“Innovation” is usually shorthand for a new product. Social innovation is broader and, in practice, considerably slower: it covers new services, new relationships between institutions, new ways of organising a community, or a policy redesigned around how people actually behave rather than how a form assumes they behave. The output is not always something you can hold — often it is a new way of doing something that already existed.
Why it does not scale like a product
A software product scales by copying code. A social innovation scales by convincing people, institutions, and incentive structures to change how they operate — and every context has a different starting point, a different set of local relationships, and a different set of actors who need to agree. A parenting programme that works because of one exceptional facilitator, or a community scheme that works because of one trusted local leader, will not automatically work somewhere else with a different facilitator and different trust. This is the central, under-discussed problem of the field: most social innovations are proven in one place and fail to generalise, not because the idea was wrong, but because the conditions that made it work were never made explicit.
Testing before scaling
The more useful question is rarely “does this work?” but “under what conditions does this work, and which of those conditions can we actually reproduce elsewhere?” That reframing changes how a pilot should be designed from the outset: instrumented well enough to separate the effect of the idea from the effect of the specific people delivering it, tested in more than one setting before anyone commits to a national rollout, and honest about the version of the idea that failed along the way, since that version usually contains the most useful information.
A more useful definition of success
A social innovation has succeeded not when it launches, and not even when an evaluation shows a positive effect in one location — but when it has been adopted, adapted, and sustained by people who were not part of the original team that designed it. That is a considerably higher bar than most pilot projects are held to, and it is the one that actually predicts whether an idea changes anything at scale.