The ICDS Methodological Series has, across its first six articles, systematically constructed and empirically tested a five-stage Signal Chain connecting social media events to consumer purchase decisions. The chain runs from a media peak (Stage 1) through informational search activity (Stage 2), sociological brand health indicators (Stage 3), transactional search activity (Stage 4), and finally to purchase or conversion (Stage 5). Articles 1 through 6 provided foundational correlational evidence for the Stage 1–2 and Stage 4 links: peaks with high expressive intensity reliably predict informational search growth at t0 and, in Cautious product categories, transactional search growth at t+1.

A critical gap remains. Stages 3 and 5 — the attitudinal and behavioural endpoints of the chain — have until now been either absent from empirical analysis or addressed only through indirect behavioral proxies. Article 4 identified this gap as the principal methodological challenge of the research programme and stated explicitly that 'Stage 3 data is not always publicly available and represents the primary measurement gap in the chain.' The present article constitutes the first systematic attempt to fill that gap, within the constraints imposed by open-access data availability.

The research impulse is straightforward. The ICDS programme has documented that certain types of social media presence — specifically, peaks with moderate-to-high expressive intensity — reliably shift consumer search behaviour. But search behaviour is a proxy for attitudinal change, not attitudinal change itself. The question that naturally follows is: does the social media environment around a brand — measured across the full year rather than around individual peaks — translate into shifts in consumer purchase intentions, and ultimately into actual purchase frequency? These are the questions this article addresses.

The answer, as will become clear, is: yes, but not in the way that a simple correlation model would predict. The relationship is real, structurally mediated, category-dependent, and partially invisible to the measurement instruments that are publicly available. Understanding why it is invisible is as valuable as measuring its effect — because the explanation points directly to what future research and commercial measurement systems would need to do differently.

Previous ICDS studies operated at monthly temporal granularity. Peaks were identified as deviations from a rolling baseline, and the dependent variables — informational and transactional search volumes — were measured at t0, t+1, and t+2 relative to peak onset. This design was appropriate for the question being asked: how does a specific media event propagate through the consumer decision process?

The present study asks a different question. We are not asking how a specific event propagates; we are asking how the cumulative social media presence of a brand over an entire year relates to annual attitudinal and purchase outcomes. This requires a shift in the unit of analysis from brand-peak to brand-year, and a corresponding shift in the treatment of the media exposure variable from event-level expressive intensity to annual Share of Voice (SOV).

Annual SOV is calculated as a brand's total social media mention volume in a given year divided by the total mention volume of all brands in the same product category for that year. This metric captures the brand's relative prominence in the category's social media discourse — not the quality or intensity of that presence, but its structural weight.

This is a deliberate simplification. The ICDS programme has established that SOV volume alone is a weaker predictor of consumer response than expressive intensity: a large but emotionally flat media presence generates less downstream activation than a smaller but highly evaluative one. By using annual SOV as the independent variable, we are testing the floor of the media-behaviour relationship — the effect that would remain even if peak quality were ignored. This conservative approach is appropriate for a first empirical pass at the annual-level question.

A non-trivial methodological complication arises from the temporal misalignment between the social media data and the attitudinal data. YouGov FMCG Rankings data for 2024 covers the period April 2023 – March 2024, while the YouScan social media data is aligned to calendar years. This creates a partial-year offset of approximately one quarter for the 2024 Consideration measurement.

In the main analyses, we treat YouGov 2024 Consideration as the outcome for social media activity measured in calendar year 2023 — a one-year lag that is theoretically coherent with the ICDS Signal Chain logic (media exposure precedes attitudinal crystallisation). For Kantar Brand Footprint data, Brand Footprint 2024 reflects purchasing behaviour during calendar year 2023. The temporal alignment is therefore consistent across outcome measures: both attitudinal and behavioural outcomes are indexed to 2023, with social media exposure indexed to 2022 (lag condition) or 2023 (concurrent condition).

The social media data derives from the ICDS FMCG corpus, collected via YouScan for the period January 2022 – December 2024. The corpus covers 71 global brands across three macro-categories — Beverages, Food, and Personal Care — and six markets: United States, United Kingdom, Germany, France, United Arab Emirates, and Kingdom of Saudi Arabia. Total mention volume across the corpus exceeds 900 million mentions.

For this study, the key derived variable is annual Share of Voice, calculated separately for each product macro-category and year. The resulting dataset contains SOV values for all 71 brands across three years.

Consumer purchase consideration data is sourced from YouGov BrandIndex's publicly available FMCG Rankings reports. YouGov defines Consideration as the percentage of adults 18+ who, when next in the market for a purchase in the category, would consider the brand. The question format is: "When you are in the market next to make a purchase, which brands would you consider?" The metric is continuous (expressed as a percentage of the adult population) and collected through daily nationally representative surveys.

Coverage in the public reports is limited to brands that appear in top-10 ranked or top-10 improving lists across the six target markets. This creates significant missingness: of 71 brands in the ICDS corpus, 51 have at least one Consideration data point for 2024, 22 have 2023 data (derived from the 'Previous Score' columns in the improvers tables), and only 9 have 2022 data. The Consideration variable is therefore substantially incomplete, and all analyses are necessarily conducted on subsets.

Actual purchase frequency data is sourced from Kantar's annual Brand Footprint reports, which rank global FMCG brands by Consumer Reach Points (CRP) — a metric combining household penetration and purchase frequency. One CRP represents a single purchase occasion of a brand by any household globally in the measurement year. The CRP metric is therefore a direct behavioural measure, not a stated-intention measure: it captures revealed behaviour extracted from consumer purchase panels across 62 markets covering 76% of the global population.

Because Brand Footprint publishes only the global Top 50 brands, coverage in the ICDS corpus is limited to large internationally distributed brands. Twelve brands from the ICDS corpus appear in the Top 50 global CRP ranking across at least two years; eight have sufficient data for the main analytical comparison.

Before presenting the correlational findings, it is necessary to examine what the YouGov Consideration metric actually measures — and, critically, what it does not. This diagnostic step is essential because the pattern of findings cannot be interpreted without understanding the structural properties of the dependent variable.

The YouGov Consideration question — 'When next in the market for a purchase, which brands would you consider?' — appears deceptively simple. In practice, it conflates three conceptually distinct consumer populations whose responses carry fundamentally different meaning for media researchers.

The first population consists of active new buyers: consumers who are currently evaluating the category and for whom the consideration set is genuinely open. These are the consumers for whom social media exposure could plausibly shift consideration. They are actively constructing or revising their brand attitudes, and media-generated evaluative content is relevant to that process.

The second population consists of habitual repeat purchasers: consumers who already buy the brand and respond 'yes' to the consideration question because they fully intend to continue buying it. Their response carries no information about media-induced attitudinal change; it is an expression of established loyalty. In FMCG categories — where repeat purchase rates are high and the average household maintains a stable portfolio of 55 brands per year — this group is numerically dominant.

The third population consists of aspirational non-purchasers: consumers who include a brand in their stated consideration set but are unlikely to purchase it given their actual purchasing context. Research on consideration set formation shows that consumers systematically report brands they aspire to but never convert — particularly for premium or globally recognised brands.

The aggregate Consideration score as published by YouGov does not distinguish among these three populations. A brand with very high loyalty penetration will show a high Consideration score that is almost entirely driven by the second group — repeat purchasers — and will therefore show almost no sensitivity to media events, regardless of how impactful those events are. The Consideration score under these conditions is a loyalty inventory, not a media response indicator.

This structural prediction is confirmed empirically. Among the 22 brands with Consideration data for both 2023 and 2024, all 22 showed a positive year-on-year change. The mean increase was +2.18 percentage points, with a median of +1.70 pp. The minimum increase was +0.3 pp (Sprite); the maximum was +12.7 pp (Dove).

This near-universal positivity is not a media effect — it is a macroeconomic signal. The YouGov 2024 measurement period (April 2023 – March 2024) followed the peak of the global cost-of-living crisis, during which brand consideration across FMCG categories had suppressed due to trading-down behaviour. The 2024 rebound reflects a general recovery in consumer confidence, not brand-specific media activity. This macro-level covariation inflates all Consideration scores simultaneously and further obscures any brand-level media signal.

These findings suggest that Consideration, as measured in population surveys, is an appropriate metric for tracking long-run brand equity trajectories and macroeconomic consumption climate — but a poor short-to-medium-term indicator of media-driven attitudinal change. The dynamic relevant to media effectiveness is not whether consumers include a brand in their consideration set (a structurally stable property), but whether consumers in the consideration set are moving toward or away from purchase intention: the conversion of consideration into intent.

YouGov publishes a separate Purchase Intent metric — 'Of the brands considered, which are you most likely to purchase?' — that would be theoretically far better suited to capturing media-driven attitudinal dynamics. This metric, however, is not available in open-access reports. The absence of public Purchase Intent data represents the single most consequential data gap for Stage 3 research in the ICDS programme.

The primary hypothesised relationship — annual SOV predicting Consideration with a one-year lag — does not achieve statistical significance at any specification tested:

The non-significant overall relationship is not primarily a failure of the media-attitude model. It reflects three overlapping methodological problems. First, the global SOV variable and the market-specific Consideration variable are measuring different things at different geographic scales — a brand's social media dominance in aggregate does not correspond to its local market salience where Consideration is surveyed. Second, the Consideration metric's inertia properties (described in Section 4) reduce its variance to a level where even a real media effect would be statistically undetectable. Third, the Beverages category shows a negative SOV–Consideration relationship, driven structurally by the mismatch between mega-brand social media dominance (Coca-Cola: SOV 43%, Consideration 32) and strong local-brand consideration in MENA markets (Almarai: SOV near zero, Consideration 57). This reversal suppresses the global coefficient.

The Beverages result deserves closer attention because it is not an anomaly — it is a theoretically coherent finding that reveals a deeper structural property of the FMCG media landscape.

In the Beverages category, the five highest-Consideration brands in the sample are all regional market leaders: Almarai (UAE/KSA, Consideration 57), Al Ain (UAE, 35), Robinsons (UK, 34), Mai Dubai (UAE, 33), and Gatorade (US, 32). None of these brands is among the top-5 SOV holders globally. Conversely, Coca-Cola — with SOV of 43% and global social media dominance — has a Consideration score of 32 in the US: lower than Gatorade (32), equal to several regional brands, and far below its global media penetration would suggest.

This pattern reflects a well-established FMCG reality: global share of voice does not translate into local purchase consideration. A brand's social media presence is distributed across markets, languages, and cultural contexts; Consideration is measured within specific market populations with specific purchasing habits. For global mega-brands, the relationship between global SOV and local Consideration is inherently weak because the majority of social media conversations about Coca-Cola occur in markets, demographics, and contexts that are irrelevant to a given local consumer's consideration set.

This structural mismatch has an important implication for methodology: testing the SOV → Consideration relationship requires market-level SOV, not global SOV. The present study, constrained to publicly available global mention data, cannot construct this variable. The correct test of the hypothesis therefore remains to be performed on market-stratified data.

The relationship between annual SOV and actual purchase frequency, measured through Kantar Brand Footprint CRP, shows a more encouraging pattern — despite the severely limited sample size of N = 8 brands with complete data for both variables.

Six of eight brands show directional agreement between SOV movement and CRP movement. The Spearman correlation between SOV_2022 and CRP_2023 is r = 0.55 (p = 0.16), consistent in sign and magnitude with the theoretical prediction but not statistically significant at a sample of eight. For CRP growth (%), the correlation with ΔSOV is r = 0.33 (p = 0.42), weaker, as expected from a scaled metric that conflates brand-size effects.

Red Bull: Falling SOV, Rising Purchases. Red Bull's global social media Share of Voice declined 7.4 percentage points between 2022 and 2023 — the largest SOV decline in the Beverages sample. Its CRP simultaneously grew by 17.8%, the fastest growth of any brand in the sample. This apparent contradiction resolves when examining the channel structure of Red Bull's media presence: the brand's Formula 1 association, innovation in product line (Organics variants), and market expansion in China, Germany, and Brazil generated purchase growth through channels — broadcast media, sports sponsorship, and in-store activation — that are not captured in social media mention volume. Red Bull's case illustrates that SOV is a social media indicator, not a total media indicator, and that for brands whose growth strategy centres on non-social channels, the SOV–purchase relationship can be systematically attenuated or even reversed.

Colgate: Stable Large Brand Immunity. Colgate's SOV declined marginally (−0.8 pp) while its CRP grew 4.4%. This is consistent with what the FMCG literature identifies as inertia-protected growth: for brands with 55%+ global household penetration, short-run fluctuations in social media presence are absorbed by the existing loyalty base without materially affecting purchase frequency. This is the CRP equivalent of the Consideration inertia described in Section 4. For brands at the apex of their category's penetration curve, media activity may matter more for defending existing buyers than for recruiting new ones — a dynamic that short-run correlation analysis cannot detect.

Dove presents the most analytically interesting case in the combined dataset. Its SOV increased modestly (+1.2 pp), its CRP grew 6.1%, and its Consideration score jumped from 32.7 to 45.4 — an increase of +12.7 percentage points, the largest in the entire 22-brand Consideration sample by a substantial margin.

This combination — moderate SOV growth, moderate CRP growth, but large Consideration growth — is the signature pattern that the ICDS Signal Chain predicts for successful attitudinal campaigns in Cautious categories. Personal Care is a Cautious category (significant psychosocial purchase risk, brand reputation matters). Dove's 2023 communications — centred on the Body Confidence initiative and the Real Beauty campaign evolution — were explicitly designed to shift brand attitudes rather than merely drive purchase frequency.

The Dove case provides early-stage evidence that, when attitudinal campaigns are effective, Consideration can move substantially within a one-year window, even against the general backdrop of inertia that characterises most FMCG brands. The case also demonstrates that this attitudinal shift may lead rather than follow purchase change: Dove's Consideration grew by 12.7 pp while its CRP grew only 6.1% — a ratio suggesting that much of the new consideration had not yet converted to purchase by the end of the measurement period.

The ICDS Signal Chain specifies five sequential stages from media event to purchase. The present study has now generated empirical evidence, of varying quality, for all five stages. It is instructive to assess the evidential status of each link:

The pattern that emerges is internally coherent. The strongest links are at the endpoints where behavioral data (search, purchase) is available; the weakest links involve the attitudinal middle of the chain where only survey data is accessible, and that survey data has structural limitations. This is precisely the profile the ICDS programme predicted: Stage 3 is the hardest to measure, not because the relationship doesn't exist, but because the available instruments — Consideration scores aggregated annually across population cohorts — are too blunt to capture the media-reactive dynamics that the theory specifies.

The present study permits a precise specification of what data architecture would be required to test the Stage 3 hypothesis properly. Four requirements emerge from the analysis:

Global SOV is not a valid predictor of market-specific Consideration. Future studies must compute SOV within each market (country-level mention volume / country-level category mention volume) and match it to country-level Consideration scores. This requires either national-level social listening data or a multi-market social listening platform capable of geo-stratified export. The ICDS YouScan corpus contains market-level mentions for all six target markets; the SOV calculation must be rerun at market level before the core hypothesis can be properly tested.

Consideration is too inertia-contaminated by habitual purchasers to serve as a sensitive media-response indicator. YouGov's Purchase Intent metric — which isolates the choice among already-considered brands — is theoretically superior and empirically more appropriate. This metric is published only in the paid YouGov BrandIndex platform. A collaboration agreement with YouGov for access to historical Purchase Intent data for the ICDS corpus brands would eliminate the most critical measurement gap in the programme.

Annual aggregation of both the media variable (SOV) and the outcome variable (Consideration) destroys the temporal information that the Signal Chain model requires. The lag structure between media exposure and attitudinal response — estimated at one to three months based on the search-behaviour findings — is invisible at annual granularity. Monthly survey data is expensive but available through YouGov BrandIndex API access. Quarterly data is potentially obtainable from the public 'improvers' tables if regional reports are collected systematically across all four quarterly reporting periods.

The current analysis is cross-sectional: it compares brands to each other rather than tracking individual brands over time. A panel design — tracking the same brands' SOV and Consideration month by month — would eliminate the confounding by cross-brand structural differences (market size, category type, distribution breadth) that inflates residual variance in cross-sectional models. The YouScan data infrastructure already supports monthly panel tracking for all 71 corpus brands. The bottleneck is the Consideration data, which needs to be sourced at monthly frequency.

Data availability constraints. The most significant limitation of this study is not analytical but infrastructural: the attitudinal and purchase outcome data available in open-access public sources is insufficient for the statistical power required by the hypotheses being tested. The YouGov Consideration sample (N = 51 for the primary analysis, N = 8 for the CRP pilot) lies below the threshold for robust regression inference. The findings presented here should be treated as directional evidence, not confirmatory results.

Geographic and temporal misalignment. The YouGov Consideration data and the YouScan social media data are not perfectly temporally aligned (the YouGov 2024 report covers April 2023 – March 2024, not the calendar year). The SOV measure is globally aggregated while Consideration is market-specific. These misalignments introduce systematic noise into all correlation analyses, biasing estimates toward zero.

Confounding by category structure. The Beverages category reversal illustrates that cross-category pooling introduces structural confounders — specifically, the global-versus-local brand composition of each category in different markets. All category-pooled results must be interpreted with this caveat in mind.

CRP pilot scope. The N = 8 pilot analysis of the SOV–CRP relationship is not powered for inference. Its value is illustrative: it demonstrates that the data can be collected and that the directional signal is consistent with the hypothesis. A properly powered version of this analysis would require 25–30 brands with complete SOV and CRP data, which would necessitate extending the corpus to include smaller brands not currently tracked in the Brand Footprint global Top 50.

Causal identification. All analyses in this article are correlational. Brands with high SOV may also invest more in advertising, distribution, and product innovation — creating multiple potential explanatory channels for Consideration and CRP growth. The temporal lag structure is consistent with, but does not establish, a causal interpretation.

This article has pursued three objectives: to extend the ICDS Signal Chain empirically from search behaviour to attitudinal and purchase outcomes; to diagnose why the publicly available attitudinal instrument (YouGov Consideration Score) shows limited sensitivity to annual social media presence; and to provide early-stage evidence on the SOV–actual purchase relationship through a pilot analysis of Kantar Brand Footprint CRP data.

The principal substantive finding is a null result with a theoretically coherent explanation. Annual global SOV does not significantly predict market-specific Consideration scores, but this reflects the structural properties of the Consideration metric and the geographic mismatch between the variables, not an absence of media–attitude linkage. When the unit of analysis shifts from the attitudinal to the behavioural — Consideration replaced by actual purchase frequency — the relationship becomes directionally positive (r = 0.55) and is consistent with the theoretical model in six of eight cases.

The most consequential substantive finding is the Consideration inertia result: all 22 brands with multi-year data showed positive Consideration trends, and the dominant driver of this trend is macroeconomic (post-inflation recovery) rather than brand-specific. This finding has a direct practical implication: brands that use annual Consideration surveys as primary media effectiveness indicators are measuring the wrong variable for the wrong question. Consideration is an equity stock indicator; media effectiveness is a flow phenomenon that requires higher-frequency, more precisely targeted measurement instruments.

The most analytically important finding is the Dove case. A brand whose attitudinal campaign succeeded on its intended audience (Personal Care, Cautious category) showed a Consideration increase of +12.7 pp against a CRP increase of +6.1% — a ratio consistent with the ICDS model prediction that attitudinal change leads purchase change, and that the conversion lag is real. This is a small-N illustration of the dynamic the entire programme is designed to detect at scale.

The research agenda emerging from this article is clear. Stage 3 of the ICDS Signal Chain requires measurement instruments that are (a) market-stratified, (b) monthly or quarterly in granularity, (c) focused on Purchase Intent rather than Consideration, and (d) tracked on a brand-panel basis rather than cross-sectional snapshot. The analytical infrastructure developed in this programme can immediately accommodate such data once it becomes available. The current bottleneck is not methodology — it is access.