Q fever is a worldwide zoonotic disease, due to the bacterium Coxiella burnetii, responsible for reproductive disorders, such as abortion in domestic ruminants. Although direct detection of C. burnetii by quantitative PCR is primarily recommended for the direct diagnosis of Q fever in abortive contexts and identification of bacterial shedding, serological approaches aiming at detecting antibodies specific for C. burnetii are also useful at both the individual and herd levels. At the individual level, ELISA tests may be used to identify animals that were previously infected (with or without clinical signs and shedding) and may still be latently infected (carriers with or without shedding). At the herd level, ELISA tests may be used to reveal a past or recent exposure to C. burnetii within a considered farm, a key tool for a rapid screening (prevalence, current evolutive circulation). Recent investigations about the diagnostic accuracy of the three ELISA tests currently commercialized for their use in domestic ruminants showed that these tests are moderately sensitive (sensitivity values ranged between 40% and 94%) and that their specificities are inferior to 100% (specificity values ranged between 95% and 99%). As a consequence, the diagnostic uncertainty should be considered to limit potential misinterpretations of the individual or herd serological status. Objectives The objectives of this study were (1) to build a methodological framework allowing calculating predictive values of Q fever ELISA tests at both the individual and the herd levels and (2) to provide a user-friendly application that could be easily used by veterinarians to interpret the results of a serological sample plan applied to a ruminant herd. Materials and methods We developed an advanced computing method based on latent class modeling, implemented using JAGS and R to calculate predictive values corresponding to the results obtained with any of the three commercialized ELISA tests, at both the individual and the herd levels. This method was integrated within an open source web application, using Shiny, to favor its accessibility to all the potential users of these ELISA tests (e.g., veterinarians, veterinary diagnostic laboratories, research laboratories). After completing the characteristics of the herd (species, herdsize, type of production) and the Q fever epidemiological context (if known), the users obtain the probability of true seropositivity of the tested herd and animals given the ELISA test results. The operation of this application is illustrated on a true clinical case where a Charolais bull originating from an apparently ‘free of infection’ herd was tested positive for Q fever by one ELISA test at introduction. To confirm or infirm these test results, additional animals in the herd of origin were tested with the same ELISA, considering firstly five heifers raised with the bull and secondly all males and females more than two-year-old (N=149). Probabilities of true seropositivity of the bull and its originating herd were calculated thanks to the developed application. Results The five other animals initially tested animals that were raised with the seropositive bull were negative to the ELISA test, which corresponds, in regard to the test used and to the herd characteristics to a probability of the bull’s true seropositivity estimated at 0.85 with a 95% credibility interval (CI) of [0.14; 0.98]. Among the animals older than two years present in the herd, 5 out of the 149 tested animals tested positive which corresponds for each positive animal to a probability of true seropositivity estimated at 0 with a 95% CI of [0 ; 0.65]. At the herd level, the probability that the proportion of seropositive animals was above 0 was estimated at 0 with a 95% CI of [0 ; 0.58]; and if the herd was truly seropositive, the proportion of truly seropositive animals in the herd was assessed to 0.06 with a 95% CI of [0.02; 0.12]. Conclusion This application assists veterinarians in a proper interpretation of the results of Q fever ELISA tests according to the sampling size and to available epidemiological and herd information. Veterinarians can therefore easily take advantage of a complex statistical model in a Bayesian framework to support their daily work related to Q fever risk analysis in ruminants. In the absence of a perfect reference test, this application could also be useful for local and reference laboratories for the confirmatory diagnosis of an ELISA test result.