Stop Making Impossible Promises with Data

Your data can guide decisions without promising an exact result.

Weather forecasters never say “it will rain tomorrow.” They say there is a 70 percent chance of rain.

That small change makes a big difference. It tells you rain is likely, but it also leaves room for a dry day. Nobody gets mad at the forecast when it does not rain, because it was never a promise.

Data work should sound more like this.

When you say a new feature will bring exactly 1,200 users, that number can sound like a promise. If the feature brings in 1,050 users, people think you missed it, even though it was fairly close.

People often remember the missed target more than how close the estimate was.

A range gives them a better view of what may happen.

Instead of saying, “The feature will bring in 1,200 users,” you could say:

We expect the feature to bring in between 1,000 and 1,400 users, with 1,200 as our current best estimate.

The team still has a clear number to plan around. They also understand that the final result may be a little higher or lower.

Why Smart Analysts Use Ranges

A range tells the truth about what you know and what you do not know yet. It shows that you have thought about what could change. You are not pretending that the future is easy to predict.

This idea works for timelines too. Instead of saying a project will take two weeks, say it should take between two and three weeks, depending on how testing goes.

That small change helps the team plan for both outcomes. They can aim for two weeks while knowing the work may take a third week.

Election forecaster Nate Silver often explains that good predictions should be honest about uncertainty. Ranges are how that honesty shows up in everyday work, not just in big public forecasts.

A Range Is Not a Random Guess

Using a range does not mean picking two numbers that look safe. A useful range should come from evidence.

You can build the range from past results, changes in the data, or mistakes the model made before.

Suppose a company ran ten similar campaigns. The weakest campaign brought in 380 customers, while the strongest brought in 610. Most of the campaigns brought in between 450 and 550.

An analyst could use that history to build a reasonable range for the next campaign.

The low and high numbers are not random. They come from what happened before.

You should be able to explain where both numbers came from.

How Wide Is Too Wide

A range can also become so wide that it stops being useful.

Saying a project may cost between $20,000 and $100,000 does not help the team make a clear plan. The difference between the low and high estimate is too large.

A wide range may show that the team does not know enough yet.

That still tells the team something important. The analyst should explain why the range is so wide and what could make it smaller.

You may need to collect more data, break the problem into smaller parts, or look at different groups separately.

For example, instead of creating one sales range for all customers, you could build separate estimates for new customers, returning customers, and large business clients.

Each group may behave differently. Combining them may hide those differences and make the final range too wide.

Breaking the data into smaller groups can create estimates that are easier to use.

The Range Plus The Best Guess

A range works best when it comes with one best estimate.

The range shows what is reasonable. The central number shows what is most likely based on the current data.

A simple format is:

We expect the result to fall between X and Y, with Z as our current best estimate.

For example:

Delivery will likely take between 8 and 12 days, with 10 days as our best estimate.

The company may lose between 250 and 320 customers next quarter, with 285 as the most likely result.

The new pricing plan may increase monthly revenue by 6% to 10%, with 8% as our best estimate.

It gives people a clear number. It also shows them the number could shift a little.

Data is a tool to help people make better choices, not a magic crystal ball. Sharing a range takes away the pressure of being perfectly right.

It makes your work more useful and helps your team prepare for different results.