Introduction: Effective adjustment to changes in metabolic demand is an important determinant of exercise tolerance in health and disease (Boone et al., 2016). This relies, in part, on the matching of skeletal muscle oxygen delivery (Q) and oxygen utilisation (VO2). The profile of the ΔHHb (deoxygenated haemoglobin + myoglobin) signal derived from near infrared spectroscopy (NIRS) can provide insights into this VO2-Q relationship (Grassi et al., 2003). Resolving the issue of accurately and fully quantifying the ΔHHb response to ramp exercise has mechanistic implications for understanding control of O2 delivery and utilisation in heart failure (HF). However, the dynamics of the ΔHHb profile during ramp incremental exercise in HF are yet to be quantified. Methods: Ten HF patients (NYHA II; EF 26 ± 9 %) performed a ramp incremental exercise test to voluntary exhaustion. The ΔHHb response was plotted as a function of absolute power output (POABS) or normalised power output (%PO) and fitted with either a sigmoidal (Ferreira et al., 2007) or double-linear (Spencer et al., 2012) model; the corrected Akaike Information Criterion (AICC) was used to evaluate model fits. Data from a representative subject was used to simulate the arteriovenous O2 difference from ΔHHb data, as in Ferreira et al. (2007). Results: VO2peak was (mean ± SD) 18 ± 5 mL/min/kg. ΔHHb increased from -3.8 to 3.9 (a.u.) during incremental exercise. The P50 (half magnitude) deoxygenation using a sigmoid fit occurred at 71.5 ± 22 W (POABS) and 59.9 ± 13 % (%PO). For double-linear fits, the breakpoint occurred at 69 ± 35 W (POABS), and 56 ± 19 % (%PO). In 6 out of 10 patients slope 1 was steeper than slope 2, indicating a progressive slowing of deoxygenation. In the others, deoxygenation accelerated, indicating a progressive shortfall in muscle O2 delivery during incremental exercise. Overall the double-linear model quantified ΔHHb data more accurately than the sigmoidal model, with AICC scores of 1592 ± 647 vs. 2354 ± 755 (POABS; P < 0.05) and 1578 ± 637 vs. 2354 ± 755 (%PO; P < 0.05) respectively. Simulations of the arteriovenous O2 difference show a rapid O2 delivery/utilisation mismatch at the onset of ramp exercise in some patients; because the linear model removes initial data, the sigmoid fit may give better mechanistic understanding of O2 delivery/utilisation mismatching at exercise onset. Conclusions: ΔHHb responses to ramp incremental exercise in HF are variable between patients, and a ‘one-fits-all’ modelling approach is not suitable for characterising the response. Muscle deoxygenation appears to show evidence of a progressive limitation in muscle O2 delivery in some patients and not others. As such, data should be interpreted on a patient-by-patient basis.