After the Buzz: What Did Anyone Actually Do?

Most brand storms blow over without a trace in the numbers. Two ICDS studies suggest how to tell, before sales move, which noise turns into behavior.

VM

Vadim Matyushkin Behavioral Scientist & Sociologist · ICDS

PublishedNovember 4, 2026

Anyone who watches brands online has seen it. A campaign is praised everywhere for a week. Another brand is cursed, and a call to boycott it trends. Then the conversation moves on, and the sales line barely bends. The boom does not arrive, and neither does the boycott.

This happens often enough that many marketers learn to shrug at online noise. The shrug has a cost, because some storms do turn into behavior: a wave of new customers, or a quiet loss of old ones. By the time either shows up in quarterly sales, the moment to act has passed.

Two questions follow. Can you tell whether the noise had an effect before it reaches sales? And is there a pattern to which noise turns into behavior and which simply fades?

Why the Usual Report Cannot Tell

The standard monitoring report starts with a count: how many times the brand was mentioned, and what share of those mentions were positive or negative. The count answers one question well, how much was said. It treats every mention as equal, whether fifty people saw it or fifty million.

That equal weighting can make the count point the wrong way. In an ICDS study of seven fintech companies over 24 months, we computed the balance of positive and negative mentions twice: once by counting mentions, once by weighting each mention by the audience it reached. The two readings disagreed on direction in 5 to 22 of the 24 months, depending on the company. For Interactive Brokers they disagreed in 22 months out of 24.

Nubank in June 2024 shows how this happens. There were about five negative mentions for every positive one, so by count the brand was having a bad month. Weighted by audience, the balance was positive. The complaints came from individual customers with small followings; the coverage that most people actually saw came from large outlets writing about the company’s growth. (We described a matched pair of such cases in an earlier note, “Reach vs Mentions: The Pair That Would Not Agree”.)

Even when the count points the right way, it measures talk. Whether anyone acted on that talk is a separate question, and the report does not ask it.

The Earliest Trace of Action Is a Search

Before people buy, they look things up. Search is the first place where talk about a brand turns into something a person does, and it shows up in the data in the same month as the peak or the month after, long before quarterly sales. Search also comes in two kinds that mark two different steps:

Searches for the brand’s own site, login or account page are left out of both. They come from people who are already customers, so they say nothing about whether the peak moved anyone new.

This is the idea behind the Media-Driven Brand Impact Index (MDBI), an ICDS method in development. It treats each surge of mentions as a small natural experiment: something happened, and we watch whether the two kinds of search rose in the weeks after it.

Whose Action Are You Waiting For?

The fintech study gives a second answer that matters as much as the first. The two ways of reading a peak, by count and by audience, each turned out to match a different kind of response.

We took seven peaks for four brokerage platforms whose outcomes were already recorded. Market and financial results went with the audience-weighted reading: in every case where the data existed, the company reported growth and its shares were up, while the count-based reading looked negative or flat.

Buyers’ search went with the count. The clearest case is Coinbase in November 2024. Searches of the looking-into-it kind ran about two standard deviations above the sector norm in the peak month, and searches of the getting-ready-to-buy kind stayed high for the next two months. By count, that month’s conversation was almost evenly split. By audience, it was strongly positive, a reading that would have predicted a quiet response.

So the same peak can move two groups through two channels. Investors respond to how many people saw the story. Potential customers respond to how many people around them are talking about it. Before judging whether a storm mattered, decide whose action you are waiting for, and read the peak the way that group experiences it.

Four Outcomes After a Peak

Put the two kinds of search side by side, and every peak falls into one of four cells. This is the pattern behind which noise turns into behavior.

Timing Depends on the Category

How long interest lasts after a peak differs by category. In considered purchases, such as phones or home appliances, most of the looking-into-it interest is still there a month later. In impulse categories, such as celebrity cosmetics, it has largely faded by then.

So the same rule cannot judge both. An impulse brand should be judged by getting-ready-to-buy searches in the weeks right after a peak. A considered-purchase brand should be judged by whether the looking-into-it interest holds.

Three Questions for Your Next Report

What this rests on

Evidence & limits

The fintech findings come from seven companies over two years, and the search result rests on one strong case. Audience figures are gross: a person reached twice is counted twice. The MDBI method is calibrated on a pilot of 73 peaks across 33 brands; its weights will change as the corpus grows, and brand-by-brand results have not been published yet. Taken together, the findings show a direction worth testing on your own brand.

Sources. ICDS, When a Mention Is Not a Unit of Impact (methodological series). ICDS, Measuring the Effectiveness of Brand Mediality (MDBI working paper). ICDS Insights, Reach vs Mentions: The Pair That Would Not Agree.