Menu
RU EN AR
Monitoring system for the national consumer goods market in the event of supply chain disruptions: metrics, indices, aggregators

Monitoring system for the national consumer goods market in the event of supply chain disruptions: metrics, indices, aggregators

A formalized system of indicators for monitoring the national consumer goods market in the context of supply chain destabilization is proposed.

 

A formalized metric system for monitoring the national consumer goods market in the face of supply chain disruption is proposed. The system is built on the infrastructure of an intelligent product catalog with fuzzy classification and substitution/analog graphs. A hierarchy of twelve metrics is defined, organized into four levels of aggregation (product → category → basket → market) and three dimensions of analysis (price, availability, quality of substitution). Each metric is provided with a precise definition, calculation formula, value range, interpretation, and alert thresholds. A composite leading indicator is formalized, enabling the detection of price pressure 5–14 days before its manifestation in retail prices. Applicability is examined using the example of the supply chain crisis in the UAE (March 2026).

Traditional government consumer price monitoring systems are based on tracking a fixed set of marker products across a representative sample of retail outlets. This approach has proven effective under steady-state conditions, but it exhibits three structural limitations when supply chains are disrupted.

Introduction

01

The empty reference problem

When a marker product disappears from the shelf, the system loses its price signal. In the standard methodology, the consumer price index (CPI) accounts for missing observations through imputation based on class averages, which masks the true dynamics during a crisis.

02

Blindness to product range degradation

Milk prices may remain stable, but consumers are forced to switch from organic to regular, from A-brand to private label, from 1 liter to 500 ml. Traditional monitoring doesn't detect this type of hidden inflation.

03

Lag

Price changes are a lagging indicator. Product range reductions, increased forced substitutions, and longer delivery times are all observed 5-14 days before retail prices react.

This paper proposes a metrics system that addresses these limitations. The system is designed to complement (not replace) the existing government monitoring infrastructure and is built on the AIHiveLab AI platform with its MinaMart field laboratory (UAE), which collects and analyzes market data across the UAE.

 

2. Monitoring Infrastructure

2.1. Master Product Catalog

The system is based on the Master Product Catalog (MPC)—a normalized product directory P = {p₁, ..., p_N}, not tied to a specific retail outlet. Each product is described by a vector of attributes:

p = (id, name, brand, brand_class, cat, pack_size, unit, storage, tier, diet_tags, risk_tags) (1)

The key derived attribute is the normalized unit price:

up(p, t) = price(p, t) / pack_size(p) [AED/kg, AED/L, AED/unit] (2)

Normalization ensures comparability of products with different packaging volumes and allows for the detection of shrink inflation.

 

2.2. Fuzzy Membership

The set of product groups G = {g₁, ..., g_m} includes categories (dairy, cereals), consumption rituals (breakfast, dinner), dietary profiles (halal, vegan), and price segments (good/better/best). For each group g, a fuzzy membership function is defined:

μ_g : P → [0, 1]                                                                                                                     (3)

where μ_g(p) = 0 — the product does not belong to the group, μ_g(p) = 1 — a perfect match. Intermediate values ​​allow a product to belong to several groups simultaneously (yogurt: μ_{dairy} = 0.9, μ_{dessert} = 0.4) and ensure smooth aggregation of metrics across intersecting sets.

 

2.3. Substitution and Price Analog Graphs

Two directed graphs are defined on top of the catalog. The functional substitution graph G_S = (P, E_S) links products that can substitute each other under OOS conditions. Each edge (p, q) is characterized by similarity s(p, q) ∈ [0, 1], type (strict/soft), and constraints. The price analog graph G_A = (P, E_A) is stored separately and enables price comparisons—the requirements for analogs are stricter (matching units, similar tiers, and brand_classes). Separating the graphs prevents false price comparisons.

 

2.4. Data Collection

The system continuously collects data on retail prices, product availability, packaging characteristics, and assortment composition throughout the UAE. For each observation, the following is recorded: product p, point of sale s, timestamp t, price(p, s, t), availability avail(p, s, t) ∈ {0, 1}, package size pack(p, s, t).

The metrics are organized into three dimensions (price, availability, quality of substitution) and four levels of aggregation (product, category, basket, market). For each metric, the following information is provided: definition, formula, value range, interpretation, and recommended alert thresholds.

Metrics system: definitions and calculations

 

3.1. The Price Dimension

3.1.1. Price Change Index (ΔP)

Definition: The relative change in the normalized unit price of a product $p$ over a period $\tau$.

$$
\Delta P(p, t, \tau) = \frac{up(p, t) - up(p, t-\tau)}{up(p, t-\tau)} \tag{4}
$$

Range: $(-1, +\infty)$. Value: $0$ — price unchanged; $0.1$ — increase by 10%; $-0.05$ — decrease by 5%.

Calculation. If a product $p$ is observed at multiple points of sale $S_p$, the unit price is aggregated as the median:

$$
up(p, t) = \operatorname{Median}_{s \in S_p} \left\{ \frac{price(p, s, t)}{pack(p, s, t)} \right\} \tag{5}
$$

The median is preferable to the mean, as it is robust to outliers (single price anomalies at individual points).

Alarm threshold: $|\Delta P| > 0.05$ over 7 days — yellow; $|\Delta P| > 0.10$ over 7 days — red.

 

3.1.2. Price Pressure Index (PSI)

Definition: Weighted average relative change in unit price for products in group $g$, with fuzzy membership weights.

$$
PSI_g(t, \tau) =
\frac{\sum_{p \in P_g^{avail}} \mu_g(p) \cdot \Delta P(p, t, \tau)}
{\sum_{p \in P_g^{avail}} \mu_g(p)} \tag{6}
$$

where $P_g^{avail}$ is the set of goods in group $g$ available at time $t$. The weights are $\mu_g(p)$, ensuring the correct contribution of partially owned goods.

Range: $(-1, +\infty)$. $PSI = 0$ — prices are stable; $PSI = 0.08$ — average growth of 8%.

Calculation. For each good $p$: (a) aggregate prices by points using the median from formula $(5)$; (b) calculate $\Delta P$ using formula $(4)$; (c) weight by $\mu_g$ and sum. Products unavailable at time $t$ are excluded from the sum (their contribution is reflected by the ASI, not the PSI).

Alarm threshold: $PSI > 0.03$ over 7 days — yellow; $PSI > 0.07$ — red. For the "essential products" group (9 UAE categories), the threshold is lower: $0.02$ and $0.05$, respectively.

 

3.1.3. Shrinkflation Index (SFI)

Definition. The proportion of products in group $g$ that exhibit a decreasing pack size with a stable or increasing absolute price.

$$
SFI_g(t, \tau) =
\frac{
\left| \left\{ p \in P_g^{avail} : pack(p,t) < pack(p,t-\tau) \wedge price(p,t) \geq price(p,t-\tau) \right\} \right|
}{
|P_g^{avail}|
} \tag{7}
$$

Range: $[0, 1]$. $SFI = 0$ — shrinkflation is not detected; $SFI = 0.15$ — 15% of the products in the group are affected.

Calculation. For each product $p$, $pack\_size(p, s, t)$ and $pack\_size(p, s, t-\tau)$ are compared at the same points of sale. If the product is available at multiple points of sale, majority voting is used (a change is recorded if observed at $\geq 50\%$ of the points).

Alert threshold: $SFI > 0.05$ — yellow; $SFI > 0.10$ — red.

The SFI detects hidden inflation, invisible to standard price monitoring: the price is stable, but the consumer receives a smaller quantity of the product.

 

3.1.4. Price Volatility Index (PVI)

Definition. The coefficient of variation of the daily unit price values of product $p$ over a sliding window of $T$ days.

$$
PVI(p, t, T) =
\frac{
\operatorname{StdDev}(\{up(p, \tau) : \tau \in [t-T, t]\})
}{
\operatorname{Mean}(\{up(p, \tau) : \tau \in [t-T, t]\})
} \tag{8}
$$

Range: $[0, +\infty)$. $PVI \approx 0$ — price stability; $PVI > 0.10$ — high price volatility. Recommended $T = 14$ days.

Group aggregation:

$$
PVI_g(t) = \frac{\sum_p \mu_g(p)\, PVI(p,t)}{\sum_p \mu_g(p)}
$$

Alert threshold: $PVI_g > 0.08$ — yellow (atypical instability); $PVI_g > 0.15$ — red.

 

3.2. Availability Dimension

 

3.2.1. Product Availability Index (AI)

Definition: The proportion of retail outlets where product $p$ is available at time $t$.

$$
AI(p, t) = \frac{|\{s \in S : avail(p, s, t) = 1\}|}{|S|} \tag{9}
$$

Range: $[0, 1]$. $AI = 1$ — product is available everywhere; $AI = 0$ — product is unavailable at all outlets.

3.2.2. Category Availability Stress Index (ASI)

Definition: The proportion of group "mass" $g$ lost due to product unavailability.

$$
ASI_g(t) = 1 - \frac{\sum_p \mu_g(p) \cdot AI(p, t)}{\sum_p \mu_g(p)} \tag{10}
$$

Range: $[0, 1]$. $ASI = 0$ — full availability; $ASI = 0.3$ — 30% of the category's product range is unavailable.

Calculation. For each product $p$: calculate $AI(p, t)$ using formula $(9)$, then aggregate with the weights $\mu_g(p)$ using formula $(10)$. Fuzzy membership ensures that a product partially belonging to a group contributes proportionally.

Alert threshold: $ASI > 0.10$ — yellow; $ASI > 0.25$ — red.

 

3.2.3. Assortment Diversity Index (ADI)

Definition. Normalized Shannon entropy of the distribution of brands/tiers across available products in a group. Measures the diversity of the offering, not the number of products.

$$
H_g(t) = -\sum_k q_k \cdot \ln q_k
$$

where

$$
q_k = \frac{\sum_{p \in P_{g,k}^{avail}} \mu_g(p)}{\sum_{p \in P_g^{avail}} \mu_g(p)} \tag{11}
$$

$$
ADI_g(t) = \frac{H_g(t)}{H_g^{\max}} \tag{12}
$$

where $k$ indexes brands (or tier classes), $q_k$ is the fuzzy-weighted share of brand $k$ in the group's available assortment, and $H_g^{\max} = \ln K$ is the maximum entropy for $K$ brands with equal shares.

Range: $[0, 1]$. $ADI = 1$ — perfectly equal representation; $ADI \to 0$ — monopolization by one brand/tier.

Interpretation. A declining $ADI$ with a stable $ASI$ means that the product range has not decreased, but the choice has narrowed—for example, premium brands have disappeared, leaving only private labels. This is a critical signal for the regulator.

Alarm threshold: $ADI < 0.6$ — yellow; $ADI < 0.4$ — red.

 

3.2.4. Extinction Rate Index (DR)

Definition. The rate of change in $ASI$. It distinguishes between slow assortment erosion and a sharp collapse.

$$
DR_g(t) = \frac{ASI_g(t) - ASI_g(t-\tau)}{\tau} \tag{13}
$$

Range: $(-\infty, +\infty)$. $DR > 0$ — availability is deteriorating; $DR < 0$ — availability is improving.

Units: per unit of time.

 

3.3. Quality of Substitution Dimension

3.3.1. Substitution Degradation Index (SDI)

Definition. The weighted average loss of quality under actual substitutions within group $g$. It measures how strongly consumer choice degrades when switching to available alternatives.

Let $ActSub_g(t)$ be the set of pairs $(p, q)$, where $p$ is a missing product in group $g$, and $q$ is the product to which the consumer actually switched (or the closest available substitute in the substitution graph $G_S$). Then:

$$
SDI_g(t) = \frac{\sum_{(p,q) \in ActSub_g(t)} w_{pq} \cdot D(p, q)}{\sum_{(p,q)} w_{pq}} \tag{14}
$$

where $D(p, q)$ is the substitution degradation measure, and $w_{pq} = \mu_g(p)$ is the weight reflecting the significance of the original product in the group.

The degradation measure is defined as:

$$
D(p, q) = [1 - s(p, q)] + \alpha_1 \cdot \Delta tier(p,q) + \alpha_2 \cdot \Delta brand(p,q) + \alpha_3 \cdot \Delta storage(p,q) \tag{15}
$$

where $s(p, q)$ is the similarity from the substitution graph; $\Delta tier = 1$ in the case of a tier downgrade (premium $\to$ economy), and $0$ otherwise; $\Delta brand = 1$ when switching from an A-brand to a private label; $\Delta storage = 1$ when the storage type changes (fresh $\to$ frozen). The coefficients $\alpha_1$, $\alpha_2$, and $\alpha_3$ define the relative importance of each degradation type.

Range: $[0, +\infty)$. $SDI = 0$ — ideal substitutions ($s = 1$, no degradation); $SDI > 1$ — severe degradation.

Recommended coefficients: $\alpha_1 = 0.5$ (tier), $\alpha_2 = 0.3$ (brand), $\alpha_3 = 0.4$ (storage). These should be calibrated on empirical data.

Alert threshold: $SDI > 0.3$ — yellow; $SDI > 0.7$ — red.

 

3.3.2. Forced Substitution Rate (FSR)

Definition. The share of products in group $g$ for which the consumer cannot purchase the original product and is forced to use a substitute.

$$
FSR_g(t) =
\frac{
\left| \left\{ p \in P_g : AI(p,t) = 0 \wedge \exists q : (p,q) \in E_S \wedge AI(q,t) > 0 \right\} \right|
}{
|P_g|
} \tag{16}
$$

Range: $[0, 1]$. $FSR = 0$ — substitutions are not required; $FSR = 0.2$ — for 20% of the group’s products, the consumer is forced to seek alternatives.

Alert threshold: $FSR > 0.10$ — yellow; $FSR > 0.25$ — red.

 

3.3.3. Substitution Depth Loss Index (SLI)

Definition. The average share of unavailable substitutes for products in group $g$. It measures how much the “substitution space” has narrowed.

$$
SLI_g(t) =
1 - \frac{1}{|P_g|} \cdot \sum_{p \in P_g}
\frac{
\left| \left\{ q : (p,q) \in E_S \wedge AI(q,t) > 0 \right\} \right|
}{
\left| \left\{ q : (p,q) \in E_S \right\} \right|
} \tag{17}
$$

Range: $[0, 1]$. $SLI = 0$ — all substitutes are available; $SLI = 0.6$ — 60% of substitution links are broken.

Interpretation. $SLI$ grows when nodes drop out of the substitution graph. This means that even with a stable $ASI$ (when product availability itself still appears normal), the system is losing resilience to future out-of-stock events.

Alert threshold: $SLI > 0.20$ — yellow; $SLI > 0.40$ — red.

 

4. Aggregation Hierarchy

 

Product- and category-level metrics are aggregated upward along two axes: the subject axis (product $\to$ category $\to$ basket $\to$ market) and the geographic axis (outlet $\to$ district $\to$ emirate $\to$ country).

 

4.1. Subject Aggregation

 

4.1.1. Basket Availability Index (BAI)

Definition. An aggregated measure of the extent to which a typical household can assemble a standard consumer basket, defined as a set of consumer rituals $R = \{r_1, \ldots, r_L\}$ (breakfast, lunch, dinner, receiving guests, cleaning, etc.).

$$
BAI(t) = \frac{1}{L} \cdot \sum_{r \in R} \min \left(1, \frac{C_r(t)}{R_r} \right) \tag{18}
$$

where $C_r(t)$ is the actual coverage of ritual $r$, and $R_r$ is the minimum required coverage:

$$
C_r(t) = \sum_p \mu_r(p) \cdot AI(p, t) \tag{19}
$$

Range: $[0, 1]$. $BAI = 1$ — the basket is fully available; $BAI = 0.7$ — 30% of rituals are insufficiently covered. The function $\min(1, \cdot)$ ensures that an excess in one ritual does not compensate for a deficit in another.

Alert threshold: $BAI < 0.85$ — yellow; $BAI < 0.70$ — red.

 

4.1.2. Basket Price Index (BPI)

Definition. The relative change in the cost of the standard basket over period $\tau$. Unlike $PSI$, $BPI$ accounts for substitutions: if a product disappears, its cost is replaced by the cost of the nearest available alternative.

$$
BPI(t, \tau) = \frac{V(t) - V(t-\tau)}{V(t-\tau)} \tag{20}
$$

where $V(t)$ is the basket cost at time $t$:

$$
V(t) = \sum_r \sum_p \mu_r(p) \cdot \tilde{U}(p, t) \tag{21}
$$

$$
\tilde{U}(p, t) =
\begin{cases}
up(p, t), & \text{if } AI(p, t) > 0 \\
up(q^{*}, t), & \text{if } AI(p, t) = 0
\end{cases}
\tag{22}
$$

where

$$
q^{*} =
\arg\min_{q : (p,q) \in E_S,\ AI(q,t)>0}
\left[
\frac{up(q, t)}{s(p, q)}
\right]
$$

is the nearest available substitute normalized by similarity.

Interpretation. $BPI > PSI$ means that the actual cost of the basket is rising faster than indicated by the average price index — this is a signal of hidden substitution inflation.

 

4.1.3. Composite Market Condition Index (CMI)

Definition. A unified aggregated indicator combining all three dimensions for group $g$ or for the market as a whole.

$$
CMI_g(t) = \omega_1 \cdot \tilde{PSI}_g(t) + \omega_2 \cdot ASI_g(t) + \omega_3 \cdot SDI_g(t) + \omega_4 \cdot (1 - ADI_g(t)) \tag{23}
$$

where $\tilde{PSI}_g = \max(0, PSI_g)$ is the truncated $PSI$ (growth only), and $\omega_1 + \omega_2 + \omega_3 + \omega_4 = 1$ are the weighting coefficients.

Recommended weights: $\omega_1 = 0.30$, $\omega_2 = 0.30$, $\omega_3 = 0.25$, $\omega_4 = 0.15$. For essential categories, $\omega_2$ is increased at the expense of $\omega_4$.

Range: $[0, +\infty)$. $CMI \approx 0$ — the market is healthy; $CMI > 0.2$ — moderate stress; $CMI > 0.5$ — severe stress.

Aggregation to the national level:

$$
CMI_{national}(t) = \sum_g \theta_g \cdot CMI_g(t) \tag{24}
$$

where $\theta_g$ is the weight of group $g$ (proportional to its share in household consumer expenditure).

4.2. Geographic Aggregation

Each metric $M \in \{PSI, ASI, SDI, ADI, \ldots\}$ is computed separately for each geographic level. Let $S_e$ be the set of outlets in emirate $e$. Then:

$$
M_g^{(e)}(t) = M_g(t \mid S := S_e) \tag{25}
$$

that is, the metric is recalculated using only the data from outlets in emirate $e$.

The national level is the aggregation over all $S$ with an optional population weighting:

$$
M_g^{nat}(t) = \sum_e \left( \frac{pop_e}{pop_{total}} \right) \cdot M_g^{(e)}(t) \tag{26}
$$

Geographic granularity makes it possible to identify where market pressure is stronger and to target inspections or interventions more precisely.

 

5. Leading Indicator

5.1. Empirical Observation

Data analysis shows a persistent time lag between the metrics: growth in $ASI$ and $SDI$ precedes growth in $PSI$ by $\delta = 5\text{–}14$ days. This is explained by the mechanism of retail pricing: the supplier first reduces supply volumes ($ASI$ rises), the retailer switches to substitutes ($SDI$ rises), and only then adjusts retail prices ($PSI$ rises). Shrinkflation ($SFI$) occupies an intermediate position.

 

5.2. Early Warning Indicator (EWI)

Definition. A weighted combination of non-price metrics and the derivative of price pressure that predicts price stress over a horizon of $\delta$ days.

$$
EWI_g(t) = \alpha_1 \cdot ASI_g(t) + \alpha_2 \cdot SDI_g(t) + \alpha_3 \cdot SLI_g(t) + \alpha_4 \cdot SFI_g(t) + \alpha_5 \cdot DR_g(t) + \alpha_6 \cdot PSI'_g(t) \tag{27}
$$

where $PSI'_g(t) = \frac{dPSI_g}{dt}$ is the derivative of price pressure (the acceleration of price growth), ensuring sensitivity to dynamics rather than only to level.

Calibration of weights. Recommended initial values are: $\alpha_1 = 0.25$, $\alpha_2 = 0.20$, $\alpha_3 = 0.15$, $\alpha_4 = 0.10$, $\alpha_5 = 0.15$, $\alpha_6 = 0.15$. These weights should be calibrated on historical data using logistic regression with the binary target “$PSI$ growth $> 5\%$ in the next $\delta$ days”.

Range: $[0, +\infty)$. Interpretation: $EWI < 0.1$ — calm; $0.1\text{–}0.3$ — increased attention; $> 0.3$ — high probability of price stress.

 

5.3. Proxy Reference Under Data Loss

When a marker product disappears from the shelf, the system computes a proxy reference price through the graph of price analogues $E_A$:

$$
ref_{proxy}(p, t) = \operatorname{WeightedMedian}(\{up(a_j, t)\}, \{w_j\}) \cdot pack\_size(p) \tag{28}
$$

$$
w_j = sim_j^{\alpha} \cdot \exp(-\beta \cdot hops_j) \cdot compat_j \cdot \exp(-\lambda \cdot age_j) \tag{29}
$$

where $a_j$ are analogues from $E_A$; $sim_j$ is similarity; $hops_j$ is the number of hops; $compat_j$ is compatibility; $age_j$ is the age of data in days; and $\lambda$ is the temporal decay parameter, which increases during a crisis.

The proxy reference ensures continuity of $PSI$ calculation even when products disappear and solves the problem of an empty reference.

 

5. Leading Indicator

 

5.1. Empirical Observation

Data analysis shows a persistent time lag between the metrics: growth in $ASI$ and $SDI$ precedes growth in $PSI$ by $\delta = 5\text{–}14$ days. This is explained by the mechanism of retail pricing: the supplier first reduces supply volumes ($ASI$ rises), the retailer switches to substitutes ($SDI$ rises), and only then adjusts retail prices ($PSI$ rises). Shrinkflation ($SFI$) occupies an intermediate position.

 

5.2. Composite Early Warning Indicator (EWI)

Definition. A weighted combination of non-price metrics and the derivative of price pressure that predicts price stress over a horizon of $\delta$ days.

$$
EWI_g(t) = \alpha_1 \cdot ASI_g(t) + \alpha_2 \cdot SDI_g(t) + \alpha_3 \cdot SLI_g(t) + \alpha_4 \cdot SFI_g(t) + \alpha_5 \cdot DR_g(t) + \alpha_6 \cdot PSI'_g(t) \tag{27}
$$

where

$$
PSI'_g(t) = \frac{dPSI_g}{dt}
$$

is the derivative of price pressure (the acceleration of price growth), ensuring sensitivity to dynamics rather than only to level.

Calibration of weights. Recommended initial values are: $\alpha_1 = 0.25$, $\alpha_2 = 0.20$, $\alpha_3 = 0.15$, $\alpha_4 = 0.10$, $\alpha_5 = 0.15$, and $\alpha_6 = 0.15$. These weights should be calibrated on historical data using logistic regression with the binary target “$PSI$ growth $> 5\%$ in the next $\delta$ days”.

Range: $[0, +\infty)$. Interpretation: $EWI < 0.1$ — calm; $0.1\text{–}0.3$ — increased attention; $EWI > 0.3$ — high probability of price stress.

 

5.3. Proxy Reference Under Data Loss

When a marker product disappears from the shelf, the system computes a proxy reference price through the graph of price analogues $E_A$:

$$
ref_{proxy}(p, t) = \operatorname{WeightedMedian}(\{up(a_j, t)\}, \{w_j\}) \cdot pack\_size(p) \tag{28}
$$

$$
w_j = sim_j^{\alpha} \cdot \exp(-\beta \cdot hops_j) \cdot compat_j \cdot \exp(-\lambda \cdot age_j) \tag{29}
$$

where $a_j$ are analogues from $E_A$; $sim_j$ is similarity; $hops_j$ is the number of hops; $compat_j$ is compatibility; $age_j$ is the age of data in days; and $\lambda$ is the temporal decay parameter, which increases during a crisis.

The proxy reference ensures continuity of $PSI$ calculation even when products disappear and solves the problem of an empty reference.

 

7. Reporting Modes


7.1. Operational Dashboard (Daily)
Contains: CMI_national, BAI, PSI and ASI heatmap by emirate, top 10 products/categories by EWI, list of products with triggered alerts (red/yellow). Designed for operational decisions.
 

7.2. Analytical Briefing (Weekly)
Contains: CMI, PSI, ASI, and SDI dynamics for the week; CMI decomposition by components; BPI and PSI comparison (identifying hidden replacement inflation); geographic detail; SLI and ADI trends (sustainability erosion warning); EWI forecast for the next week. Length: 2-3 pages.
 

7.3. Strategic Report (Monthly)
Contains: Retrospective analysis of EWI performance (how many triggered alerts were confirmed); Structural shifts in product assortment (ADI trends); assessment of the stability of the substitution graph (SLI trends); recommendations for revising alarm thresholds.
 

8. Application: UAE Supply Crisis (March 2026)


In March 2026, the blockade of the Strait of Hormuz led to a massive disruption of food supplies to the Persian Gulf region. The UAE, which imports approximately 85% of its food, faced rising prices for certain categories, disappearances from shelves, and a surge in panic buying. The Ministry of Economy and Tourism intensified monitoring: over 7,100 inspections were conducted, and 567 price violations were identified. The Ministry's electronic system, covering 627 retail outlets, tracks prices for nine categories of essential goods.
 

The proposed system expands the capabilities of government monitoring in several areas:
Continuity of supply during product disappearances. Proxy reference (28–29) ensures PSI calculation even when marker goods are withdrawn, solving the empty reference problem.
Detection of hidden inflation. The BPI–PSI gap (formulas 6 vs. 20) quantitatively shows how much faster the real cost of the basket is rising than nominal prices. SFI (7) detects shrinkage.
Lead time of 5–14 days. EWI (27) combines ASI, SDI, SLI, SFI, and DR, each of which reacts to supply disruptions earlier than retail prices. This gives the regulator time for proactive measures.
Geographic granularity. Formula (25–26) calculates all metrics broken down by emirate, allowing for targeted inspections and interventions.


Coverage beyond 9 categories. The Gold Catalog covers the entire spectrum of retail goods, while the BAI (18–19) measures the affordability of consumer baskets, not limited to regulated categories.
 

9. Conclusion


This paper proposes a formalized system of fifteen metrics for monitoring the national consumer goods market, organized into three dimensions (price, affordability, quality of substitution) and four levels of aggregation (product, category, basket, market). For each metric, the following are defined: a precise calculation formula, a range of values, an interpretation, and alarm thresholds.
 

Key contributions of the paper:


1. Formalization of fuzzy-weighted indices (PSI, ASI, SDI) using μ_g(p) for correct aggregation across overlapping product groups.
2. Introduction of indices invisible to traditional monitoring: SFI (shrinkflation), ADI (diversity entropy), SLI (substitution depth), and BPI (basket price index accounting for substitutions).
3. Designing a composite leading indicator (EWI) consisting of six components with calibrated weights, providing a 5-14-day price stress warning.
4. Defining an aggregation hierarchy with geographic and subject-specific granularity, including a single composite CMI index.
The system was implemented on the AIHiveLab platform with monitoring across the UAE and tested during the March 2026 supply crisis. Future developments include calibrating the EWI using historical data, integrating a predictive layer based on lead time and supplier availability scores, and expanding to other Gulf markets.

Literature

01

UAE Ministry of Economy and Tourism. Pricing Policy for Essential Consumer Goods in the UAE. Ministerial Decision No. 246 of 2024. Abu Dhabi, 2024.

02

Emirates Food Security Council. National Food Security Strategy 2051. UAE Cabinet, 2018.

03

Dubai Food Security Committee. Food Security Dashboard: Technical Documentation. Smart Dubai, 2020.

04

World Food Programme. WFP warns rising food and fuel prices risk pushing global hunger higher. Situation Report, March 2026.

05

Kpler Commodity Intelligence. Strait of Hormuz: Fertilizer Supply Disruption Assessment. March 2026.

06

Shannon C.E. A Mathematical Theory of Communication. Bell System Technical Journal, 27(3), 1948. Pp. 379–423.

07

Zadeh L.A. Fuzzy Sets. Information and Control, 8(3), 1965. Pp. 338–353.

08

ILO, IMF, OECD, Eurostat, UN, World Bank. Consumer Price Index Manual: Concepts and Methods. Washington, D.C., 2020.

09

Diewert W.E. Index Number Theory Using Differences Rather Than Ratios. American Journal of Economics and Sociology, 64(1), 2005. Pp. 311–360.

10

Silver E.A., Pyke D.F., Thomas D.J. Inventory and Production Management in Supply Chains. 4th ed. CRC Press, 2017.

11

Persson L., Rosén B. On the OECD–FAO Agricultural Outlook methodology for price index construction under product substitution. OECD Papers, No. 181, 2023.

12

Google OR-Tools: CP-SAT Solver Documentation. https://developers.google.com/optimization.

Language
RU EN AR

Write to us

Fill out the form and we'll get back to you

Let's meet for coffee

Leave your name and phone number — we'll get in touch