The issue being, that operators seemingly (in my opinion), make decisions based on a small proportion of the population and assume direct proportion applies to the return.

I know, you know and most other people know that you won't get the same answers from two sets of 5 people, two sets of 50 or two sets of 5000.

Yet as passengers, our services change, because 5% (as in the case of the 1/24) of the population think it should.

Three or four methods have been used and after a period, the method is adapted or changed - with what looks like, not many more responses than the previous method.

Rather than use one method and hope for the best, incurring the associated costs - to then repeat this as the changes haven't worked out, go for several methods in one go.
The initial costs may be higher, but surely the increased sample will lead to changes that are suited to an increased population, rather than assume direct proportion will apply - realise it doesn't and go back to the beginning.

When Nexus did their initial bus strategy consultation in 2009, they went to community centres, booked space in libraries, set up websites and managed to engage a quite staggering amount of people.
They listened, acknowledged and took on board what people wanted.

edit:
An excellent read.
It talks about sample sizes, how bigger samples give better results, direct proportion and much more.
http://home.ubalt.edu/ntsbarsh/business-...urveys.htm
If operators are insistent on going down the route of consultations, maybe they will take on board the comments of this academic - leaving passengers with a network that works and isn't constantly being meddled with.

I reckon circa 600 people need to be spoken to (using a 4% margin error and the reported population of Sunderland, 275,000). In the Stockton example, 83,000 residents equal a similar sample size.

Is it possible to get 600 people spoken to in an 8 hour day at Park Lane or 2 hours in Stockton?
http://www.surveysystem.com/sscalc.htm#one

Change the margin error to 1%, all of a sudden the sample size needs to be in the thousands!