Maintenance
Part of Only one of the two engines behind a viral trend can be planned
Viral reach: six properties that decide whether strangers ever see it
Viral trends platforms and the structural routes to strangers: six properties that decide whether wide spread is possible, and how to read a surface.
Whether something can spread at all is a property of the surface it is on, not of the material. Put the same clip on two products and one of them can carry it to strangers while the other cannot, because the second has no path from a person to someone who does not already follow them.
This page is about those paths: which structural features make wide spread possible, which suppress it, and what each implies for what you should even attempt.
What to take away
- Spread needs a route to people outside your own audience. If the surface has no such route, no amount of quality produces reach.
- The route matters more than its width. A surface with a narrow but genuine path to strangers beats one with a wide path to people who already follow you.
- Whether a surface is built on similarity or on interest decides what kind of material can travel on it.
Six structural properties
| Property | The question it answers | Effect on spread |
|---|---|---|
| Out-of-network distribution | Can the system show your item to non-followers | Without it, ceiling equals your audience |
| Republication mechanics | Can one person put your item in front of theirs | Multiplies reach along social ties |
| Format reuse | Can someone build on your item rather than only pass it | Turns one item into a family of items |
| Graph openness | Are connections public and one-directional | Open, one-directional graphs spread further |
| Persistence | Does an item stay findable after its first day | Decides whether a second wave is possible |
| Unit size | Is the atom a clip, a post, a thread, an album | Governs how much context survives a hop |
The first two are the load-bearing ones. Everything else modifies the rate; those two decide whether the question is live at all.
Format reuse deserves its own note because it changes the arithmetic rather than the speed. When a surface lets people make their own version of a thing, participation replaces transmission, and the item stops being one object that travels and becomes a template that regenerates. That is a different mechanism from passing something along, and the same distinction sits behind the two engines described in the overview.
Similarity versus interest
Surfaces differ in what they use to decide who sees what, and the difference shows up as a hard limit on which material can travel.
A surface built on connections is built on homophily, the strong tendency of connected people to resemble each other. Material spreads well within a cluster and stalls at its edges, because the ties that carry it also define the group's boundaries. What crosses those boundaries is unusual and usually not what the maker expected.
A surface built on inferred interest is not bounded that way. It can put an item in front of anyone whose behavior resembles that of people who responded to it, whatever their social position. That removes the boundary problem and introduces a different one: your existing audience is no longer a reliable route to anyone, so a following is worth less than it looks.
Neither arrangement is better. They favor different material, and knowing which you are on tells you whether to build ties or to build for strangers. The differences between surfaces are set out further across the seven feed surfaces.
Why the network shape sets the ceiling
Two ideas from network research do most of the explanatory work here, and both are worth having in mind rather than in an argument.
Short paths exist between almost any two people, which is the ordinary content of six degrees of separation. The paths existing is not the same as anything traveling along them. A route can be three hops long and still carry nothing, because each hop requires a person to act.
Networks with mostly local ties plus a few long ones behave very differently from purely local networks. That is the small-world property, and the few long ties are what make wide spread possible at all. None of it is visible from outside in any detail, which is the limit described in an outside view of a closed system.
In practice a small number of accounts with unusual reach across clusters matter far more than the total number of connections, which is why reach is so lumpy and so hard to plan around.
Reading a surface before you commit to it
Four questions, answerable by observation in an afternoon.
Can you find items from accounts you do not follow, in ordinary use? If not, there is no out-of-network route worth building for.
When something reaches you, can you tell whether it came through a connection or through inference? The answer tells you which mechanism you are working with.
Do items from months ago still surface? If not, everything you make is a one-day asset and your planning horizon should match.
Does the surface reward remaking or only passing along? This decides whether your best move is a finished object or a usable template.
The answers give you the ceiling. What you do under it is a separate question. Structural pressures on product design also mean these properties move over time, so a reading from a year ago may no longer hold.
Common questions
Can a surface with no out-of-network route ever produce wide reach?
Only through people carrying items to other surfaces by hand, which happens and is slow. If that is your route, the item has to survive being stripped of its context, because nothing travels with it.
Does a large following help on an interest-based surface?
Less than it once did, and less than people assume. It helps as a starting signal and as an audience that reliably responds, which can matter for whatever the system does next. It is not distribution in itself.
Is a closed or private surface worthless for spread?
For measurable spread, largely yes, and that is not the same as worthless. Material moves through private channels constantly and leaves no trace you can count, which is one more reason the numbers you can see are not the whole picture.
Which property should I optimize for?
None of them, because they are properties of the surface rather than of your work. What they are for is choosing where to spend effort and setting an honest expectation before you start, which is the same job the durability tests do for a change.
