Your Super Bowl Ad Worked. Your Website Didn't.

In less than two weeks, companies will spend upward of 2.7 million dollars per thirty seconds of Super Bowl airtime. A predictable fraction of them will spend the following Monday explaining why the website in the ad did not work.

It happens every February with the reliability of the game itself. The spot airs and a national audience reaches for laptops. The campaign microsite, built beautifully and tested by four people on an office connection, collapses inside ten minutes. The traffic the ad was designed to create becomes the traffic the infrastructure was never asked about.

We have spent enough Januaries hardening sites for scheduled moments that the playbook is worth publishing. Not Super Bowl scale. Admissions deadlines, giving days, product launches, broadcast mentions. None of it is exotic. All of it is regularly skipped, because it lives in the gap between the marketing meeting and the operations meeting, and that gap is where launches go to die.

The math nobody runs

Start with the number the campaign implies. A spot reaching ninety million viewers that moves one-tenth of one percent of them produces ninety thousand visitors. They arrive not across a day but across the minutes after air, which means peak load in the tens of thousands at once.

Even the humbler versions of the scheduled moment obey the same math. An email to fifty thousand addresses with a strong call to action can put four figures on your site inside the first half hour. Compare that to the Tuesday average your infrastructure was sized against, a few hundred at once for a healthy mid-sized site, and the problem stops being abstract.

The question is not whether your normal capacity handles the moment. It is what multiple of normal the moment represents, and nobody at the launch meeting has usually calculated it. So calculate it, crudely. Crude is fine. Audience size, times an honest response-rate range, spread over the arrival window the media plan implies. If the top of that range exceeds your tested capacity, you now have an engineering requirement with a deadline. That is infinitely more useful than a hope. Capacity is a marketing budget line. If the ad works, the server bill is the good news.

The playbook, in the order we run it

• Load-test the real paths. Not just the homepage. Test the actual path the campaign drives: landing page, form, confirmation. Forms and database writes bottleneck long before static pages do, and a test that skips them is a test of the wrong site. Tools exist at every budget, from open-source JMeter upward. Find the number where response times bend, because that number, not your hosting plan’s marketing page, is your capacity.

• Cache like the spike is certain. Everything from the caching ladder last summer applies double here. Opcode cache on, full-page caching for anonymous visitors, expires headers set. Campaign visitors are almost all anonymous and almost all looking at the same three pages, which is the perfect caching scenario. A cached landing page can serve ten times more visitors from the same hardware.

• Prepare the static fallback now. The highest-value insurance in the kit: a plain, pre-built HTML version of the landing experience, deployable in minutes if the dynamic stack starts drowning. Serving a simpler page that works beats serving a sophisticated one that does not. The switch is only fast if it was rehearsed.

• Offload the heavy assets. Video, hero imagery, and PDFs are the bytes that multiply fastest under load. They belong on separate infrastructure from the application: storage services like S3, or your host’s static tier. The application server’s job during the spike is applications, not shoveling video.

• Have the hosting conversation before launch week. Your host has seen spikes and has options: temporary upgrades, burst arrangements, or at minimum awareness so their monitoring does not mistake your success for an attack. Calling them in advance versus calling them during is the difference between a plan and an apology.

Watch the write path: it drowns first

A refinement the load tests keep teaching: read traffic and write traffic fail differently, and the write path fails first. Ten thousand people viewing your landing page is a caching problem with a caching solution. Ten thousand people submitting your form is a database-write problem, and writes cannot be cached. Each one is real work at the moment of arrival.

This is where campaign infrastructure actually buckles. The page survives, the form times out, and the visitors you paid most to attract are the ones the failure selects.

Defenses, in order of cheapness. Trim the form to essentials, since every field is write-load as well as friction. Queue the expensive follow-ons, like the confirmation email and the CRM sync, rather than doing them inside the request. And if a third-party vendor hosts your form or donation flow, ask them the capacity question directly and in writing. Their Tuesday sizing is now your launch-day risk, and ‘we’ve never had a problem’ is an anecdote, not an answer.

Instrument the moment itself too. Someone should be watching response times and error rates live during the window, with the authority to act. The difference between a five-minute wobble and a forty-minute outage is usually whether anyone saw the wobble start.

The degradation decision, made in daylight

One more preparation separates the composed from the chaotic: deciding in advance what switches off first. Under real strain, something will give. The only question is whether you choose it or the stack chooses for you.

The choosing version looks like this. The campaign form is sacred. The video becomes a link to YouTube. The related-content blocks go static. The site search goes dark if it must. Write the ladder down, know who is authorized to climb it, and the worst hour of launch day becomes procedure instead of improvisation.

Whose budget is this, anyway?

Here is the framing argument, and the reason this piece is aimed at marketing leaders as much as technical ones: peak capacity is a campaign deliverable. If the media buy is designed to create a surge of attention, then surviving that surge is part of the campaign’s cost, exactly like the creative and the airtime. A percent or two of the total, typically, for testing and hardening that determines whether the other ninety-eight percent pays off.

The failure mode we keep seeing is not technical ignorance. It is organizational. Marketing plans the moment, operations discovers it live, and the invoice for the gap arrives as downtime during the most visible ten minutes of the year.

So put the line item in the campaign budget, and put the meeting on the calendar. Media plan on the table, arithmetic on the whiteboard, playbook assigned. For the readers with February moments, including our higher-ed clients facing decision-notification season, there is still time to run it this cycle.

‘We went down’ remains the most expensive sentence in digital marketing. It is also, with two weeks and a modest checklist, among the most preventable. The teams that treat their best day as an engineering requirement get to experience it as a triumph. The rest get the Monday meeting.