Despite the meteoric rise in big-data approaches, funders also need to recognize that some of the most pressing problems must instead rely on the intelligent use of small data sets.

Value-of-information analysis can evaluate whether big-data collection is worthwhile (R. Schlaifer and H. Raiffa Applied Statistical Decision Theory; Clinton, 1961). Collecting big monitoring data for threatened or invasive species, for instance, risks delaying decisions on protective measures. It might be better to fund direct action, as for killer whales in the Georgia basin (E. McDonald-Madden et al. Trends Ecol. Evol. 25, 547–550; 2010).

Urgent decisions may be necessary when information is sparse. In agricultural systems, a fast response to a new pest or disease outbreak can make the difference between eradication and decades of costly quarantine programmes. In ecology, population sizes and detectability are often too low to create big data sets. In health, defence and social sciences, collecting big data risks violating human ethics.

Where data are limited, scientific solutions underpinned by strategies such as adaptive management can optimize decision making (I. Chadès et al. Theor. Ecol. http://doi.org/br9s; 2016). Artificial intelligence, for example, can inform adaptation strategies for sea-level rise to protect migratory shorebirds worldwide (S. Nicol et al. Proc. R. Soc. B 282, 20142984; 2015).