AI Transparency Statement

This AI Transparency Statement (“Statement”) explains how artificial intelligence and machine-learning functionality is intended to be used within PRE-ORDER, what role AI outputs play in the Platform, what information may support those outputs, how human oversight is maintained, and what safeguards apply to transparency, data protection, monitoring and user control.

This Statement should be read together with the Privacy Policy, GDPR Notice, Terms of Service, Platform Terms, Seller Agreement, Buyer Agreement and other applicable PRE-ORDER policies. It describes the intended and deployed role of AI functionality at a general level. Specific features may be introduced, changed, limited or withdrawn as the product develops and as technical, legal and validation requirements evolve.

1. Operator and scope

1.1 The PRE-ORDER Platform is operated by “PRE-ORDER Europe” EOOD, UIC/EIK 208904127, registered at Arthur Residential Complex, Block C, Floor 3, Apartment C44, 8229 Sveti Vlas, Nessebar Municipality, Burgas District, Bulgaria (“PRE-ORDER”, “Operator”, “we”, “us” or “our”).

1.2 This Statement applies to AI-supported functionality made available through the PRE-ORDER website, mobile application, administrative interfaces, brand tools, analytics modules and related services where PRE-ORDER determines or controls the relevant AI use case.

1.3 PRE-ORDER may rely on technology suppliers, cloud providers, development contractors, data processors or specialised model providers. Their involvement does not change the transparency principles in this Statement. Where a third-party AI service materially affects a user-facing feature, PRE-ORDER will seek to provide appropriate information in the product, documentation or privacy materials.

2. Purpose of PRE-ORDER AI

2.1 PRE-ORDER is designed as a fashion pre-order and demand-information platform. Its AI functionality is intended primarily to support business users with demand forecasting and related decision support before or during production planning.

2.2 Depending on product rollout and available data, AI-supported functionality may assist with:

• forecasting likely demand for a collection, product or SKU;
• identifying demand patterns and changes over time;
• analysing pre-order, conversion and relevant historical demand signals;
• supporting production-volume recommendations;
• highlighting uncertainty, trends or anomalies in available demand data;
• supporting inventory and overproduction-risk analysis;
• supporting validation and analytics during project pilots.

2.3 AI functionality is not intended to replace professional judgement, contractual obligations, statutory duties, product-safety responsibilities or decisions that Brands, Sellers, Buyers or other users are legally required to make themselves.

3. What PRE-ORDER AI is not intended to do

PRE-ORDER does not design its demand-forecasting AI to:

• make legally binding decisions about individuals;
• determine a person’s access to essential services;
• autonomously decide whether an individual may obtain credit, employment, housing, healthcare or another essential opportunity;
• autonomously make financial decisions about individuals;
• autonomously initiate production without an authorised human decision;
• replace the Seller’s responsibility for product safety, conformity, delivery, refunds or contractual performance;
• guarantee future demand, sales, profitability, production success or environmental outcomes.

If a materially different AI use case is introduced in the future, PRE-ORDER will assess that use case separately and update relevant transparency, privacy, risk and governance documentation where required.

4. Decision support, not autonomous production

4.1 AI outputs used for production planning are advisory. An authorised Brand user remains responsible for the final production decision and can accept, modify or reject an AI-supported recommendation.

4.2 The intended decision flow is: data and demand signals → forecast → recommendation → human review → human decision → production or other business action.

4.3 Production is not intended to be automatically initiated solely because an AI model generated a recommendation. Where the Platform provides workflow automation, the relevant Brand user remains responsible for confirming the business action unless a separate, clearly disclosed and lawfully assessed automation is introduced.

4.4 PRE-ORDER may retain the original recommendation, the subsequent human decision and relevant changes for audit, validation, quality improvement and dispute-support purposes, subject to applicable data-protection and retention rules.

5. High-level description of the forecasting process

5.1 PRE-ORDER may use statistical, machine-learning or neural-network techniques to estimate future demand. The exact model family may vary by feature, product category, data availability, validation results and deployment stage.

5.2 At a high level, the forecasting process may involve:

• collecting or receiving relevant operational and demand data;
• cleaning, validating, aggregating or pseudonymising data where appropriate;
• creating model features representing demand behaviour, seasonality, order activity or other relevant factors;
• training or calibrating a model on appropriate datasets;
• producing a forecast for a defined product, SKU, collection or period;
• applying disclosed business constraints or decision rules where relevant;
• presenting the resulting recommendation together with available confidence or uncertainty information and meaningful explanatory context.

5.3 The recommendation layer may apply business constraints such as confirmed pre-orders, paid demand, minimum order quantities at collection, SKU or colour level, production limits or other rules configured by the Brand or relevant workflow. A forecast and a production recommendation are therefore not necessarily the same number. Where these indicators are presented, cart activity, submitted pre-orders and financially confirmed demand should not be represented as equivalent.

6. Data that may support AI and analytics

Depending on the feature and lawful purpose, AI-supported demand forecasting may use data such as:

• product, SKU and collection information;
• historical and current sales or order information;
• pre-order volumes and status information;
• conversion and engagement indicators such as product views, collection views, wishlist activity or add-to-bag events;
• payment-status or paid-demand indicators where appropriate, without using full card credentials as model inputs;
• production volumes and lead times;
• inventory and availability information;
• seasonality and timing information;
• sustainability or operational indicators where relevant to the specific analytical purpose;
• aggregated or derived variables such as Demand Scores, conversion indicators and forecast variables.

The actual input set may differ by model and customer configuration. PRE-ORDER aims to use data that are relevant to the specific forecasting purpose and to avoid unnecessary personal data.

7. Data not intended for AI model features

The following categories are not required for PRE-ORDER demand forecasting and are not intended to be used as AI model features:

• passwords;
• payment-card numbers;
• CVV or card security codes;
• authentication credentials;
• identity documents;
• private payment credentials;
• free-text private chat content;
• special-category personal data under Article 9 GDPR, unless a future, separately assessed use case lawfully requires such data.

If a user incidentally enters sensitive information into a free-text field or chat, PRE-ORDER does not intend to extract that information for demand-forecasting model training. Users should nevertheless avoid submitting unnecessary sensitive information through Platform communications.

8. Personal data, pseudonymisation and aggregation

8.1 AI and analytics may involve personal data where user behaviour, transaction metadata or business-contact information can be linked to an identifiable person. Such processing is subject to the Privacy Policy, GDPR Notice and applicable data-protection law.

8.2 Where feasible and appropriate, PRE-ORDER intends to separate direct identifiers from analytical datasets and use pseudonymised or aggregated data for forecasting, model validation and project-level reporting.

8.3 The use of personal data for AI-related purposes does not remove data-subject rights. Where profiling or automated processing is legally relevant, PRE-ORDER will assess the applicable transparency, objection, access and other rights and provide additional information where required.

9. Training and validation data

9.1 During development and pilot validation, PRE-ORDER may use synthetic, anonymised, aggregated, pseudonymised or real operational company data where appropriate and lawfully available.

9.2 Real pilot-company data may include product and SKU data, collection data, historical or current sales, pre-order volumes, order status, production volumes, production lead times, inventory information, conversion indicators and sustainability or operational indicators.

9.3 Where company datasets contain personal data, such as information about company representatives or sole traders, those elements remain subject to applicable data-protection requirements.

9.4 Training and validation datasets should be documented sufficiently to support provenance, quality review, performance evaluation and later investigation of material issues.

10. Human-readable explanations

10.1 PRE-ORDER aims to provide a transparent and auditable decision-support rationale rather than claiming that a complex neural network can provide a perfect causal reconstruction of every internal mathematical operation.

10.2 For a material demand forecast or production recommendation, the user-facing explanation may include, where technically appropriate and sufficiently validated:

• what was predicted and for which product, collection, SKU or forecast horizon;
• the forecasted demand or relevant output;
• the model name or version where appropriate for audit and support;
• the level of confidence, uncertainty or prediction interval where supported;
• the main input-factor groups or variables associated with the output;
• relevant production or business constraints;
• the rule or logic used to translate a forecast into a recommendation where such a rule applies;
• the recommendation presented to the Brand;
• the fact that the recommendation is advisory and requires human review.

10.3 Where post-hoc explainability methods are used, their output should be understood as an interpretation of factors associated with model behaviour, not as a guaranteed causal explanation of the internal reasoning of a neural network.

11. Confidence, uncertainty and insufficient data

11.1 Forecasts contain uncertainty. PRE-ORDER does not represent AI forecasts as guaranteed outcomes.

11.2 Where available data are insufficient, model confidence is inadequate or an explanation cannot be provided reliably enough for the intended use, PRE-ORDER may:

• display a low-confidence or insufficient-data warning;
• show a wider uncertainty range;
• require additional human review;
• use a simpler or more interpretable baseline approach;
• withhold a production recommendation;
• limit the feature until additional validation is completed.

11.3 A Brand should not treat a high-confidence score as a guarantee of market demand. Unexpected market conditions, supply-chain events, consumer behaviour, pricing, marketing activity, product quality and other external factors may materially affect actual outcomes.

12. Accuracy and project targets

12.1 PRE-ORDER evaluates model performance using suitable validation metrics. Depending on the forecasting use case, these may include MAE, MAPE, RMSE, forecast accuracy percentages or other metrics appropriate to the model and dataset.

12.2 The PRE-ORDER project includes a target of at least 85% forecasting accuracy where sufficient validation data are available. This is a project development and validation target, not a contractual guarantee that every individual forecast, product category, Brand, collection or deployment will achieve that level.

12.3 Performance may vary depending on data quantity, data quality, product type, seasonality, market volatility, model maturity and the forecast horizon.

13. Monitoring and lifecycle management

PRE-ORDER intends to maintain proportionate monitoring and model-management procedures, which may include:

• model and configuration versioning;
• training- and validation-dataset documentation;
• performance measurement and comparison with actual outcomes;
• monitoring for model drift and anomalous errors;
• review of low-confidence or materially incorrect outputs;
• retraining and recalibration procedures where justified;
• rollback or disabling capability where a model or feature creates unacceptable risk;
• logging sufficient information to investigate significant recommendations and changes.

Model monitoring does not eliminate all error. It is intended to help identify degradation, unexpected behaviour and material limitations so that appropriate corrective action can be taken.

14. Fairness, bias and data quality

14.1 Demand forecasting can reflect limitations and biases in historical or operational data. A model trained on incomplete, unrepresentative or systematically distorted data may produce less reliable forecasts.

14.2 PRE-ORDER intends to address relevant risks through data-quality checks, validation, comparison across datasets or periods, monitoring of unusual errors and human review. The exact controls will depend on the model and deployment context.

14.3 PRE-ORDER AI is intended to forecast commercial demand and support production planning, not to infer sensitive personal characteristics or make discriminatory decisions about individuals.

15. User responsibilities when relying on AI

Users who receive AI-supported forecasts or recommendations should:

• review the underlying collection, product, pricing, pre-order and operational context;
• consider whether the data supplied to PRE-ORDER are complete, current and accurate;
• consider uncertainty, warnings and model limitations;
• apply professional judgement and any legal, safety, financial or operational requirements that apply to their business;
• avoid representing an AI forecast to third parties as a guaranteed future result;
• report significant errors or unexpected outputs through available support channels.

A Brand remains responsible for the production quantity it ultimately selects, even where PRE-ORDER supplied an AI-supported recommendation.

16. AI, environmental and overproduction information

16.1 PRE-ORDER is designed to help align production more closely with observed and expected demand. AI-supported forecasting may contribute to that objective by giving Brands additional information before production decisions are finalised.

16.2 Any environmental, waste-reduction, CO₂ or overproduction indicators shown by PRE-ORDER may depend on underlying assumptions, coefficients, data quality and validation methodology. Project targets and pilot results should not be interpreted as already-proven universal outcomes.

16.3 Where AI is used to support sustainability or ESG-related indicators, PRE-ORDER intends to distinguish measured data, estimates, modelled values and project targets where reasonably practicable.

17. AI-generated or AI-assisted content

17.1 PRE-ORDER’s principal documented AI use case is demand forecasting and production decision support. The Platform may in the future include additional AI-assisted tools, such as content assistance, product-data enrichment or image-related functionality.

17.2 If such functionality is introduced, PRE-ORDER will seek to apply appropriate disclosure, intellectual-property, consumer-protection, data-protection and content-moderation controls. Users remain responsible for ensuring that content they publish is lawful, accurate and not misleading.

17.3 AI-generated imagery must not be used to materially misrepresent a Product’s essential characteristics where such presentation would mislead a Buyer or Consumer.

18. Logging, traceability and auditability

18.1 For material AI outputs, PRE-ORDER intends to preserve sufficient information to reconstruct the decision-support context to a reasonable degree. Depending on the feature, the record may include model/version information, time, prediction object, forecast horizon, input-factor groups, output, confidence or uncertainty, business constraints, recommendation, human decision and later actual outcome.

18.2 Logs may be used for quality assurance, model validation, security, troubleshooting, compliance, support, dispute handling and investigation of significant errors, subject to access controls and retention requirements.

18.3 Traceability records are not intended to expose proprietary source code, trade secrets, confidential model parameters or information that would compromise security or the rights of other users.

19. Data protection, security and access controls

19.1 PRE-ORDER intends to apply data-protection-by-design and data-protection-by-default principles to AI-related processing.

19.2 Relevant safeguards may include encryption in transit, encryption at rest for appropriate categories, role-based access, least-privilege permissions, audit logging, secure API authentication, separation of environments, pseudonymisation and monitoring of security-relevant events.

19.3 Access to model-development data, AI logs or sensitive analytical information should be limited to authorised personnel and service providers who need access for legitimate operational, technical, security or compliance purposes.

20. Risk assessment and governance

20.1 PRE-ORDER will assess AI use cases in light of their intended purpose, affected users, data, level of autonomy, potential impact and applicable law. Where a use case could create a high risk to individuals, additional assessment, safeguards or a Data Protection Impact Assessment may be required before deployment.

20.2 Governance responsibilities may be allocated among product, technical, AI/ML, security, privacy/compliance and operational functions. Relevant service providers may be contractually required to support security, data protection and technical documentation obligations.

20.3 PRE-ORDER is being developed with the objective of meeting applicable EU requirements relevant to AI, data protection, digital services and consumer protection. Regulatory classification can depend on the exact deployed use case and may be reassessed as the product evolves.

21. User questions, concerns and challenges

21.1 Users may contact PRE-ORDER if they have a question about an AI-supported feature, believe an output is materially incorrect, need clarification about the role of AI in a workflow or wish to raise a privacy concern.

21.2 Contact: info@b2b-agency.cc. Users should provide enough context to identify the relevant feature, order, collection or recommendation, but should not send passwords, CVV codes or full card details.

21.3 Where applicable under data-protection law, users may also exercise rights described in the Privacy Policy and GDPR Notice. This Statement does not limit rights that cannot lawfully be waived.

22. Changes to this Statement

PRE-ORDER may update this Statement when AI functionality, model architecture, data sources, legal requirements, governance controls or user-facing explanations materially change. Material changes will be communicated in an appropriate manner. The current publication will apply from the effective date shown above.

Annex 1. Minimum transparency record for material AI outputs

Where relevant to the use case, PRE-ORDER’s internal or user-facing AI record may include:

• Forecast or Recommendation ID;
• model name and version;
• date and time;
• Brand / Collection / Product / SKU;
• forecast horizon;
• relevant input-factor groups;
• forecasted demand or other output;
• confidence, uncertainty or prediction interval where available;
• applicable business constraints, including MOQ, paid demand or production limits;
• decision rule(s) translating the forecast into a recommendation, where applicable;
• optimisation objective or criteria, where applicable;
• alternative production options considered where the logic explicitly evaluates alternatives;
• main influencing factors;
• AI recommendation;
• human decision: accept / modify / reject;
• final human-selected production quantity or action;
• actual outcome once available;
• validation status, warnings and material limitations.