Q&A: why default algorithms do not beat commercial solver ?

docs/make.jl

@@ -27,7 +27,8 @@ makedocs(
             "Configuration" => joinpath("man", "config.md"),
             "Algorithms" => joinpath("man", "algorithm.md"),
             "Callbacks"   => joinpath("man", "callbacks.md")
-        ]
+        ],
+        "Q&A" => "qa.md"
     ]
 )
 

docs/src/qa.md

@@ -0,0 +1,22 @@
+# Question & Answer
+
+#### Default algorithms of Coluna do not beat the commercial solver I usually use. Is it normal ?
+
+Yes it is.
+
+Solvers such as Gurobi, Cplex ... are handy powerful black-box tools. 
+They can run a very efficient presolve step to simplify the formulation,
+automatically apply lots of valid inequalities (such as MIR or cover cuts), 
+choose good branching strategies, or also run heuristics.
+However, when your formulation reaches a certain size,
+commercial solvers may run for hours without finding anything.
+This is the point where you may want to decompose your formulation.
+
+Coluna is a framework not a solver.
+It provides algorithms to try column generation on your problem very easily.
+Then, you can devise your own branch-cut-and-price algorithm on top of Coluna's algorithms.
+to scale up and hopefully beats the commercial solver.
+
+To start customizing Coluna for your own problem, 
+you can [separate valid inequalities](../user/callbacks/#Separation-callbacks) 
+or [call your own algorithm that optimizes subproblems](../user/callbacks/#Pricing-callback).