The primary differentiating characteristic is the question of where and how the writer puts together his texts and formats them.

Advantage: Texts and meta-information can be created and managed together. There are no additional license costs for an external editor. Only one software package has to be installed and updated.

Disadvantage: The integrated editors often don’t provide the same comfort and efficiency as the major text processors. If you want to use macros or auto text, for example, you’ll often have to use a third-party utility.

Other tools follow the contrary approach and use an external editor for creating the texts – mostly Microsoft Word, but also, to an increasing extent, FrameMaker, HTML editors, or XML editors.

Advantage: You can use all functions of your favorite editor.

Disadvantage: Often, operation is inefficient. Because the source data is mostly loaded with additional meta-information (for example, in the form of hidden texts or with the help of field functions), the translation process becomes particularly error prone. Add to this the typical, well-known problems of the various editors. For instance, those who are reluctant to work with Microsoft Word will never be happy with a tool that integrates Microsoft Word.

In this group, text creation and help generation are largely decoupled. Complete user manuals (mostly written with Microsoft Word or FrameMaker) are converted into online help only in the second step. The basis for formatting is usually a specific template, which must be adhered to strictly while writing the texts. Moreover, similar to the tools that use an external editor, often some additional meta-information has to be incorporated in the text.

Advantage: This procedure allows for efficient single source publishing especially if the material already exists in the form of printable user manuals.

Disadvantage: The book metaphor can mislead the author easily into writing contents that’s not as modular as required for online help. The results often are mere books on screen instead of media-adapted online help. Another disadvantage is the fact that the documentation for different product versions can hardly be generated from the same source.

Other converters are based on XML-DITA. This avoids the disadvantages listed above, as it forces the author to develop a modular, online-ready structure right from the beginning.

If online help needs to be Unicode-compatible for certain languages, this restricts the choice of tools. Not all programs support Unicode. There are also significant differences in the translation workflow. Many tools completely ignore translation requirements. Others provide sophisticated export and import options for texts that are to be sent out to external translators, or allow the author to edit several languages in parallel.

Only few of the classical help authoring tools explicitly support the simultaneous editing by several authors. In some cases, this can also be achieved manually to a certain extent at the file level, but this only works for the tools that save each topic into an individual file. If your team is working at different locations, a remote access might be required.

Other considerations are:

Does the tool have a built-in version control system or can it be connected to an external VCS?

Does the tool provide any support for the review process?

Does the tool provide some possibility for subject matter experts to contribute information in an efficient and easy way?

The majority of tools claim to support single source publishing, which means creating texts once, and then publishing to different formats and media. As long as these are just different help formats, such as compiled help or browser-based help, this usually works smoothly. However, when it comes to also generating printable user manuals or PDFs, the following aspects are what separates the grain from the chaff:

If documentation not only for different media but also for different product versions has to be generated from the same source (such as for a standard-version and a pro-version, for example ) single source publishing acquires one more dimension. Some tools allow you to define so-called conditional texts or build flags to hide specific sections of the text as required. Other important features are variables and text snippets (sometimes also called embedded topics), which allow you to use individual text sections many times over such as product names or a warning, for example.

Can the production of documentation for multiple product versions be automated? Is it possible to invoke the tool from the command line or from scripts? Can the tool be invoked along with all the settings and parameters that are required?

Can the tool be integrated into an overall build process?

Some tools have specialized on code, assembly and component documentation. They scan the source code and automatically generate a completely linked basic framework for the documentation. Citations from the source code are formatted automatically (syntax highlighting feature).

Using a low-cost tool can prove to be a costly exercise when a project grows lager than initially expected, or when it has to be revised later on. Most low-cost tools don’t support definable paragraph and character styles and format each word manually. When a global change to the layout becomes necessary, then comes the rude awakening.

Only for those who only need to create a small online help occasionally and don’t have to maintain and update this help file regularly, a low-cost solution might be an economical alternative.

Although all current help formats are backed by the source format HTML, only a few tools give you full control over the generated HTML, JavaScript and CSS code. Here you’ll have to make a fundamental decision: Do you want to be able to tune the code manually, or are you rather looking for a tool that spares you from most technical details?

Finally, there are some questions regarding performance, efficiency, and the scope of functions:

Apart from the set of features of the tool, and its performance and usability, you should also take into account the following worst-case scenarios: