Soil microbial ecology needs robust tools to elucidate ecological questions, such as the impact of fertilisation on soil microbial communities. However, the methods and data analysis used can directly affect the biological conclusions. In this study, the sensitivity of terminal-restriction fragment length polyphorism (T-RFLP) to four restriction enzymes (RE), six peak area thresholds (PAT) from 0 to 10 % and two matrices (presence/absence and relative abundance) was assessed on soils subjected to eight different long-term amendments. The T-RFLP profiles were analysed using a three-step multivariate analysis approach: (i) cluster analysis and non-metric multi-dimensional scaling, (ii) ANOSIM and PERMANOVA and (iii) correlations. The application of organic and mineral fertilisers over 53 years changed the bacterial community composition regardless if the RE, PAT and matrix were used. However, the clustering of the community, the strength of these differences, the correlations with environmental variables and, subsequently, the biological conclusions varied with the use of RE, PAT and matrix. Hence, the bacterial community composition was found to be either highly sensitive to any changes in soil organic matter strongly correlated to C and N concentration, or only affected by large inputs of C or soil management. Different REs can reveal different bacterial populations affected by different drivers, but PATs 0.5 and 1 % should be used especially when using presence/absence matrix. This study also shows the complexity of the effect of organic and mineral amendment on bacterial community composition and stresses the importance to inform on methodological and data analysis parameters.