You should read Stanley Lieberson's paper in Social Forces about Small N studies.  It was published around 1995 or so.  It addresses some of the issue you raised. 

Validity & reliability tests are desirable, more particularly so in most of the qualitative studies. Instruments for measurement of any variable (observation) is expected first, both to be valid as well as reliable.

Trust, enough literature can be found on this (particularly from any standard 'Research Methodology' book).

I agree with Rabinarayan. You will have to test reliability and validity as an essential part of your research. 

Thanks a lot for you all!

I am not sure yet though!

 Many researchers are sceptical about the construction of such beliefs and

criteria and the hegemonic influences arising from such notions (Johnson,

Buehring et al. 2006). In terms of internal validity, questions are often

raised about how do remote actors establish truths within another's

research and also claim that many truths often exist, but are often neglected

(Johnson, Buehring et al. 2006). With respect to external validity, many

difficulties arise from the identical transferability of criteria from one

setting to another, this is due to a range of un-replicable, often temporal

factors (Johnson, Buehring et al. 2006).

Thanks Saleh.

Due to the potentiality of the application of the research output - validity is required and similarity the reliability.

In questionnaire survey, measurements are not from machine and thus have human biases, therefore, it is essential.

If you have proper justification and ensured these during your data collection processes even qualitatively, there is need to report it during your writings.

Thanks.

Saleh, I think they are important in all cases when "you want to measure what your study intends to measure" and when you want your "data to be consistent any time you measure them again".

It sounds like you are doing a mixed methods study with a design that amounts to: QUAL --> quant. They basic issue is thus what you want to accomplish through the additional quantitative work that you are doing.

For example, you might want to do a preliminary test of a hypothesis you generated through the qualitative data, so you would need to operationalize the relevant variables. For what amounts to a pilot study, you would probably have a small sample and just a few variables, so you would essentially be relying on face validity rather than more technical approaches.

The key point here is that a small, supplementary "quant" study may not need the complete apparatus of reliability and validity that would go into a piece of full-scale quantitative research. Of course, not having proven reliability and validity would limit would you could say, but you would only be trying to make a limited statement in this case.

Thank you all for your advice, I am really happy to read all your comments. This shows me how academics have the commitment to the community. 

in fact, I found that there are two dimensions of the phenomenon that I am studying "Managers response strategies to negative customer reviews". Based on these two dimensions, I developed four main response strategies and ended up with a "lovely" taxonomy.

Now, some managers follow one single strategy, and therefore can be assigned to one of these main strategies. However, others follow different strategies in different occasions.

The problem is that my supervisor suggested that I use something like "perceptual map" to show where each manager fits (my sample is up to 100 manager). Therefore, I developed a "Behaviour Rating Scale" to do that. I ended up with pretty diagram that shows the idea that there are some variance within each group and some managers fit somewhere in between these groups. However, I have to admit, it is not reliable, not valid. I used single-item measures to measure these dimensions and found that they need multi-items really.

I found that I have to use more complex techniques to measure these two dimensions. 

Do you have any advice please?