Costs
Part of How to tell whether sharing or recommendation carried a viral spike
The risks of going viral arrive in a predictable order
Viral trends risks in the order they arrive: the mismatched audience, the unreadable baseline, and the pressure to repeat that forms afterward.
A hit is treated as a windfall, so its costs go unbudgeted. Some arrive immediately and are obvious once you have lived through one: the support load, the moderation load, the week you lose.
Others arrive slower and cost more. The worst is that a hit changes who your audience is without asking you. This page sets out what a burst of attention costs, in the order the costs arrive.
What to take away
- The audience a hit brings is selected by the hit, not by you, and it is usually a poor match for everything else you make.
- The most durable damage is to your instrumentsafter a hit, your own baseline is unreadable for as long as the spike lasts and some time after.
- A hit is a poor input to a decision, because it is one observation drawn from the extreme tail of your own distribution.
The costs, in arrival order
| When | Cost | Why it is missed |
|---|---|---|
| Hours | Support and moderation load on a team sized for a normal week | Nobody plans capacity for the outlier |
| Days | Attention displaced from everything scheduled | It feels like the right call at the time |
| Weeks | An audience acquired on one thing, expecting more of it | The new numbers look like growth |
| Weeks | A baseline that no longer describes anything | Instruments are assumed to be neutral |
| Months | Pressure to repeat, applied to every subsequent decision | The pressure is internal and rarely named |
| Months | A public association with a single item | Only visible when you try to move on |
Audience mismatch is the expensive one
People who arrive during a burst came for the item. By construction, that item is unrepresentative of your work: it is the thing that traveled, almost always more extreme, funnier, angrier or simpler than your usual output.
Before vs during the burst
Before burst
- Came for
- your usual work
- Selected on
- normal output
- Response rate
- stable baseline
- Reading it as
- quality signal
During burst
- Came for
- the one item
- Selected on
- extreme or simpler
- Response rate
- falls, denominator changed
- Reading it as
- quality problem (wrong)
The audience that follows it is selected on that property. The route that brought them was set by the structural conditions on the surface, not by you.
What happens next is predictable. Your ordinary material now reaches a group that did not come for it. Response rates fall, not because the material got worse, but because the denominator changed.
If your instruments report engagement as a rate, they will report a decline unrelated to anything you did. Reading that decline as a quality problem, and adjusting the work to fix it, is how a single hit reshapes an operation over a year.
Keep the cohorts apart: measure people who arrived before the burst and people who arrived during it separately. Measure for as long as it takes for the difference to stop mattering; aggregating them destroys exactly the information you need.
The mechanics of that separation are the same as in reading a cohort rather than a total.
The incentive that forms afterward
Once a burst has happened, it becomes the reference point. Every subsequent piece is judged against the best week you ever had, and the pressure to reproduce it is applied by people who mean well, including you.
This is an ordinary perverse incentive: a measure meant to show things were going well becomes the thing pursued, and the pursuit degrades what it measured. In practice it pushes work toward whatever traveled last time.
That target is narrower than your actual subject, and it stops working as soon as several people aim at it.
Pressure is worse when the thing that traveled was provocative. Material that sparks a strong reaction spreads efficiently. That is a fact about the mechanism, not the material's worth.
Chasing that efficiency deliberately leads toward clickbait, a term for a real, specific failure: packaging promises what contents do not keep. The cost is paid the second time, when the audience remembers.
The wider version of this trade, where attention rather than money is the scarce resource being competed for, is the subject of the attention economy, and it explains why the pressure is structural rather than a failure of character.
Costs you do not control
Some of the risk is not yours to manage. A burst can attract organized hostility, or a wave of accounts with no relationship to your subject, or an association with a dispute you had no part in. None of these are preventable and only the response is in your hands.
Prepare by deciding in advance what you will not do. Choose comment sections you will not enter, claims you will not respond to, and when to stop reading. Deciding during a burst is deciding badly.
General form of the argument: some risk in any system belongs to somebody else. It is set out in the risks that are not yours.
What a hit is actually good for
It is good for two things. It is evidence that a route to a wide audience exists from where you are, which you may not have known. And it is a supply of new material, in the form of what people said when they arrived, which is often the most honest description of your work you will ever get.
It is not evidence that the thing you made was good, and it is not a plan. One draw from the tail tells you almost nothing about the distribution.
Common questions
Should I avoid trying for reach at all?
No. The argument is for pricing it, not for refusing it. Knowing that a hit brings a mismatched audience and an unusable baseline lets you decide whether this attempt is worth those costs, which is a decision you can only make before. Price it against the ratio worth measuring, not against a total.
How long does the baseline stay unreadable?
Until the burst audience has either engaged or gone quiet, which takes longer than the spike does. There is no fixed figure and anyone offering one is guessing. Watch for the point where your rates stop moving, then treat that as the new normal rather than trying to restore the old one.
Is a smaller, slower audience genuinely better?
For most operations, yes, because it is an audience selected by your ordinary work rather than by one outlier. That is a defensible position and not merely a consolation, though it is worth being honest that it is also slower.
What if the hit is on a subject I do not want to be known for?
Then the association is the main cost and it will outlast the numbers. The only real response is to keep producing what you actually do, accept a period of poor rates while the mismatched audience settles, and resist the pull described above. The decision about what not to do is worth writing down before you need it.
