Accessibility has been a bit of a hot topic for me recently. I was at a netimperative.com networking dinner recently, where the speaker was Julie Howell (currently Fortune Cookie, ex RNIB). Her take on accessibility was particularly interesting - do it, because it makes sense, and not because it's "worthy". Making access to your site difficult to people with disabilities, either because you're unaware, lazy or dismissive, doesn't make good business sense.
For a company like ours, which is nearly all pure B2B (you can't pop along one day and buy a couple of tonnes of bauxite from Rio Tinto with your credit card) accessibility is a tricky one. For instance, we have tonnes of legacy content from previous iterations of our website going back 10 years or so, including thousands of speeches, presentations and investor documents that are in an "inaccessibile" PDF format. For us, we need to make a call on how many of these documents we republish, documents that are rarely (if ever) viewed - would you want to see a press release dating from 1998 about production reviews at a mine we no longer own, for instance?
This is further compounded by (ok, at this stage fairly annecdotal) evidence that we have fewer disabled viewers than many other sites. Our stakeholder groups that the site is aimed at (investors, journos, job seekers, governments etc) had lower presentation of disabled people than the general public. This doesn't mean that we ignore them, or dismiss them (far from it), however it does lead to difficult decisions - how much do we engineer areas of the site and our content (eg re-editing videos and adding captioning at huge expense) for a very small audience segment. We took the view that we should do everything that people would reasonably have expected us to do (eg making the site at least Priority A, and AAA where we could), but where the effort and cost was prohibitive, and where the content was of minimal or minority interest, to leave the legacy content as it was. We're reviewing all of our old content on a regular basis, and if we start to see there is a demand, then obviously we'll retro-fit this to current accessibility standards.