What to do when a link goes viral

The redirect is rarely what breaks

A redirect is a small, cacheable, stateless response. A properly configured redirect layer will absorb an enormous multiple of normal traffic without difficulty, because it does almost no work per request. What breaks is the destination: a page that queries a database, renders a template, calls three services and loads a heavy set of assets.

This asymmetry is useful. It means your short link keeps working and keeps counting while the destination struggles, which gives you both a measurement of the surge and a control point for responding to it. The link is the only part of the chain you can change instantly without a deploy.

Recognise it early from the shape, not the volume

Viral traffic does not look like a campaign. It arrives from a referrer you did not plan for, concentrates in a country or a time zone you were not targeting, and grows on a curve rather than stepping up at a send time. By the time a total looks unusual, the shape has been unusual for a while.

An alert on rate of change, rather than on an absolute threshold, catches this earlier. A link doing ten times its own trailing average is interesting regardless of whether that number is large in absolute terms, and it is the earliest signal available from data you already have.

Find the source before deciding anything

The response to a link circulating in a community that likes your product is different from the response to a link circulating in a thread criticising it, and different again from the response to a link being used in a scam impersonating you. All three produce the same graph. Look at referrers, search for the link text publicly, and read what is actually being said before you act.

Do this first even under pressure. Most bad decisions in a traffic surge come from responding to the number rather than to the cause, and the two most common overreactions — disabling the link and rewriting the destination page — are both wrong in at least one of those three scenarios.

Use the link as a valve

If the destination cannot cope, repoint the link. A lightweight static page carrying the essential message, hosted somewhere that cannot fall over, is a far better experience than a timeout, and it can be prepared in advance and left dormant. Repointing takes seconds and requires no deployment.

Keep the same short link throughout. It is already published in places you cannot edit, and changing the URL abandons everyone who has the old one. Changing where it goes preserves the audience while changing what they get.

Get the caching right at the destination

The single highest-leverage technical change during a surge is making the destination cacheable. A page that can be served from an edge cache for even sixty seconds turns an unbounded load into a bounded one. Look for the things that prevent caching: a per-request personalisation block, a session cookie set on every response, a cache-control header set to no-store by a framework default.

Removing personalisation temporarily is usually acceptable and usually sufficient. A visitor arriving from a viral link has no relationship with you yet and loses nothing by seeing the generic version of the page.

Bot traffic will be a large part of it

Attention attracts crawlers, preview fetchers, security scanners and scrapers. Every platform where the link is shared will fetch it to build a card, sometimes repeatedly. A surge is therefore always inflated relative to human interest, and the inflation is not uniform across sources.

Do not clean this up in the moment. Record it, note that the figure is unfiltered, and reconcile later against destination-side measurement. Announcing a headline number during a surge and revising it downwards a week later is a predictable and avoidable embarrassment.

Decide what the moment is for while it is happening

A surge is a one-off supply of attention that will not repeat. The question is what you want from it: sign-ups, an email address, a follow, or simply that people leave with an accurate impression. Answer that explicitly, because a destination optimised for none of those wastes the entire event.

Small changes matter disproportionately here. A single clear next step on the page, without a modal covering it, will convert better than an elaborate funnel assembled under time pressure. Prefer the smallest change you can make confidently.

Write down what happened within a week

Once traffic subsides, record where it came from, what the curve looked like, what you changed and what the result was. This is the only version of this event you will ever have, and the details will be gone from memory in a month.

The most valuable output is usually a shorter list of preparations for next time: the dormant static page, the alert on rate of change, the cache header that had to be fixed by hand. Turning those into standing configuration is what makes the second surge routine rather than dramatic.