Introduction: Vapour sensing plays an important role in various safety, security, and healthcare applications. However, it is perhaps surprising that despite modern technologies sniffer dogs are still considered the gold standard in vapour sensing. Achieving high sensitivities often remains challenging. Furthermore, existing technological solutions often require relatively large pieces of equipment, or pre-concentration steps [1].Conductive polymer (CP) based chemiresistive vapour sensors offer a promising solution because of the large range of materials available and their relatively cheap and easy processing methods. Furthermore, their small scale makes it possible to use them in small scale devices. CP based sensors described in literature often use a CP thin film as the sensing layer. Here we demonstrate a significantly improved sensitivity by using a percolation network of CPs instead of a CP thin film. Concept: Electrical percolation describes the likelihood of a current running from an electrode on one side of a network to an electrode on the other side. If there are no electrical connections between the electrodes the conductance is zero. When enough connections are added to the network to form a conductive pathway between the electrodes the conductance increases sharply. When more connections are added to the network the conductance keeps increasing and eventually reaches a plateau. Traditional chemiresistive vapour sensors, consisting of a CP thin film, are operated in this flat part of the percolation curve. We’ve shown that by creating sensors in the steep part of the curve, close to the percolation threshold, a significantly higher sensitivity can be achieved. Experimental: Our sensors consist of a polypyrrole (PPy) network between interdigitated Pt electrodes on a glass substrate. Au nano-islands were used as nodes in the PPy network. Sensors with different polymer coverages, corresponding to different points along the percolation curve, were exposed to various concentrations of ammonia. Results and conclusion: Sensors with polymer coverages corresponding to the steep part of the percolation curve give a significantly higher sensitivity than those in the thin film regime. One would perhaps also expect that the closer to the percolation threshold the better the sensor, however in practice this doesn’t hold true. Moving closer to the percolation threshold from this optimum does result in a higher absolute sensor response, however due to an increase in relative noise the sensitivity decreases when a sensor is too close to the percolation threshold. At this optimum, in the steep part of the percolation curve, we have achieved sensitivities as low as 9 ±2 ppb. Outlook: Further avenues of research that we are currently exploring include increasing the range of sensing materials and analyte vapours, as well as using this sensor technology for specific applications such as the detection of improvised explosives. We are also using percolation networks to improve the sensitivity of sensors on flexible substrates, and for the inclusion of this technology into wearable devices. Finally, we have developed an integrated device that not only allows us to take this technology out of the lab and test it in real-life scenarios, but also allows us to use multiple sensors at the same time in an electronic nose type setup. This allows us to increase our selectivity in addition to the increased sensitivity achieved by using percolation networks of CPs instead of thin films.[1] Lefferts, M. J. & Castell, M. R. Vapour sensing of explosive materials, Analytical Methods 7, 9005-9017 (2015); DOI: 10.1039/C5AY02262B