...in a non-self-helpy-buzz-wordy sort of way.
Being something of a nerd (you'd never guess ;) ), I keep detailed statistics on the performance of The Second Realm. This involves a huge 5-page spreadsheet, most of the content of which is obscure and unhelpful. What it actually boils down to is a handful of graphs, and I thought it might be useful to talk a bit about what they are and what I use them for.
The Most Important Graph:
Well, the effect is psychological. This graph exists to take the sting out of bad weeks by showing how insignificant they are in the grand scheme of things - can you even spot the most recent bad week? It's not the dip in early January, it's the tiny little blip at the start of April. What this graph shows, more than anything, is that whatever momentum I have managed to establish isn't really diminishing at all.
In an ideal world, this would show an upward curve, indicating growing momentum, but I'll settle for what I've got for now. The reason this one is so important is that it keeps me going when all the other graphs turn discouraging.
This is probably the one I use most (and you've seen it before, a couple of times). The blue line is the number of total downloads across all episodes for a given week, and thus tracks overall performance on a weekly basis. The red line is a four-week rolling average (the total for the current week and the preceding three, divided by four), and the green is a twelve-week version of the same.
The red line is a bit obsolete now, but the purpose of the two averages is to even out the big spikes in the blue line (most of which represent the release of a new episode). What they reveal is that the end of last year went badly for me, but things have been picking up since the nadir in mid-February. It's worth noting, too, that the actual traffic nadir will have been earlier than the dip in the green line - this is the disadvantage of the rolling average measure.
The usefulness of this graph comes from the limit it places on what might be going on. A flat green line would mean steady traffic - losing as many readers as I gain. Crucially, it would also mean that my reach - the number of new readers trying out one of my stories for the first time - wasn't growing.
The slight upward trend means one of three things; growing reach (more readers trying me out), growing conversion (more reached readers coming back for another episode), or growing devotion (regulars coming back more often). Since all these cases indicate growing interest, they're all good things. Now, last time I made any predictions about the performance of The Second Realm, I was proved wildly optimistic, but I remain quietly content this time. We'll see how the rest of the summer plays out, which brings me too:
Traffic per Episode:
Anyway, the red bars are basically the bars from the last few weeks superimposed on the equivalent weeks last year, which will in theory allow me to compare this year's performance to last year's. Hopefully this will allow me to start identifying seasonal effects (i.e. do people just download more free short stories in the summer? Is there a key popular time of year?), though obviously with only two years' data to draw on there's a limit to the conclusions I'll be able to draw.
As far as reading this graph goes, the important thing is the angle of the slope within a given segment. Each jag represents a new episode, and the angle of the slope tells me how well that episode did at generating new traffic for the series as a whole. There is a caveat to this, which is that because it's an average, as time goes on and more and more episodes appear, each individual download of an individual episode becomes less significant. I haven't quite worked out how to compensate for that yet.
A steady upward trend (such as appears for about the first third of the graph) would indicate that the increase in traffic each month was keeping up with this effect - and would actually be equivalent to a significant upward curve on both the previous graphs. Clearly that's not happening, but much of the drop-off in the last six months doesn't actually represent a lack of growth, just growth that's not keeping pace with the appearance of new episodes.
And finally, just for kicks, a graph that isn't useful at all:
blogging as promotion last month, but didn't actually show it because, well, just look at the thing. It's a bit... chaotic, no? The blue line should be familiar, where it's visible at all - it's the same blue line as was on the second graph, above. The red and green lines here, however, are completely different.
What they actually measure is my activity on this blog (i.e. number of posts per week) as a proportion of my average activity (total number of blog posts divided by total number of weeks). The key lists them both as percentages, but the green line is actually per-fifty rather than per-hundred. Basically, towards the right-hand end of the graph, you can kind of see a trend line curving gently downwards - that represents weeks where I actually blogged twice. spikes above it mean three or four posts, below means one or two. The red line is just crazy and should be ignored.
What I was looking for was evidence that activity on this blog translated into traffic for The Second Realm, because I'd developed the conviction that whenever I had a week away from blogging, traffic really dipped. If that conviction had been correct, though, there'd be a consistent correlation between dips in the green line and dips in the blue, but for every week that appears to validate the belief, there's one that refutes it. I keep the graph going these days mainly out of aesthetic curiosity and chagrin.
Anyway, that's a few of the perks and pitfalls of maths for tracking your progress as an author. Don't feel you need all (or any) of these graphs, but if you think you'll find them useful, excel makes things fairly simple (although, apparently, too complicated for Harvard economists) to implement. Just make sure you've got plenty of data to draw on - the more data you have, the less discouraging an individual bad point will be.
I should add - these are very simple measures, and I'm no statistician. There are all sorts of other techniques I could use, except that I have no idea how to set them up or what they'd mean.