COVID-19 has presented unprecedented challenges for epidemiologists, as well as health care, the economy and society. Keeping up with the dizzying pace of change in the epidemic and producing consistently policy-relevant modelling predictions has been more challenging than ever before. In this context, a range of models have been used, and although they differ considerably in their construction, many of the messages have been consistent and accurate. Producing precise predictions of the epidemic trajectory has been more challenging for several reasons, including that the data available is surprisingly sparse. Months into the pandemic and with millions of cases worldwide, much of our understanding is still derived from small groups of well-described patients.