First you get the Page Views, then you get the Contributors, then you get the Money…then you get the Power?

Jai’s post showing that total dollars raised is a direct function of number of contributors made me very curious. Number of contributors alone leads to ~87% of the variation being retained in the model. Wow.

But something troubled me. I had shown a decent (not as good) relationship with number of project views and dollars raised. In some ways this made more sense, as a huge number of project page views might mean there was something inherently awesome in the project that might mean each individual donor might give more money.

Do both number of page views and number of contributors jointly play a role in determining how much money a project raises? Understanding both might help guide further explorations into what sorts of activities either just bring in more raw page views, or bring in more actual contributors. I.e., do we need to worry about quantity of viewers or quality (willing to open their wallets) of viewers?

First I thought maybe there was some interaction effect – $ = project page views * # of contributors. But, when I tested that model, no dice (F test p>0.2).

Still, it seemed to me something else might be going on there to explain total amount raised. Maybe the signal from page views to total raised was clouded by number of contributors, and that there is some additional information about the quality of a project or the quality of its outreach that explains total amount raised after the signal connecting pageviews to donors is taken out. Maybe there is a direct and indirect effect of project page views on total raised.

And that means we need to do some Structural Equation Modeling to tease those apart. Oh yeah, baby. Time to get my lavaan on. (For those who don’t know me, I <3 SEM and even teach workshops in it). (Also, for those who don’t know SEM, which is probably most of you, it is a technique that, among other things, allows one to build multivariate models that address the strength of direct versus indirect effects.)

So the question is this – is the effect of number of page views on the total amount raised explained soley by the relationship between number of views and number of contributors? Or, is there some additional effect of page views on the total amount raised? This is a classic test of mediation. So I constructed the following two models using pre- and post-goal page views (i.e., number of page views before a project hits its 100% mark, and number of page views after) as I’ve shown previously they do slightly different things:

Note – I thought perhaps post-goal page views might be a function of pre-goal, so I included that path in both models.

For a test of mediation, for the non-SEM-nerds in the house, one runs both models, and then compares their Likelihood Ratio Chi-Square values. If they are not different (typically p>0.05 – that’s right, greater than!), one goes with the simpler model.

So….I ran the models (on the data without the outlier from before) and did the test, and, indeed, dropping those direct paths from project page views to total amount raised doesn’t affect model fit. So it’s all about indirect pathways from page views -> contributors -> $$$. There’s also no real relationship between pre- and post-goal page views, so I dropped that as well. Here’s the final model.

I feel quite good about this, as not only is the R2 for total about 0.86, but, it’s 0.81 for contributors. This model explains a lot of what is going on in the data.

So what does it mean. A few interesting points. First, the average contributor is worth about $60. Before you hit your goal, you’re likely to get about 1 contributor for every 100 page views. But, after you hit 100%, you get 1 contributor for every 20 page views. So, again, success breeds success. (And incidentally a lower goal will get you to ‘success’ faster.)

Getting eyeballs again seems to be key. So good outreach is going to be essential to the success of a #SciFund project. And getting eyeballs on a project that people are excited about seems to be even MORE important.

Now, there are a few open questions and issues left – 1) ok, so, given the data we have, what leads to pre and post-goal pageviews? Can we quantify that good outreach? Also, 2) It’s still possible that we need to fill in some additional variables here – things about a proposal or the kinds of people coming in as viewers that modify whether they are likely to go from just viewing a page to becoming a contributor.

Still, a good baseline model, and a lot to chew on.

Share

Trackbacks

  1. [...] our research on the first round of #SciFund shows that people are FIVE TIMES more likely to give you money once you have hit your [...]

  2. [...] some analyses on which campaigns were the most successful, and sees that, perhaps not surprisingly, the number of pageviews correlates to the amount raised. Further, he finds that the size of network of the scientist directly relates to the amount of [...]

  3. [...] As we’ve seen, to understand how well a project is funded, we need to know how many eyeballs are coming in to view a project. A blog can be one key portal to bringing folk in to see your project. In particular, a blog can be HUGE early on, as once your project goes live, it can be an access point through which you can funnel your readership to your project. [...]

  4. [...] Fortunately, we’ve got a number of metrics as grist for this mill – number of twitter followers of participants, number of project tweets, Facebook friends, Facebook likes, and blog presence. And we know that ultimately they need to be impacting page-views to have an impact on total raised. [...]

  5. [...] As we’ve seen, to understand how well a project is funded, we need to know how many eyeballs are coming in to view a project. A blog can be one key portal to bringing folk in to see your project. In particular, a blog can be HUGE early on, as once your project goes live, it can be an access point through which you can funnel your readership to your project. [...]

  6. [...] very curious. Number of contributors alone leads to ~87% of the variation being retained in…Via scifund.wordpress.com Like this:LikeBe the first to like this [...]