Network Science Research · Behavioral Science Analysis
Cutting an adoption timeline from a year to months with behavioral network data
A person's social context shapes their choices. As the lead researcher, I combined survey data with a map of who works with whom to cut one organization's product-adoption timeline from a year to months.
The problem
The organization had switched to a new project management tool, but adoption was sluggish: as many as 70% of people had heard of it, yet many were still on the fence about actually using it. Stakeholders needed a way to understand why adoption was stalling and how to turn insight into action.
Approach
I combined two kinds of data: users' attitudes toward the new tool and their knowledge and use of it, and the map of who they most frequently worked with and how strong those working relationships were. Overlaying behaviour onto the social graph made the underlying adoption dynamics visible.
What we found
Adoption behaved like a complex contagion: people didn't switch on a single recommendation, they switched once enough of their working circle already had. That tipping point averaged about 65%, but it wasn't the same for everyone. It ran higher for the more change-resistant, and, tellingly, the most resistant people tended to sit at the chokepoints between teams. So the real blocker wasn't company-wide indifference; it was a handful of well-placed holdouts, each gating the teams behind them. Overlaying attitude onto the working graph made those choke points, and the user groups stranded behind them, visible for the first time.
A note on method
Any serious research on behavior change will note that the mechanisms that shape it are almost always multi-layered. Complex contagion is one mechanism behind adoption, and our goal was to measure it in the context of our client's real organization rather than assume it from theory. The ~65% threshold we found for adoption came out of a logistic regression on real ego-network data, but we knew that couldn't be the full story, or else no one would adopt the change until the majority of people did. That's a chicken and egg problem, but getting deeper into the weeds to hunt for causality, with expensive studies to separate true peer influence from people who just already think alike and naturally cluster together, wasn't feasible. Taking a pragmatic view, we realized we didn't actually need to solve that problem to meet our goal of improving adoption. Instead, we used the threshold as a planning lens for where to intervene, not as a claim about cause. The data we had was already enough to act on.
Outcome
Knowing where adoption would catch and where it would jam turned a budget problem into a targeting problem. Rather than blanket the whole org with endless email blasts or court every possible influential stakeholder, we did two things. We seeded adoption in the clusters where it would spread on its own, and we spent scarce, high-touch effort only on the small set of holdouts who both sat at a chokepoint and held formal authority, the people who could otherwise stall everyone behind them. That combination pulled adoption from a projected year-plus down to roughly three months.
- ~65%
- Tipping point distribution & average
- ~3 Months
- Adoption, down from a year+