If I were to suggest another idea, it would be a interactive graphics chart (Flash or Java) of people's click throughs. You are two steps away from doing this today: (1) The graphics part (effectively moving it from "HTML list structure" to a 'flow' in Flash) and (2) the move to a graphic UI would allow you to use symbols and labels to present additional info (again, interactively) about the pages the user selected (e.g. percentile of exit-pages per page, lenght of stay for this user and on average etc). May sound like a huge undertaking, but for someone skilled in Flash it would likely be a fairly quick development. Obviously the Flash part would have to communicate with the engine, but that should be easy as that already happens today.
.. and if I were you, I would create a quick design for this, set up a poll to ask for how many would license it separately (in addition) to Sawmill - and how much people would be willing to pay for that extra feature. I'm thinking $150. Any higher ?

I should probably mention that there are/were examples of this type of functionality(one from MIT that I recall trying out), but the problem was that these systems had none of all the other logic that Sawmill has - which meant that I would have to a) filter my data prior to loading it into these tools, b) maintain two (or three or four...) different systems for different but highly overlapping use and c) it's just too much work to do that.

Take care, guys.

titles = {
_index_html_id_12 = "Page 12 Description"
_index_html_id_13 = "Page 13 Description"
_index_html_id_14 = "Page 14 Description"
}

What I've done here is taken the page, /index.html?id=12, and converted all /, =, ?, ., etc. characters to underbars. Then I've defined a map from that "internal" representation of the URL, to the page description. Having done this, the log filter becomes:

v.page_name = page;
v.page_name = replace_all(v.page_name, '/', '_');
v.page_name = replace_all(v.page_name, '=', '_');
v.page_name = replace_all(v.page_name, '?', '_');
v.page_name = replace_all(v.page_name, '.', '_');
page = node_value(subnode_by_name('titles', v.page_name);

This log filter will change the page field value to the label specified in the file. With this in place, all you need to do is create that titles.cfg file. If you need more sophisticated logic, it's available; for instance, you could use a regular expression to extract the id 12 from the URL, even if there are multiple parameters, and then use it to look up the page. This is actually simpler; for instance, if id defined a unique page, then the log filter can just be:

if (matches_regular_expression(page, 'id=([0-9]*)')) then
page = node_value(subnode_by_name('titles', $1);

Where the titles.cfg file nodes are named after the ids instead of the full URLs. Lots of variants are possible; see http://sawmill.net/cgi-bin/sawmill7/docs/sawmill.cgi?dp+docs.technical_manual.filters for some fancy examples.

I'm going to answer the "flash" suggestion in a separate post, since it really is a complete different topic.

Great. I'll look into how I can do this using filters. It's a little bit of a hassle as a database dump would only provide me with all the written pieces, not all the other pages that are static or dynamically generated (e.g. search, FAQ, filebrowsing, downloads etc). The content of these pages would only be indicated by the title, not the URL. I suppose I could rewrite the publish tool to output the title in a tag in the URL and use internal redirection, but that also isn't ideal.

But - clearly better than nothing, so it will get me started!

Thanks & have a good night, Greg.
-D

>Regarding "titles," I should point out that though Sawmill
>does not include specific functionality to do the actual
>HTTP spidering, it *is* possible to convert URLs to titles
>if you do the spidering separately, in advance. Sawmill 7's
>log filters are extremely powerful (Sawmill 7 has a
>programming language built-in now, and filters are written
>in it), and it is particularly easy to set up a "map" file
>which maps URLs to file titles. For instance, you could
>create a file called titles.cfg, in LogAnalysisInfo, with
>this content:
>
> titles = {
> _index_html_id_12 = "Page 12 Description"
> _index_html_id_13 = "Page 13 Description"
> _index_html_id_14 = "Page 14 Description"
> }
>
>What I've done here is taken the page, /index.html?id=12,
>and converted all /, =, ?, ., etc. characters to underbars.
>Then I've defined a map from that "internal" representation
>of the URL, to the page description. Having done this, the
>log filter becomes:
>
> v.page_name = page;
> v.page_name = replace_all(v.page_name, '/', '_');
> v.page_name = replace_all(v.page_name, '=', '_');
> v.page_name = replace_all(v.page_name, '?', '_');
> v.page_name = replace_all(v.page_name, '.', '_');
> page = node_value(subnode_by_name('titles', v.page_name);
>
>This log filter will change the page field value to the
>label specified in the file. With this in place, all you
>need to do is create that titles.cfg file. If you need more
>sophisticated logic, it's available; for instance, you could
>use a regular expression to extract the id 12 from the URL,
>even if there are multiple parameters, and then use it to
>look up the page. This is actually simpler; for instance, if
>id defined a unique page, then the log filter can just be:
>
> if (matches_regular_expression(page, 'id=([0-9]*)'))
>then
> page = node_value(subnode_by_name('titles', $1);
>
>Where the titles.cfg file nodes are named after the ids
>instead of the full URLs. Lots of variants are possible; see
>http://sawmill.net/cgi-bin/sawmill7/docs/sawmill.cgi?dp+docs.technical_manual.filters
>for some fancy examples.
>
>I'm going to answer the "flash" suggestion in a separate
>post, since it really is a complete different topic.

What I had in mind was a slightly different idea. The problem with the current view is that it does not offer an intuitive view to understand the browsing pattern of the end user(s).

I did a quick google to see if I could find some examples of what I am describing, and I came up with Statviz and Pathalizer (both on Sourceforge). I'm sure you know them both. What I am suggesting is an interactive Statviz (hence Flash or whatever) that uses your engine to provide the input.

To explain the background in a little more detail:
When I study click patterns to see if a particular web site has been successfully designed etc I've found it's necessary to go beyond "entry pages" and "exit pages". By only looking at Exit Pages you don't get a very good idea of what did the person read *before* he/she exited, and generally what information did he not seek out. As an example: A person entered Software,Inc's Site from Google, thoroughly browses almost every page related to the product features. He then does a Search for "Pricing", gets a page that tells him there's no relevant results and he clicks (out of frustration?) Support which only gives him Patches and Updates Downloads. He looks at this page for 1 sec then exits.
Now, doing the analysis, one could imagine the Patches and Uploads page to be a problem, when it clearly isn't : The user was interested in the product (he spent a lot of time looking at those pages), and then tried to find the price list but failed to understand (case in point) that the page is called RTL (runtime license costs) and so it didn't turn up in the search.

One could argue that this is a constructed case, and yes it is. One could also argue that using LiveHelp or something like that might have solved that issue, but that's a little besides the point : I'd like to be able to analyse the pages and understand what triggered people to read more or leave..

So, over to the functionality:
Statviz etc offers an intuitive and highly visual (obviously) way of seeing the a) click pattern per session or b) aggregate view. Since the image is generated 'statically' it is not very much more useful than the current Sawmill functionality, but it is a step forward since non-trained people can understand this more easily.

Now, I could imagine the StatViz image to be rendered dynamically, effectively combining both the per session info and the aggregate info + overlaying info (like Statviz) is doing now but based on real time Sawmill filters rather than statically defined info that Statviz relies on.

Did this give you an idea?

-D

For the specific example you're describing, you're really looking at a specific session, rather than doing session aggregation (like the Paths reports do). If you click that user's session in the Individual Sessions report, and zoom on Log Detail, you'll see the session, click-by-click, and you'll be able to make sense of what happened there. Are you asking for a graphical Statvis-style depiction of a *single* session? Or are you looking for a graph of all sessions, aggregated? It sounds like you're asking for an aggregated report which somehow morphs into a single-session report, maybe through filters. Would it suffice to have controls in the Paths reports which zoom in on the sessions being described by a particular page, or a particular path segment, so you could look at those in aggregate, and zoom in on particular sessions by clicking something there?

I guess what I'm asking is: can the existing reports be enhanced to provide the same functionality? Without jumping to a completely new Flash report, could the information you want be embedded in the existing Paths reports?