The Signal Chain: an old complaint, finally written down

Not a discovery — a refusal to keep skipping the links nobody wants to measure.

VM

Vadim Matyushkin Behavioral Scientist & Sociologist · ICDS

PublishedOctober 2, 2025

I should say up front: there was no unsolvable problem behind this one, no dead end that forced a breakthrough. What sits behind the Signal Chain is closer to an old complaint that finally got written down properly — the kind of thing a person mutters for years before deciding it deserves a framework instead of a grumble.

01

The first substitution

The complaint is this. Most digital measurement quietly substitutes one thing for another and hopes nobody notices. A platform counts mentions, applies a sentiment classifier, and reports the result as if it had learned what people think and feel about a brand — as if it had looked into something. It has not looked into anything. It has counted words and guessed at their polarity. That is a perfectly legitimate thing to do, but it is not the same claim as “we know the consumer’s attitude,” and treating the two as interchangeable is the first substitution I got tired of watching.

02

The second substitution

The second substitution is subtler, and in some ways worse, because it happens even among people who are careful enough to admit they are only measuring observable behavior. They measure one observable thing — share of voice, a spike in mentions, a jump in branded search — and then hand it universal explanatory power, as if that single indicator were the whole story rather than one link pulled out of a chain nobody bothered to trace. A media peak gets treated as a sales event. A search-volume jump gets treated as a change of heart. Nobody asks whether the signal actually traveled anywhere, whether it reached the next stage at all, or simply lit up and died where it started.

03

The evidence problem

And then there is the evidentiary habit that bothers me most: the one or two stories. A campaign goes viral, sales follow, and the case becomes a slide, then a rule of thumb, then something people cite as if it were a law of consumer behavior. Or the opposite happens — a brand crisis, a mention spike, no visible sales damage — and the case becomes proof that “online reputation doesn’t matter.” Both conclusions are drawn from an N of one. You cannot build a model out of a story. You can build a story out of a model, and only after the model has been tested across enough cases to have earned the right to explain any single one of them.

04

Five links, not two

So the Signal Chain, honestly, is not a discovery. It is a refusal to keep doing what I have just described. If a media event is going to be credited with a purchase decision, then something has to be shown to have happened at every stage in between — not assumed, shown. That meant naming the stages explicitly rather than leaving them as an unexamined black box between “event” and “outcome”: the media peak itself, the informational search it may or may not trigger, the shift in attitude that search may or may not reflect, the transactional search that attitude may or may not produce, and finally the purchase that all of it may or may not lead to. Five links, not two. And the rule that made the framework worth anything at all: no stage in that causal sequence may be skipped, even when it is skipped in the measurement, because a stage we didn’t measure is not a stage that didn’t happen.

05

An argument, not just a method

This discipline is also a quiet argument against a paradigm, not just a methodology against sloppiness. The dominant habit in reputation and social analytics is implicitly behaviorist: treat the consumer as a stimulus-response mechanism, connect the input to the output, and bracket everything in between as unmeasurable or irrelevant. That bracketing is a choice, not a technical necessity, and it is a convenient one, because behavioral data is cheap and abundant while the cognitive and affective middle is expensive and often proprietary. Convenience is not the same as correctness, and a field that measures only what is easy to measure will eventually mistake the shape of its dataset for the shape of the phenomenon.

06

Where the chain breaks

Being honest about the chain also means being honest about where it currently breaks. The attitudinal checkpoint — the stage sitting between informational search and transactional search, the place where a consumer actually forms or revises an opinion — has no reliable free proxy. Search volume is not attitude. It is a footprint attitude sometimes leaves behind, and sometimes doesn’t. We did not paper over that gap with a stand-in metric dressed up as the real thing. We named it as the primary unmeasured link and said so plainly, which is a less satisfying thing to do than declaring victory, but it is the only honest way to treat a gap you have not yet closed.

07

What it actually took

None of this required an unusual insight. It required refusing three ordinary temptations at once: the temptation to let a behavioral proxy speak for an internal state it cannot access, the temptation to let one measured link stand in for a chain nobody traced, and the temptation to let a single vivid case do the work that only a longitudinal body of cases can do. Take those three temptations away, and what is left is not a clever new theory. It is the ordinary, unglamorous insistence that if you are going to claim a signal moved from a headline to a shopping cart, you should be able to point to it moving, one stage at a time, instead of simply believing it did because the two endpoints happened to line up.