{{title}}
{{textarea}}
{{task}}
{{select}}
{{team}}
{{input}}
{{solution}}
{{result}}
ICDS Research Brief Methodology
Beyond Description: How ICDS Moves from Observing Media Events to Explaining Their Influence on Human Behaviour
Standard media analytics records what happens. ICDS explains why — and what will happen next. Here is what the difference is, how it works, and why it changes how organisations should make decisions based on media activity data.
VM
Vadim Matyushkin Behavioral Scientist & Sociologist · ICDS
Imagine your brand has just come through a major media peak. One million mentions in three days. Sentiment: 74% positive. Reach — the highest of the year. The marketing team reports success. Three months later, the sales team cannot explain why sales have not grown.
This is not an exception. It is the norm for organisations using standard media monitoring. And the reason is straightforward: standard analytics describes what happens to the media field. It does not explain why some peaks convert into sales and others do not. And it does not predict what will happen to human behaviour after the peak has passed.
This gap — between description and explanation — is the central problem that the ICDS methodology addresses.
2 yearsResearch programme66 FMCG brands · 26 countries · 10 categories · consumers and investors
5 stagesIn the Signal ChainFrom media event to purchase — each measurable, each with its own indicator
2,200+Publications reviewedNone replicates the tri-metric ICDS architecture (media + search + finance)
01
The Blind Spot of Standard Analytics
Contemporary media analytics works professionally with the two endpoints of the consumer journey. At the input end it records media events: publication volumes, sentiment distributions, reach. At the output end — commercial outcomes: sales, conversions, customer acquisition. Everything between these two points it systematically fails to see.
This is not a technical limitation — it is a conceptual choice. The dominant paradigm in media analytics is implicitly behaviourist: it connects stimuli to responses while bracketing the mental processes between them. This approach is convenient because behavioural data is cheap and accessible. But it generates systematic errors in attribution, timing, and strategic inference.
| What standard monitoring measures | What it does not see | What ICDS adds |
|---|---|---|
| Mention volume (mentions, reach) | Expressive intensity — the share of evaluative mentions within the peak | Six-level taxonomy of expressive intensity + optimal conversion corridor 10–39% |
| Overall sentiment (% positive/negative) | The structure of the evaluative field — the ratio of positive to negative within the expressive subset | Balance of Expression: four levels, each predicting behaviour differently across categories |
| Response at the peak (t0) | The delayed effect at t+1 — where the real purchase decision sits in Cautious categories | Chain temporal profile: t0 = cognitive activation, t+1 = behavioural intent |
| Whether a peak occurred (yes/no) | Why the peak occurred and which of the five trigger types caused it | Trigger typology + content source analysis (organic vs. paid) |
| Overall brand search index | The difference between informational (Stage 2) and transactional (Stage 4) search | Intent-based query split: ‘[brand] reviews’ vs. ‘[brand] buy’ |
The key consequence: organisations using only standard monitoring regularly make two errors. First — declaring a media event ineffective because they cannot see its effect at t+1 (the delayed response in Cautious categories). Second — attributing sales growth to the most recent visible media event, when the real trigger may have been a peak from three weeks earlier.
02
The Signal Chain: Five Stages from Event to Action
The central methodological concept of ICDS is the Signal Chain: a five-stage causal chain from media event to purchase. Each stage has its own observable indicator and its own measurement instrument. The chain is grounded in the three-component model of attitude (Howard & Sheth, 1969; Rosenberg & Hovland, 1960): every media peak that reaches a person potentially changes their affective, cognitive, and behavioural orientation toward the brand.
| # | Stage | Indicator | Data source | Status in ICDS programme |
|---|---|---|---|---|
| 1 | Media peak | Mention volume, expressive intensity, Balance of Expression | YouScan, Brandwatch, GDELT | Confirmed (N=66–85). Optimal expressive intensity corridor 10–39% established. |
| 2 | Informational search | Growth in [brand] + informational modifier queries | Google Trends, KeywordTool Pro | Confirmed. Growth at t0, brand-specific, durable in Cautious categories. |
| 3 | Attitude change | Spontaneous awareness, brand attitude, purchase intention | YouGov BrandIndex, Kantar BrandZ | Gap. Instruments insufficiently sensitive to media-reactive changes. |
| 4 | Transactional search | Growth in [brand] + transactional modifier queries | Google Trends, KeywordTool Pro | Confirmed in Cautious (ρ=+0.75, N=18). Effect at t+1, not t0. |
| 5 | Purchase / conversion | Sales, market share, website conversion | Nielsen/Kantar (proprietary), SimilarWeb | Pilot only (N=8, r=0.55). Requires data provider partnership. |
The key unexplained fact
At t0 — the moment of the media peak — the commercial effect is often invisible. The business looks at the data and concludes the campaign did not work. In fact, in Cautious categories t0 is the moment of cognitive activation, not transaction. The buyer begins to gather information. Transactional search rises 4–8 weeks later. Companies unaware of this lag systematically underestimate the effectiveness of their media activity.
03
Three Variables That Explain What Description Cannot
1
Expressive intensity — not volume, but the share of evaluative mentions.
96.5% of all brand mentions are neutral. They evaluate nothing. Only 3.5% are evaluative (expressive). It is this share — not the total peak volume — that determines whether the event activates a cognitive process in the consumer. Optimal corridor: 10–39% expressive intensity. Below this, the peak is too informational and generates no evaluative engagement. Above, it overheats and is perceived as provocation or crisis. Established on N=66–85 brands; replicates across FMCG and fintech.
2
Balance of Expression — the structure of the evaluative field, not its level.
Standard sentiment says: ‘70% positive.’ Balance of Expression asks: what is the ratio of positive to negative within the evaluative subset — and how does it interact with the category type? In Cautious categories the optimum is 75–90% positive in evaluative mentions (Strong Positive Disbalance). Higher — the environment is perceived as inauthentic. Lower — no clear signal for the central route. Correlation with transactional search at t+1: ρ = +0.75 (p < 0.001). In Spontaneous categories it does not work at all. Different categories, different rules.
3
The chain’s temporal profile — exactly when to look for the effect.
Standard monitoring looks at the effect at the moment of the peak (t0) or aggregates over a period. ICDS introduces three time points: t0 (moment of peak), t+1 (month after), t+2 (two months after). For Spontaneous categories: peak search interest at t0, rapid decay. For Cautious: informational search grows at t0, transactional search at t+1, and in 72% of brands the effect continues at t+2. This is not a delay — it is a structural property of the central route: the buyer thinks first, then searches for where to buy.
04
The Signal Chain Works for More Than Consumers
A defining feature of the ICDS architecture is that the same Signal Chain, with the same input variables, describes the behaviour of two completely different audiences: consumers and investors.
For the consumer
media peak → informational search → attitude change → transactional search → purchase
Time horizon: 4–12 weeks.
For the investor
media peak → Tonality Jump → algorithmic positioning → price volatility → price correction
Time horizon: days to weeks.
The mechanisms differ, the audiences differ, the time horizons differ — but the input variable is the same: the structure of the media peak. This is not coincidental. It means that a media peak carries information simultaneously relevant to the consumer decision and to the financial valuation of the brand. This is precisely why we call it a signal, not an event.
TDI — a candidate for a universal ICDS law
Tonality Jump (TDI) — the change in the tonal index from t−1 to t0 — is the only statistically significant predictor of stock volatility (ρ=−0.337, p=0.014, N=53). The same coefficient sign replicates on the FMCG corpus (N=66) and the AI-sector corpus (N=10, ARPI). Three independent corpora — three markets — one pattern. A candidate for a cross-sector universal law of media dynamics.
05
Why This Is Original: What Nobody Else Does
Independent verification across 2,200+ publications from 2022–2025 (Long & Guo, 2025; Future Business Journal, 2025) showed that none of the three adjacent research streams replicates the tri-metric ICDS architecture. ICDS occupies the space between them: it takes real media data at market scale (from Social Listening), applies attitude theory and central/peripheral processing routes (from behavioural economics), and closes the chain onto measurable financial outcomes — both consumer and investor (from financial analytics). This space was unoccupied before 2022.
| Research stream | What they do | What they do not do |
|---|---|---|
| Media analytics / Social Listening | Mention monitoring, sentiment, reach, Share of Voice | Do not construct a causal chain to behaviour. They describe the media field, not its influence. |
| Behavioural economics / Consumer Research | Persuasion experiments, attitude theory, purchase decision | Do not work with real media data at scale. Laboratory conditions, not market dynamics. |
| Financial analytics / Sentiment Trading | Sentiment scores for stock forecasting, event studies | Do not connect the financial effect to consumer behaviour through a single causal model. |
06
ARPI: When Explanation Becomes Prediction
The Signal Chain explains the past: why a media event did (or did not) convert into behavioural change. ARPI — Aggregate Risk-Predictive Index — takes the next step: using the same structural variables to predict future price movement.
ARPI = normalize(TDI×2.0 + IAI×1.8 + LPD×2.2 + AS×1.5 + FDI×1.8) × SBI | Metric | Weight | What it measures | Basis |
|---|---|---|---|
| TDITonality Jump | ×2.0 | Tonality change: 14-day rolling average vs. 90-day baseline | ρ=−0.337, p=0.014, N=53 |
| IAIInformation Asymmetry | ×1.8 | ‘Quiet news + positive surprise’ — harbinger of L-pattern sustained decline | NVDA Q4 + GPT-4.5: both L-pattern at IAI > 0.6 |
| LPDL-Pattern Detector | ×2.2 | Neutral immediate reaction + tonality deterioration = quiet disappointment | 2/2 accurate triggers on N=10, highest weight |
| ASAmbiguity Score | ×1.5 | Share of negative content in the 45–65% zone — maximum speculative volatility | Robinhood: 6.3% volatility vs. 2% in unambiguous environment |
| FDIFrontier Decoupling | ×1.8 | Tonality of private AI companies minus public ones — first proxy for private-company risk | New instrument, cross-corpus validation |
| SBISemantic Bleed (deflator) | ×0.7 | Share of irrelevant narrative (name overlaps with pop culture) — deflates the signal | Stargate/Grok anomalies in GDELT |
ARPI is currently undergoing prospective verification. Retrospective calibration on N=10 events (December 2024 – June 2025) demonstrated the system’s capacity to lead market consensus by 10–13 months in the AI sector. The next step is to log signal dates in real time and compare against actual outcomes.
Methodology note
How the study was built
The research is based on Articles 4, 5, 6, 7, and 8 of the ICDS Methodological Series and the ICDS ARPI v4 document. Signal Chain: theoretical development and primary empirics 2023–2025. Corpora: FMCG N=66 (26 countries, 10 categories, YouScan + Google Trends); Fintech N=12–53 (public companies, Yahoo Finance); AI sector N=10 events (GDELT + Yahoo Finance / yfinance). Originality verified across 2,200+ publications (Long & Guo, 2025; Future Business Journal, 2025). ARPI was calibrated retrospectively — prospective verification is underway.
Full research articles available at icds.institute.
VM
About the author
Vadim Matyushkin
Behavioral Scientist & Sociologist · ICDS
Vadim Matyushkin is a psychologist and sociologist with close to twenty years of research into digital behavior, trust, and information dynamics, and a researcher behind the Institute of Communication and Data Science (ICDS). His work has supported organizations including Coca-Cola, PepsiCo, Mars, Danone, Nestlé, and Bayer in moving from self-reported survey data toward direct behavioral evidence.
ICDS — Institute of Communication and Data Science is an independent research institute focused on understanding how trust, behavior, and reputation are formed in digital environments. ICDS operates as an intellectual institution — not an agency, not a platform, and not an educational provider.