I am ambivalent as to the need for indexes. Don't get me wrong, indexes are a great labor saving device. But the very nature of indexes discourages in-depth searches into the original records that likely contain just the information you are looking for even though it does not appear in the index. Statements about indexes have the tendency to portray them as the ultimate solution to research problems. It is true, that much of the easy research is aided by indexes. It is equally as true that indexes sometimes prevent the very research they are supposed to assist.
The tragedy in genealogical research occurs when the neophyte researcher relies entirely upon the index and assumes that the records is missing rather than the fact that the index is simply defective. It is a fact of life that when an index is off even one letter the resultant entry can be completely unconnected to the original. When I say this, think what happens if the initial letter is misread. One example from my own genealogy involved the indexing of "Tanner" as "Danner." An index-only search in that database fails to return even one "Tanner" entry.
When you begin your research you may think that your ancestors have simple surnames but your quickly disabused of this belief by the variety of ways the original names were recorded in the source records and even more by the variety of index entries that have misread the originals. There is always a point in any researcher's experience with genealogy that spelling variations in names becomes an issue and indexes do not handle name variations well. If your ancestors spelled their name "Smith" as "Smythe" on occasion, then a very good index would move that spelling completely out of the "Smith" returns to a place where you might not think to look.
I could write a very, very long post about all the incidents of spelling variations and all the people who got hung up on the "way my name is supposed to be spelled" but the idea here is that all these variations show up in indexes by moving the entries from one results list to another. Some of the changes come from the way the individual spelled his or her name in the first place. You have a very hard life as a genealogist if you don't make allowances for spelling variations at all levels of record keeping. But all of these variations make indexes less than perfect. How many "brick wall" situations are simple spelling issues? I would guess quite a few.
The idea here is not that indexes per se are bad, but that relying on them exclusively is foolhardy. The classic example is the ancestor who seems to disappear from the U.S. Census in certain census years. The vast majority of these disappearances turn out to be spelling based errors occurring somewhere in the process of making an index. As some of my friends would say about now, "Well, James why don't you tell us what you really think about indexes?" No, I am not advocating a wholesale abandonment of the indexing industry, I just got fed up with bad entries lately and needed some space to think about the subject.
Back to the issue at hand. Some people seem to think that indexes are proper in all instances, that is, every single source record should appear in an index. I disagree with that broad brush approach to indexing. Most of the indexing efforts pick and choose which parts of the original records are included in the indexing and sorting process. Valuable information in original records can be and often is, completely obscured by failing to examine the original and relying on an index. There is often an excessive reliance on surname indexing that is inappropriate with certain naming conventions. For example, Spanish language surnames may contain two or more "surnames" such as "Maria Gonzalez y Rodrigues." So is this indexed under "G" or "R" or both? What if one of my ancestors used the name "Tanner-Roberts" or some variation with a hyphen, where does the name fall in an index? Is it under "Tanner" or "Roberts" or both? It is my experience that most indexing programs or procedures would not include the name twice.
I am always amazed at how many people, when using Ancestry.com, stop at the page where the summary of the record is given and fail to examine the image of the original record closely. How much of the original is in the summary? Yes, of course, the original record could also be a mess, but how do you know that if you rely exclusively on indexes? Some original records should not be indexed in my opinion. Indexing is a waste of time and misleading. For example, Spanish and English language parish registers. I have yet to see an index that includes an index entry for the witnesses to baptisms and marriages. They might exist, but most of the indexes are doing good to index the main name of the people baptized or married. Leave this type of record alone and let the researchers search through the entire record themselves. How many children and others are missed in families simply because the index is faulty?
My conclusion is don't believe everything you see and read. Take time to think about alternatives and don't get hung up on spelling variations. Finally, don't rely too heavily on indexes, they lie.