One of the hot topics in constructionist approaches is which traits are to be considered essential to posit a construction, as can be attested in the wide range of definitions that have been put forward when stating what a construction looks like. Construction grammarians such as (Kay 2013) draw a clear-cut distinction between constructions proper and patterns of coining, the latter lacking full productivity. Against this background, I highlight that the frontier between proper constructions and patterns of coining is rather blurred, and speakers cannot know without experimentation whether each string is learnt individually and whether speakers are likely or not to build new expressions analogized to these patterns. Using a corpus-based methodology, I examine the construction [Dime X y te diré quién eres], a semi-schematic pattern emerging from a proverb intrinsically related to the notion of “Modelle der Analyse” (Fleischer 1997), “E-creativity” (Sampson 2016) or “snowclone” (Hartmann & Ungerer 2024). With 1,036 occurrences extracted from the esTenTen18 corpus, I scrutinize the productivity of the construction by means of the role of high token frequency in the postulation of analogical extensions (productivity as extensibility) and of semantic connectedness in the distribution/clustering of the types (productivity as regularity).