Crossing Hurdles

A hurdle is an obstacle in the sense of a predetermined benchmark. Someone sets a number, and only once a result exceeds that number does something happen. Crossing Hurdles describes exactly this moment of surpassing. In finance, the number is usually a minimum return, meaning the minimum profit an investment must generate. In technology, it can be a test score a program must reach before it is released. What both cases have in common: below the threshold, the result doesn’t count, above it, it does.

Why a single number decides bonuses and budgets

Thresholds make success verifiable. Without them, anyone could claim a result was good enough. A fixed hurdle preempts this discussion. All parties know in advance what they’ll be measured against.

A particularly large amount of money hinges on this in funds, meaning large pooled sums of money from investors. There, a hurdle of around eight percent annual return is common. Only once the fund exceeds this eight percent may the management retain a share of the profit. If it stays below, only the standard management fee applies. For a fund team, crossing this line can mean millions.

In tech companies, the principle works similarly, just without a dollar amount. An AI model is often only allowed to ship to customers once it reaches a certain accuracy rate in testing. This threshold protects against half-finished systems making their way into products. But it also shifts the focus: teams then work specifically toward hitting that number.

How a threshold is set and measured

It starts with choosing the metric. For an investment, this is usually the annual return in percent. For a software test, it can be the share of correctly solved tasks. It’s important that the metric can be calculated unambiguously. Otherwise, everyone ends up arguing about the measurement rather than the result.

Next, the level of the threshold is determined. For investments, this is often based on a low-risk alternative, such as government bonds. These are loans to a state, considered very safe, that yield a small, reliable interest rate. The hurdle is then set above this, because a riskier investment needs to yield more to be worthwhile.

There are two common variants when it comes to crossing the threshold. In the hard variant, only the profit above the threshold counts. In the soft variant, with a so-called catch-up, the management also retroactively receives a share of the profit below the threshold. The difference sounds technical, but it significantly changes the payout. Anyone reading a fund contract should therefore pay close attention to this wording.

From fund contracts to AI benchmarks

In business news, the term mostly appears in the context of private equity firms and hedge funds. It’s often reported that a fund missed its hurdle rate and therefore received no performance fee. Companies themselves also work with thresholds: an investment project is only approved if it exceeds an expected minimum rate of return.

In the AI industry, this principle shows up with benchmarks. These are standardized test suites used to compare models. Companies then proudly report that their new model has surpassed a certain score. Such announcements should be read with caution.

Because a well-known weakness of the principle is the temptation to optimize exactly for the threshold. A fund manager might take on higher risk shortly before year-end just to clear the hurdle. An AI team might heavily tune its model to the test tasks without it actually getting better in everyday use. A threshold, then, doesn’t measure quality—only what it measures. Anyone reading reports about crossed hurdles should therefore always ask which number is meant and who set it.