Past environmental information is typically inferred from proxy data contained in accretionary sediments. The validity of proxy data and analysis workflows are usually assumed implicitly, with systematic tests and uncertainty estimates restricted to modern analogue studies or reduced-complexity case studies. However, a more generic and consistent approach to exploring the validity and variability of proxy functions would be to translate a sediment section into a model scenario: a “virtual twin”. Here, we introduce a conceptual framework and numerical tool set that allows the definition and analysis of synthetic sediment sections. The R package sandbox describes arbitrary stratigraphically consistent deposits by depth-dependent rules and grain-specific parameters, allowing full scalability and flexibility. Virtual samples can be taken, resulting in discrete grain mixtures with defined parameters. These samples can be virtually prepared and analysed, for example to test hypotheses. We illustrate the concept of sandbox, explain how a sediment section can be mapped into the model and explore geochronological research questions related to the effects of sample geometry and grain-size-specific age inheritance. We summarise further application scenarios of the model framework, relevant for but not restricted to the broader geochronological community.