Forecasting the strength of the sunspot cycle is highly important for many space weather applications. Our previous studies have shown the importance of sunspot number variability in the declining phase of the current 11-year sunspot cycle to predict the strength of the next cycle when the minimum of the current cycle has been observed. In this study we continue this approach and show that we can remove the limitation of having to know the minimum epoch of the current cycle, and that we can already provide a forecast of the following cycle strength in the early stage of the declining phase of the current cycle. We introduce a method to reliably calculate sunspot number second differences (SNSD) in order to quantify the short-term variations of sunspot activity. We demonstrate a steady relationship between the SNSD dynamics in the early stage of the declining phase of a given cycle and the strength of the following sunspot cycle. This finding may bear physical implications on the underlying dynamo at work. From this relation, a relevant indicator is constructed that distinguishes whether the next cycle will be stronger or weaker compared to the current one. We demonstrate that within 24–31 months after reaching the maximum of the cycle, it can be decided with high probability (0.96) whether the next cycle will be weaker or stronger. We predict that sunspot cycle 25 will be weaker than the current cycle 24.