Biodiversity is declining fast, and with limited funding and resources in conservation, it matters that these resources go towards actions that genuinely make a difference. So how do we know which actions deliver results, and which don’t? The answer depends on the quality of evidence behind them – which is largely contingent on the study design used. New research from Staples et al (2026) in the Journal of Applied Ecology examined the study designs used to evaluate conservation actions to understand whether we are becoming better at testing what works, and in which contexts.
The importance of study design in conservation
When evaluating the effectiveness of a conservation intervention, the key challenge is estimating the counterfactual scenario: what would have happened if the action hadn’t been taken?
Different study designs answer this at different levels of confidence. At the weak end, After designs simply measure an outcome once the action has happened – no baseline or control (untreated unit) to compare it to. Before-after designs measure providing a temporal comparison but still lack a control group. Control-impact designs compare treated and untreated sites, but only after the fact. Before-After-Control-Impact (BACI) designs and randomised controlled trials (RCT) combine temporal and spatial controls, and reduce bias, providing more reliable estimates of what changes can be attributed to an action.
What we did
We analysed over 8,500 studies from the Conservation Evidence database – spanning over 50 years of studies testing conservation interventions. After classifying each study into one of the five designs outlined above, we used a Bayesian multinomial model to test how design use shifts over time, and how it varies with publication language, national income, biodiversity intactness, and subject area.
What we found
The encouraging news is that conservation is gradually shifting away from weaker study designs, towards more robust ones. After designs made up 74% of studies in 1970, which has gradually decreased to only 33% in 2023. Meanwhile, more robust designs have increased, with Control-Impact studies now being the most common, and RCTs have gone from being vanishingly rare in the 70s to now making up around 15% of studies.
The less encouraging news is that despite their promising growth, robust designs remain rare and unevenly distributed, revealing some key barriers to their use.
Geographically, robust designs remain concentrated in North America, Western Europe, and Australasia. Many of the world’s most biodiverse and environmentally degraded countries, particularly in Africa, had very few rigorous study designs, if any studies at all. Lower-income countries showed a similar pattern, with an underrepresentation of robust designs relative to wealthier nations. Studies published in non-English languages also had substantially higher proportions of weaker study designs, a gap that persisted even after accounting for other variables in the multinomial model.
Design choice was also strongly context-dependent. Studies on habitat and soil-based topics used randomised experiments far more often, while studies on animals leaned disproportionately on weaker designs. This makes sense: mobile, wide-ranging, or elusive species are much harder and more resource-intensive to manipulate experimentally, assign randomly to treatment groups, or track consistently over time than patches of soil or field margins.
What this means for practice
Our findings suggest that critical barriers to improving the evidence base remain. The persistent dominance of simple designs in non-English language studies and lower-income countries points to structural, resource-driven inequalities. It is clear that targeted capacity building and longer funding timelines are necessary to close this gap – particularly in the biodiverse, lower-income countries that have the fewest robust evaluations, and the most to lose from getting conservation decisions wrong.
We don’t think the answer is to demand gold standard RCTs everywhere. In practice, decision-makers frequently cannot wait for the perfect experiment before acting, especially when it means withholding help from an endangered population or not intervening in rapid invasive species spread. However, while designing robust studies may cause a short-term delay, they can save time by preventing the prolonged use of actions that turn out not to work, which can be particularly crucial when deciding on long-term investments, policies or large-scale projects.
The good news is that meaningful upgrades don’t require jumping straight to a randomised trial. Adding even a single matched control site, or collecting baseline data before an action begins can substantially improve what a study design can tell us. Imperfect action, backed by imperfect but improving evidence, is better than no action at all. The goal isn’t a rigid application of gold-standard designs, but matching the level of rigour to the consequences of being wrong, and building evidence into the adaptive management cycle – especially in the places currently left furthest behind.
Read the full article ‘Temporal and spatial trends in study designs used to understand what works in conservation’ in Journal of Applied Ecology.
