Social Network Webbing: Finding the Instructors Other Instructors Copy

Course Guru shipped completion nudges — automated check-ins to learners who stall mid-course — to every instructor on the platform in July. By September, 400 instructors were eligible and 26 had turned it on. The feature worked; the instructors using it saw meaningfully better completion. It just wasn’t spreading, and our growth-shaped instincts were all pointed at the wrong lever: another in-app announcement, an email sequence, maybe an incentive.

We ran Social Network Webbing on the instructor population instead, with a room that had no instructors in it, which is an unusual way to run this structure and needs justifying.

The structure

Social Network Webbing maps the informal relationships around a specific challenge and then picks next actions from the map. Small groups, one big sheet each: name the challenge, list every relevant person or group on sticky notes, place them, draw the connection lines, then read the resulting web for hubs, brokers, clusters, and gaps. Finish by naming two or three concrete conversations to have next. Sixty to ninety minutes.

Normally the people drawing the map are the people in the network. Our network was 400 external instructors, and we weren’t going to convene them for a workshop. So the drawers were seven of our own people who each held partial knowledge of that network: two support reps, the instructor-success lead, a community manager, a solutions engineer, and two people from marketing who ran the newsletter. Each knew a different slice. The map was assembled from overlapping partial views rather than lived experience, which makes it weaker evidence — and we treated it that way, as a set of hypotheses to check rather than findings.

What the sheet showed

The challenge sentence was “get completion nudges used by 150 instructors by January.”

The first surprise was structural. Three of the four groups drew clusters that were not platform-shaped at all: a Discord server for photography instructors with about 90 of our users in it, a private Slack for a cohort of business-course creators who’d all come off the same course-creation bootcamp, and a much looser group around one YouTube channel about course marketing. None of those spaces belonged to us and none of them appeared in any of our segmentation.

The second surprise was who the hubs were. Our top instructors by revenue — the ones on our advisory panel, the ones we emailed first with every feature — had very few advice-lines drawn to them. They were, as the community manager put it, admired rather than consulted. The heaviest hub on the sheet was an instructor with a mid-sized course who answered questions constantly in the photography Discord and had, according to two support reps independently, been the origin of at least four platform habits that had spread widely, including a course-structure convention we’d once mistaken for our own onboarding working.

Nothing in our rollout had ever reached that person deliberately.

The gap the map made visible: we had strong lines from us to individual instructors and essentially no lines into the places instructors talk to each other. Every distribution channel we owned was one-to-many from the platform; the network was many-to-many and elsewhere.

Next steps were three conversations, not a campaign: ask the Discord hub to try nudges on her next cohort and say whatever she honestly concluded; ask the bootcamp-Slack cluster’s most active member what a nudge would need to do to be worth turning on; and stop treating the advisory panel as a distribution channel, since it plainly wasn’t one.

By January we were at 168.

Why the drawing surfaces this and a survey doesn’t

A survey asks people what they do. The web asks a room what it knows about who influences whom, and then makes the answer geometric — clusters and empty space are visible in a way that tabulated responses aren’t. The gap we found wasn’t a fact anyone stated; it was a shape on the paper, an entire region of the sheet with no lines connecting it to us.

Running it with staff rather than network members is a real compromise and worth naming. What made it defensible was multiplicity: four groups drew independently, and the claims that appeared on three or four sheets — the Discord cluster, the advisory panel’s low advice-weight — were the ones we acted on. Anything that showed up once, we treated as one person’s impression. That rule kept the session honest.

When to skip it

Don’t do the staff-drawn version if you can plausibly get the network in the room; you’ll get a better map from six actual instructors than from twelve employees. And don’t do it at all if you can’t act on what it finds — the natural conclusion of a map like ours is “go participate in communities you don’t own,” which requires a genuine willingness to show up somewhere you can’t control the messaging. If leadership won’t sanction that, the session just produces a well-drawn frustration.

It also fails when the adoption barrier is product-shaped rather than social. If instructors weren’t turning on nudges because the setting was buried three screens deep, the map would have shown a healthy network cheerfully not using a hard-to-find feature. Check the obvious mechanical explanations before you convene anyone.