This paper proposes a numerical encoding method based on two-bit symbols selected from the set {00, 01, 10, 11}. Each symbol is assigned a specific weight—1, 3, 7, and 9 respectively—and the symbols are distributed across a graduated scale such as 10^-3, 10^-6, 10^-9, and beyond. A numerical value is therefore represented as the sum of weighted symbolic contributions across multiple levels. Unlike conventional binary encoding, Gray coding, or Huffman coding, this method emphasizes decomposition, reconstructability, and layered numerical structure rather than storage efficiency or error minimization. The system is simple, flexible, and can be converted into byte-oriented sequences for practical processing. Consequently, it offers a viable alternative for specialized numerical encoding and analysis tasks.