Hamming Quasi-Cyclic (HQC), recently selected by NIST for standardization, does not employ ciphertext compression, unlike its lattice-based counterpart, Kyber. In lattice-based encryption, ciphertext compression is a standard post-processing step, typically implemented via coefficient-wise rounding. In contrast, analogous methods have not yet been explored in code-based cryptography. We address this gap by developing techniques to reduce ciphertext sizes in schemes defined over the Hamming metric, with a particular focus on HQC. To support this approach, the decryption failure rate (DFR) analysis is generalized. Specifically, we revisit the modeling of the error that must be correctable with probability \(1-2^{-\lambda }\) to achieve \(\lambda \) bits of security, which was previously tractable only under an independence assumption. We propose a more accurate model of the error distribution, which takes dependencies between the coefficients into account. Confirmed by extensive simulations, the proposed model sharpens the DFR analysis and thereby improves our understanding of the security of HQC. Building on this generalized framework, we present a ciphertext compression mechanism that enables a precise DFR analysis and is therefore transparent with respect to security. This is achieved by carefully designing a quantization code with a direct-product structure, aligned with HQC’s error-correcting code. For the parameters proposed in the round 4 submission, our techniques reduce HQC ciphertext sizes by up to 4.7%; a proof-of-concept implementation confirms that this improvement comes without noticeable loss in efficiency. Reductions of up to 10% are achievable via a trade-off with public-key size.