Conclusions are what really separate the rhetorical argument from the research report. In a research report we might read the literature and, well, report on it. But in an argument we extend what we knew before and create new knowledge. We use the conclusion of an argument to answer questions like these:

Validity	Does it do what it's supposed to?
Repeatability	Does it do it consistently?
Reliability	Can we trust this result?
Scope	Is this result consistent with our intent?
Range	Did we consider all reasonable possibilities?
Predictions	What's going to happen next?
Judgment	What value does this result have?
Implications	How does this affect our shareholders/customers?
Concerns	Do we see something we need to improve?
Recommendations	What must we then do?

We can also summarize here. But this is not the place to introduce new subjects apart from the above, nor is this the place for cliches.

Dangerous Conclusions

The ill-fated Space Shuttle Challenger launch was based in part on self-contradictory conclusions:

The lack of a good secondary seal in the field joint is most critical...

Analysis of existing data indicates that it is safe to continue flying existing design.

The analysis in this case was based on computer simulations only. (Is that enough to make such a judgment?)

Demonstrations for the Rogers Committee investigating the Challenger disaster were performed by managers. Their argument: "We think we do a good job, and we wanted to show what we do." (Is that the time to have the managers leading?)

And clearly reports leading up to the Challenger launch were required to generate at least one new piece of knowledge: should the next launch take place on that given day, in that given temperature? A lesson we might learn is that conclusions must be consistent with what we've observed, or we might be better off not making them.

The probability of mission success is necessarily very close to 1.0.

Does that mean it IS? Or SHOULD BE? Within NASA, a range safety officer told Feynman it was 0.999 (one Shuttle failure expected in 1000 launches); the official NASA position was 0.99999 (one failure in 100000 launches). If we take a position like this in advance, and the way we interpret it varies from person to person, then we can easily bias our conclusions to match -- even if we have to ignore some of the actual data to do it. Sobering, right?

References

Leighton, R. and R. Feynman. What Do You Care What Other People Think?. Norton & Company, 2001. ISBN 0-393-32092-8
Vaughan, D. The Challenger Launch Decision. University of Chicago Press, 1997. ISBN 0-226-85176-1