh it? Can they tell when something looks wrong?

Those questions matter more than how impressive the dashboard looks.

Leave a runbook.

Document where the data lives, who owns it, how the system refreshes, and what to do when something goes wrong. A short recorded walkthrough can help too.

Staff turnover is real. The person who knows how everything works today may not be the person doing the job next year.

Cross-train people.

At least two people should understand the system well enough to keep it moving. Ideally, there is a program owner, someone who understands the technical side, and a leadership sponsor who understands why the data matters.

You should never be one resignation away from losing your data system.

Use AI as an assistant, not the foundation.

AI can help speed up work that is already organized and well governed. It cannot fix a system where nobody agrees on definitions, ownership, or where the data lives. In that situation, AI can just help you create confusion faster.

A sustainable data pipeline is not the most sophisticated system an organization can afford. It is the simplest reliable system the organization can maintain.

The goal is not to become a technology company. The goal is to become a learning organization.