When a seemingly promising idea loses efficacy or profitability as it expands, we call it a “voltage drop.” These failures to scale never happen because of one single reason.

Over the last 25 years as a behavioral economist, consultant to companies large and small, and former White House economic adviser, I’ve identified five causes, what I call “vital signs,” of voltage drops.

[Here are the first two.]

1. False Positives 

This occurs when you interpret a piece of evidence or data as proof that something is true, when in fact it isn’t — for example, as we’ve seen with inaccurate Covid test results. For scaling, a false positive is an erroneous sign that an idea has voltage when it really doesn’t.

Sometimes a false positive occurs because of a statistical error, as was the case with the famous drug abuse prevention program, D.A.R.E. After an independent study showed promising short-term results, the program received an influx of funding from the U.S. Department of Justice. However, there were several problems with the study: It excluded drugs like alcohol and marijuana, focusing on tobacco; it was based on a small sample size; and later studies and even metanalyses could not replicate the results.

In other cases, false positives result from intentional lying. Think of Elizabeth Holmes and the purportedly groundbreaking blood-testing technology of Theranos, which didn’t actually exist.

When possible, the solution for rooting out false positives is to have at least three independent replications of the idea that show early promise. In companies with confidential research, employees must be incentivized with financial rewards that encourage them to question results.

2. Biased Representativeness of Population

Once you’ve reliably demonstrated the efficacy of the endeavor you hope to scale, the next step is to answer the question “How broadly will the idea work?”

All ventures must understand their potential audience. The first way to do this is by making sure your test samples in the small scale reflect the larger population at scale. Otherwise, you’ll be like McDonald’s, which fell victim to selection bias when it launched the unsuccessful Arch Deluxe. Focus group participants loved the new product, but they weren’t representative of the majority of Americans, who simply wanted to keep eating their Big Macs.

To weed out such biases, make sure your early adopters are a random sample. You should also make sure that your survey respondents have appropriate incentives to tell you the truth. A focus group participant who says they would purchase a product if it was introduced could simply be saying, “I would love the option to consider that product in the future,” as opposed to “I will be purchasing the product in the future.”

Another strategy is to create models that don’t rely on top-tier talent. As you scale, finding and paying high-performers will become prohibitive. The solution is to create products that can give their full value to customers even with average performers delivering it.

When I give talks on this topic, I like to invoke the famous opening line of Leo Tolstoy’s novel Anna Karenina: “Happy families are all alike; every unhappy family is unhappy in its own way.” Similarly, scalable ideas are all alike; every unscalable idea is unscalable in its own way. The difference with scaling is there are only five main obstacles you face. And once you anticipate and avoid them, you can scale your idea for the highest voltage possible.