As in any SIP experiment, the experimental design should be carefully considered in advance. Particular attention should be given to the length of incubation and the number and type of controls. Incubation of the sample in the presence of the 15N-labelled substrate should be prolonged enough to ensure that the DNA or RNA are sufficiently labelled above the detection limit. The detection limit, in turn, is itself not a fixed value but will depend on the number of fractions being collected from each gradient, the relative abundance of the particular labelled taxa and on which method is being used to analyse the data (see Section \ref{316470} for more details). In general, it is assumed that DNA or RNA molecules should be labelled to at least 30 atom% to differentiate them from unlabeled molecules in a density gradient \cite{Buckley_2007a,Cadisch_2005}. In contrast, long incubation times bear the risk of labelling community members that do not perform the metabolic activity in question but were labelled through cross-feeding. The issue of cross-feeding is of general concern in SIP experiments and has been mostly discussed for 13C-based SIP experiments (e.g. \citealt{McDonald_2005,DeRito_2005}), but diazotrophs have also been shown to release substantial amounts of fixed nitrogen through cross-feeding or leaching \cite{Belnap_2001,Adam_2015}. Since cross-feeding in a microbial community cannot simply be put to a halt, the typical way of dealing with this issue is to sample at several time points, limit the incubation time to the minimum necessary for labelling and combine complementary lines of evidence when concluding that a specific taxa indeed performs the metabolism in question. Diazotrophy is a slow and costly process, and, incubation times are accordingly relatively long compared to incubations with a 13C-labelled substrate. Consequently, 15N-SIP incubations targeting diazotrophs would require incubating the samples for several days or even weeks, depending on the specific level of activity of the system \cite{Angel_2017,Buckley_2007,Pepe_Ranney_2015}. However, for targeting the assimilation of biologically available N-forms such as ammonium, urea or amino-acids incubation times should be reduced to several hours to few days, since the process is much more rapid and requires only little energy from the cells \cite{Alonso_Pernas_2017,Bell_2011}. The number of fractions collected can also affect the detection limit. While a higher number of fractions will most likely increase the sensitivity, it also entails higher sample processing efforts and costs. In addition, more fractions also mean less template per fraction and thus also an increased difficulty to amplify the target and a higher chance of contamination. Typically 12--20 fractions are collected, of which about 10--16 end up being analysed. As in any SIP-experiment, appropriate no-label controls should be set up in parallel to avoid detecting false-positives. Particularly with the growing use of high-throughput sequencing and statistical models to detect labelled OTUs the need to include more no-label controls in the experiment in order to correctly detect labelled phylotypes has been growing. The exact number of no-label controls will depend on the exact statistical method used to analyse the data, but also on the type of SIP being performed since DNA-SIP is more prone to detecting false positives than RNA-SIP because of the effect of the G+C-content bias (see \ref{316470} for more details). Ideally, every labelled sample will have its parallel no-label control.