On Wednesday 19 December 2018 the European Centre for Medium-range Weather Forecasting (ECMWF) hosted a Royal Meteorological Society national meeting on ‘Progress in weather forecasting – the user's perspective’. The aim of the meeting was to share examples and experience of how weather forecasting has progressed over the last few decades, with particular emphasis on how this has impacted the people who use and communicate the forecasts, especially for warnings of severe weather. It soon became apparent that there has been a considerable shift towards probabilistic forecasting in recent years which is likely to continue. This has happened in conjunction with a much greater emphasis on getting the forecast message across and seeking a more multi-disciplinary approach, including more interaction with social scientists. The meeting began with Florence Rabier welcoming everyone to ECMWF and showing a fantastic picture of Hurricane Irma. Florence made the point that hurricane forecasts have vastly improved, not just because of better forecast models, but also from the use of satellite observations in initiating forecasts. This was demonstrated by showing forecasts of Hurricane Irma with and without currently available satellite information. The forecasts with the satellite data produced a good representation of the hurricane, whereas those without had only a very weak signal. The main body of talks started with Charlie Pilling from the Flood Forecasting Centre, who talked about the North Sea coastal floods associated with the great storm of 1953, which killed over 2500 people, and contrasted this with a similar event that occurred in 2013. Charlie very nicely demonstrated how far we've come in 60 years in terms of forecasting this sort of event. There were 71 flood warnings and 160 000 warning messages issued in 2013, compared to zero warnings in 1953. We now have the capability to run ensembles of atmospheric forecasts and coastal surge heights. The generation of warnings from a probability-based flood-risk matrix along with better communication and improved sea defences all help to keep people much safer today. Brian Golding continued the wind storm theme by looking back at the Great Storm of 16 October 1987 and some major storms since then. Brian pointed out that we are in much better shape now, especially with the advent of kilometre-scale models, ensemble forecasting and knowledge of the ‘sting-jet’ phenomenon. He praised the use of probabilities and the risk matrix approach but said that there was still much more work to be done on the communications side and, in collaboration with social scientists, to identify weaknesses in warning systems. Christel Prudhomme moved the discussion from high winds to flooding, and presented a talk on the European Flood Awareness System (EFAS). Flood events can sometimes span very large transnational geographical areas, with huge societal consequences. In the 1990s it was realised that useful forecast information did exist but it was difficult to access. The concept of building a full flood forecasting chain – from precipitation forecasts to hydrology to warnings and dissemination – was first tested in 2002, building on the existing European-scale Flood Forecasting System (EFFS), and since then it has grown into a pan-European consortium effort, going fully operational in 2012. Tim Hewson then shifted the emphasis from rain to snow and reminded us of when we were shivering through the ‘Beast from the East’ early in 2018. He described the larger-scale temperature evolution leading up to the event and showed how the transport of warmer air towards the pole led to a blocking pattern and cold easterly flow. The balance of probabilities tipped in favour of the very cold weather in the ECMWF ensemble around seven days ahead, and we can now routinely expect to predict these sorts of events many days ahead. Tim also discussed the influence of Sudden Stratospheric Warmings (SSW) on cold easterly outbreaks over the UK and showed that the chance of a cold easterly on a given day is only increased from 3.3% to 4.5% following an SSW. His talk finished by examining more localised and unusual weather, such as freezing rain, that still pose a very difficult forecasting problem. The final event of the meeting gave the audience the exciting opportunity to participate in a ‘live science session’ to identify probability thresholds on which people base their decisions. This was led by John Hammond, Sara Thornton and Mark Rodwell, and involved posing scenarios and asking whether people would act one way or another given a particular probability of a weather event. For example, the audience was asked to consider the following scenario: It is a Monday in summer, and you are thinking of going to the beach at the weekend with family, friends or on your own. The forecasts give you a probability that it will be warm and dry (greater than 20°C and with less than 0.5mm rain in 24h). Using Plickers cards, each member of the audience was asked to indicate whether they would go to the beach given different probabilities of the event. It was emphasised that there were no right or wrong answers. The results were almost instantly available, and it turned out that the critical probability (that changed the decision) was made at 60–80% (>60–80%, ‘I'll go’; <60–80%, ‘I won't’). It was shown statistically that it might have been better to use a probability threshold of 30% for this example, but that doesn't account for all of the important personal factors people used in their decision-making. The audience were invited to give their reasoning, in order to see how the variety of priorities people had, led to a range of different decisions. This is an important area of study if we want to make the best use of our probability forecasts for individuals and/or corporate stakeholders. It was a very enjoyable meeting, with informative talks, plenty of questions and discussion, and a lot of fun in the live session. The benefits of probabilistic forecasting came out over and over again, along with how to communicate and make the best use of probabilities for practical decision-making. The message is this: embrace probabilities! I would also like to mention the impressive movie-loops provided at the break from a variety of research centres and say thanks for the festive canapés reception hosted by ECMWF, along with carols from the Met Choir.