Purpose:　Collaborative　research　and　development　have　remained　a　pertinent　mechanism　for　conducting　technological　innovations.　With　the　lens　of　knowledge-based　view　(KBV),　this　study　aims　to　examine　the　role　of　changes　in　knowledge　couplings　and　network　cohesion　to　elevate　innovation　performance.　Design/methodology/approach:　Data　analysis　has　been　performed　on　53,459　patents　through　regression　analysis　with　random　effects.　These　independent　and　joint　patents　are　extracted　from　Derwent　Innovation　Database.　Findings:　Findings　explicate　that　change　in　external　existing　or　existing　and　new　knowledge　couplings　have　inverted　U-shaped　effects　on　a　firm’s　innovation　performance.　Changes　in　internal　existing　or　existing　and　new　knowledge　couplings　have　direct　positive　effects　on　firm’s　innovation　performance.　The　moderation　effect　of　network　cohesion　flattens　the　inverted　U-shaped　effect　of　external　new　and　existing　knowledge　coupling,　whereas　it　has　no　significant　effect　on　external　existing　knowledge　coupling.　Network　cohesion　further　elevates　the　effects　of　internal　knowledge　couplings　–　existing　or　existing　and　new.　Research　limitations/implications:　This　study　theoretically　contributes　to　KBV　and　innovation　management　literature　by　highlighting　the　scope　of　changes　in　internal　and　external　knowledge　couplings　and　subsequent　output.　Network　cohesion　flattens　the　curviness　of　changes　in　external　new　and　existing　knowledge　couplings,　which　is　a　contribution　to　strategic　management　literature.　Practical　implications:　Organizations　need　to　carefully　manage　changes　in　knowledge　couplings　and　ensure　their　benefits　(obtain　new　knowledge　domain　or　new　combination)　outweigh　liabilities　(damages　to　organizational　routines　or　increase　in　collaboration　costs).　Managers　must　consider　four　kinds　of　knowledge　coupling　changes　along　with　developing　network　cohesion　as　an　R&D　strategy.　Originality/value:　This　study　is　one　of　its　types　to　flatten　the　curve　through　network　cohesion.　This　study　divided　the　changes　in　knowledge　coupling　into　four　types　and　two　dimensions;　external　existing　and　new　and　existing　knowledge　couplings　and　internal　existing　and　new　and　existing　knowledge　couplings.　©　2021,　Emerald　Publishing　Limited.