Background There are racial, ethnic and geographical differences in complete blood count (CBC) reference intervals (RIs) and therefore it is necessary to establish RIs that are population specific. Several studies have been carried out in Africa to derive CBC RIs but many were not conducted with the rigor recommended for RI studies hence limiting the adoption and gener-alizability of the results. Method By use of a Beckman Coulter ACT 5 DIFF CP analyser, we measured CBC parameters in samples collected from 528 healthy black African volunteers in a largely urban population. The latent abnormal values exclusion (LAVE) method was used for secondary exclusion of individuals who may have had sub-clinical diseases. The RIs were derived by both parametric and non-parametric methods with and without LAVE for comparative purposes. Results Haemoglobin (Hb) levels were lower while platelet counts were higher in females across the 4 age stratifications. The lower limits for Hb and red blood cell parameters significantly increased after applying the LAVE method which eliminated individuals with latent anemia and inflammation. We adopted RIs by parametric method because 90% confidence intervals of the RI limits were invariably narrower than those by the non-parametric method. The male and female RIs for Hb after applying the LAVE method were 14.5–18.7 g/dL and 12.0–16.5 g/dL respectively while the platelet count RIs were 133–356 and 152–443 x103 per μL respectively. Conclusion Consistent with other studies from Sub-Saharan Africa, Hb and neutrophil counts were lower than Caucasian values. Our finding of higher Hb and lower eosinophil counts compared to other studies conducted in rural Kenya most likely reflects the strict recruitment criteria and healthier reference population after secondary exclusion of individuals with possible sub-clinical diseases.