This is an excellent publication from the National Education Policy Center.
"Notwithstanding the federal enthusiasm for test scores, many researchers have warned against using a single measurement of any kind as the primary basis for such important personnel decisions as teacher retention, dismissal or pay.... [P]olicymakers would do well to pause and carefully examine the issues that make teacher assessment so complex before implementing an assessment plan.
Steps toward that goal include that policymakers:
• Be clear about the purposes of any assessment before selecting strategies. Where formative and summative assessment are to be combined, plan to address the challenges of dual-purpose systems.
• Involve all key stakeholders in system design.
• Rather than employing a single assessment tool, gather evidence from multiple sources. Combine strategies so that the weakness of any single tool is offset by the strengths of another.
• Be sure that the criteria for assessing performance, artifacts or other factors are credible and are well understood by teachers and assessors.
• Provide high-quality, ongoing training for assessors and routinely calibrate their efforts to ensure consistent application of criteria.
• Look to high-quality research on existing tools and programs to inform the design of assessment systems.
• Commit sufficient resources to produce high-quality, productive assessment.