Abstract How does one define reservoir characterization? The answer depends on who is asking and for what purpose. Are they looking for volume, shape, fluid type, flow rates, or behavior over time — or perhaps mechanical response to drilling or hydraulic fracturing, maximum economic return in a particular time period, or minimal environmental impact? What are the underlying reservoir properties that control these desired features, and, more fundamentally, what is the physics that underlies these properties, individually and in combination? Can multiple properties be predicted simultaneously yet distinguished from one another? At what resolution? Can these properties be validated and deterministically identified remotely, or do approximations and probabilities need to be introduced? Can we trust the validation information or does that also contain uncertainty? How do we harmonize scales, competing objectives, limited resources, and entrenched attitudes? Despite all these questions, the overriding purpose of any reservoir characterization is to enlighten, to inspire confidence, and to provide utility. It must be trusted enough to be consulted and then ultimately improve future decisions and outcomes. Furthermore, rather than a rigidly derived specific answer, a dynamic workflow is preferable — one that is adaptable, flexible, and customizable. With a clear understanding of these objectives, a systematic approach, and multidisciplinary input, diverse data can be distilled into useful representations of key properties with maximum credibility. Achieving this end will require human ingenuity, machine assistance, science, and art.