I guess my complaint is more to do with the submission title :)

However... there is an intriguing idea that you might be more successful by only doing it the dropbox way. It seems counterintuitive to restrict your options, but then you really would be productizing the dropbox approach - you'd be offering not only the implementation of it, but also a guide to it, an explanation for why it works etc. Not technology, but valuable business guidance. It would also be simpler and more convenient, by being less customizable to use; and more definite and concrete in your marketing materials, making it easier to understand and pass on, and be stickier.

But developers hate artificially limiting their technology; and I'm far from able to guarantee it would sell better (or, indeed, at all).

I guess you could simulate it, by having a form/wizard, that guides the user through the issues of a dropbox-style referral - sort of like "referral model advice" - but one that ideally would also create the system on the spot! I'm thinking that many people have much less idea of referral programs than you do (and value their ignorance); making the choices simpler is a big plus for these folk. Of course, I don't know if referral systems are well-understood enough yet to be reduced to simple rules that always work (or work reasonably reliably, well enough for a first iteration for the client).

We started with one simple model under the assumption there's a "best" model. Now we're discovering there's more than one "ideal" model - different models work better on different sites (although some things work well consistently, and some things consistently don't). For example, on some sites providing an incentive for the referrer increases conversion (not surprising).. but on other sites, it hurts conversion - the social value of altruistically sharing a good deal outweighs to social cost of sharing something to your own benefit.

Now our product vision is to offer a small variety of simple models proven to work well, to get people started. But we also let hackers tweak stuff so it can work more seamlessly with their products.

Allan

Iterative experiments to discover what works, then encapsulating it in rules - sounds extremely promising.

This (i think) would increase implementation difficulty. While the walkthrough on "how to use it" at the front end would be an easier on-boarding process, but the complexity of figuring out what unit "more dropbox" was in most sales wouldn't be as easy as letting the seller define any offer and defaulting to the easily understood/implemented discount offers.

re: ethics, if we look at curebit as a platform that's encouraging more ethical referrals (by focusing on pre-built social network relationships and encouraging retailers to use best practices), we shouldn't expect them to have to limit their growth by forcing early users into unfamiliar patterns in order to be the moral police on anyone wanting to try their platform.