In the past, I’ve found Thomas’s Kickstarter analysis for game-projects very useful. Today I found something that complements them: an excellent visualization of Kickstarter projects.
It takes a little thought to understand how to use it, so here’s some ultra-quick example analysis…
WTF is this thing?
You get an odd scatter graph, with a precise readout attached to your mouse position (gives you the exact X and Y co-ords of your position):
- You can only see a tiny fraction of the graph by default, you have to click the “zoom” word in bottom left, and click the – minus button repeatedly
- Most of the axes data is missing, to make it “look cooler”, perhaps (hmm). Instead you need to move your mouse around the screen to see what you’re looking at
Some interesting things you can find quickly
Lines drifting up the screen
…represent “most people do Kickstarters at round-number target amounts (x-value is the same), but the actual amount raised varies widely (y-value has many different values)”
Bright lights along the y=x (Gradient:1) slope
At all scales, you see a cluster of lights in a diagonal line going up and to the right:
The line represents “project raised EXACTLY its target amount”.
Immediately we can see something interesting: apart from “tiny” projects ($3,000 or less), the vertical upwards streaks are considerably brighter than the y=x diagonal line. In fact, most of the y=x line is missing – we see instead an implicit line, made up of the bottom edges of bright streaks.
If we didn’t already know better (from watching Kickstarter), we’d expect the y=x line to be a cluster that blobbed-out both above and below the strict y=x line. Instead … it’s a hard-edge below, and blurred upwards above.
i.e: Projects over $3k generally succeed wildly – with lots of variation in the “overshoot” – or fail by a wide margin (blackness beneath the y=x line)
How fractal is it?
We’d expect this graph to be fractal and self-similar: as you zoom out, it should look almost the same as when you’re zoomed in.
Let’s compare the graph at 0-$500:
$0 – $2500:
$0 – $10k:
$0 – $50k:
and finally: $0 – $200k:
Some quick broad-strokes observations (NB: these are not statistically accurate – I’m simply pointing out the ultra-fast, at-a-glance analysis you can do here. If you’re following Kickstarter a lot, and don’t have time to keep re-analysing in detail, this gives you a fast view):
- this part is easier to see for yourself by actively zooming the app in/out, so you can see the animated change in self-similarity.
- Most of the range is self-similar.
- Below $300-$500, it’s very different – the normal trends of KS don’t apply
- Projects over $30,000 funding target do badly on KS: much higher rate of failure
- Projects $1k-$8k tend to do well: much higher rate of success
If you need less than $30k, it doesn’t matter how much you ask for
Look at the left-hand edge of the $200,000 graph again:
Looking “above” the x=y line, we see that the density of dots is almost constant left-to-right. If that’s accurate, it means that the final amount raised is largely independent of the amount they originally asked for.
That’s interesting for projects that intend to set their funding bar lower than possible, in order to raise the chances of getting “some funding” as opposed to “nothing”.
This analysis sucks
Why, yes! – yes it does. If you want accurate analysis, it’s not enough to take quick guesses (as above). You absolutely need to look at calculated co-dependence/independence, rather than guessing (because those figures are almost impossible to do correctly in your head – you want a spreadsheet or computer to calc them for you).
But these days it’s so easy to collect metrics that we often find ourselves overwhelmed by numbers. Many of us fall back to glancing at the raw numbers and guessing trends from them – which is the very worst option of all (humans are terrible at accurately doing stats-analysis by gut feel).
In those situations, where we’re too lazy / overwhelmed to look at the correct figures, visualizations like this can help a lot.
Also … although I’ve looked at many spreadsheets and presentations on KS stats, I’d never internalized quite how much people stick to the “round numbers” of funding targets. This visualization made that a lot more obvious to me. Similarly with the variance on “actual amount received, when successful” – I knew it usually overshot by a large amount, but I had the impression (from Kickstarter.com’s own press releases and conversations) that the overshoot was clustered … and yet clearly it’s not.