12. Technical Architecture, Data and Security
The experience eBattery offers the user rests on different services working together within a single app. A station found on the map can be added to a route plan; a trip that has been created can be shared in the community; a charging session can be started at a supported operator, and the completed session can be reflected in the user’s spending history.
Behind this integrity lies a shared technical structure in which each module keeps its own responsibility. The map, AI, messaging, Game Zone, payment and device management work with different types of data. Besides connecting these areas, separating access permissions and transaction limits correctly is also important.
eBattery’s architectural approach is based on four core principles: modularity, data accuracy, service continuity and user control. These principles guide all technical decisions, from the app’s day-to-day operation to integrations in new countries.
For a user, the counterpart of technical quality is concrete: the map opens on time, how up to date the station information is can be understood, the status of a sent message can be seen, and when a payment transaction is repeated, no second charge occurs. In terms of privacy, the user can manage who their location is shared with and which of their data is used for which service.
This section defines eBattery's architectural design, its approach to data management, and the security principles that will be maintained as services grow. The charging, payment, marketplace and home charging functions to be opened after launch are tied to the same technical and governance framework.
12.1 System Architecture and Integrations
Shared user experience, specific service responsibilities
eBattery is designed to be integrated from the user's perspective and modular in terms of technical responsibilities. While the user can access different services with a single account, each module reaches only the data and transaction permissions it needs to perform its task.
For example, the vehicle profile supports station filtering and route planning. The same profile can also contribute to assessing product compatibility in EV Market. By contrast, a seller does not need access to the user's private messages or past location records. A shared account does not mean all data is opened to all services.
Similarly, the real vehicle profile, the virtual vehicle in the Game Zone and the investor identity on the Web3 side are separate records. The relationships needed between them can be established, but a level, authority or status in one is not automatically carried over to another.
This approach makes it easier to add new modules while also preserving the boundaries of existing services.
Core layers of the architecture
eBattery's technical structure is addressed through layers of responsibility, from user interaction to data storage.
| Layer | Core responsibility |
|---|---|
| Mobile app | Screens, user interaction, device permissions and accessibility |
| Access and identity layer | Session verification, authorization and request checking |
| Product services | Business rules for map, route, community, game, orders and charging |
| Integration layer | Connection with map providers, operators, payment institutions, devices and AI models |
| Data layer | Transaction records, geographic data, content, events and analytics |
| Operations and observability | Error tracking, performance measurement, security events and service health |
These layers do not mean that each function must run on a separate server. The priority is to separate responsibilities and data access clearly. As transaction volume and operational needs grow, components are organized so that they can scale independently.
In this way, technical development does not progress merely by adding more infrastructure. It is measured which service needs which resource, and capacity is increased accordingly.
Division of tasks between the mobile app and the server
The mobile app manages the user's transactions in a clear and smooth way. Precise authorization checks, financial transactions, reward balances and critical business rules are verified on the server side.
For example, the EP balance shown on screen in the Game Zone reflects the records on the server. Changing a value on the user’s device does not change the real balance. The same principle also applies to order amounts, coupon use, reservation entitlement and charging session ownership.
The information sent to the mobile client is limited to the needs of the relevant screen. Server-side secret keys are not placed in the app package. Client keys required by mobile SDKs are managed with the app ID, usage scope and the restrictions supported by the provider.
Requests made with an older version of the app are also taken into account. Thanks to API versions and compatibility rules, updates are rolled out without disrupting users’ ongoing transactions.
Map and location services
Google Maps is one of the core integration components for eBattery's map experience and related location services. Map display, address search, place information and route services are applied according to the scope of the product used and the relevant access terms.
A charging station's operating status, connector availability and current tariff are linked to the relevant operator or authorized data source. A place appearing on the map does not mean that live charging data or remote-operation authority exists for that place.
For this reason, eBattery does not reduce a station record to a single provider's place ID. It sets up controlled matching between its own station ID, the map provider's place ID, the operator's station ID and the socket records.
This matching reduces the same station appearing again under different names and makes it possible to trace which source the data came from.
OCPP and OCPI distinction in charging integrations
In charging infrastructure, different protocols manage different relationships. eBattery treats this distinction as a core part of the architecture.
OCPPfocuses on the communication between the charger and the central charging management system. Device status, transaction records, remote commands and supported smart charging functions are considered within this relationship. The protocol version and the features the device implements determine the scope of integration.
OCPI, supports the exchange of data and services between charging network operators and mobility service providers. It is used in areas such as station information, authorization, sessions and completed charging transaction records.
Accordingly, eBattery accessing an operator's network and eBattery controlling a device it manages itself are not the same technical connection. On the operator side, OCPI or an API offered by the partner may be used, while OCPP-based connections may come into play for direct device management.
Opening an integration is carried out together with protocol support, a commercial agreement, transaction authorizations and end-to-end verification. Supporting a standard does not automatically make all of the counterparty's functions available.
State management in charging and payment transactions
A charging session and a payment transaction are linked processes, but they have separate statuses.
The user pressing the "Start Charging" button, the command being passed to the operator, the operator accepting the command and the device starting to transfer energy are different stages. The app does not merge these stages into a single success message.
Similarly, card verification, pre-authorization, final capture and refund are separate financial transactions. Between the completion of charging and the verification of the final fee, transaction recording and reconciliation may be needed.
So that repeated requests do not produce duplicate transactions, each critical transaction is associated with a unique identifier. If the network connection is lost, the user retrying the same transaction should not cause the same charge or collection to be created a second time.
A transaction that gets no response is not treated directly as failed. First, the transaction status is queried from the provider concerned and the records are compared. This approach prevents a charging session that has actually started from being restarted, or a payment that has actually gone through from being collected again.
Separating AI from transaction systems
eBattery AI uses language models to understand user intent and explain results. Route feasibility, energy calculation, connector compatibility, transaction authority and payment rules are evaluated with verifiable application logic.
For example, when a user says “Tomorrow morning I want to make two short stops on my way to İzmir,” the model interprets the context of the request. The route, energy need, station options and time conditions are obtained from the relevant services. The result is built from structured data.
Access to different model providers, such as the Gemini and GPT families, is handled through a common model connection layer. A change of provider does not become an element that changes the app’s account, payment or route rules.
A model suggesting an action is separate from that action being carried out. Actions that have financial consequences or change the account status on the user’s behalf go through authorization and approval checks.
Event-based communication and consistency across modules
Completing a transaction may involve more than one module. A completed charging session can update spending history, create a notification and trigger the evaluation of a suitable loyalty campaign.
Secondary operations like these are carried out through event logs without making the main operation wait unnecessarily. If the same event is received again, the results are not duplicated.
A failure to send a notification does not cause a completed payment to be lost. A temporary slowdown of the analytics system also does not stop the user's core transaction. Each component tracks its own result and error state.
Shared route, trip and efficiency cards are also tied to versioned content records. Opening a link does not replace authorization to access the content; for private content, the authorization check is applied again.
12.2 Data Layer and Analytics
At eBattery, the value of data is determined by its quality before its quantity. Knowing a station’s location does not mean knowing that the station is operating right now. A user searching a region also does not show that they will complete a charge there.
Therefore, in the data architecture, each piece of information’s its source, time, scope and intended use considers together.
Shared data model
Different operators and service providers may define the same concepts in different ways. One provider may use the name "station" for a physical location, while another may register a single device under that name.
eBattery organizes these differences in a shared data model. The physical location, charging equipment, socket, operator, tariff and session are kept separate from one another. Each record has its own identity and relationships.
This distinction matters for giving users accurate information. Having four sockets at a location does not, on its own, show that four vehicles can charge at the same power at the same time. Power sharing and simultaneous-use conditions are assessed using the relevant equipment data.
Tariff records hold the currency, unit of measure, validity time and applicable fee components. kW, which expresses power, and kWh, which expresses the amount of energy, are not confused with each other. Currency, date and unit formats used in different countries are presented to the user in a suitable way.
Source and freshness of the data
An operational record does not consist of a value alone. Where possible, the following information is evaluated together in the relevant record:
The system or authorized source that provides the data.
The event time reported by the source.
The time the data was received by eBattery.
The relevant station, socket or transaction ID.
Validity period of the data.
Verification and quality status.
Update or correction history.
This helps prevent a late-arriving status message from mistakenly overwriting newer information. When the same record arrives through different channels, duplicates can be filtered out.
When sources conflict, the system does not merge them invisibly. For example, a user’s “station closed” report and an operator’s “available” status are different types of data. The user report is taken into consideration; its relationship to the operator record and the time difference are preserved.
Missing information is not interpreted as a zero value. Failing to get a price does not mean the service is free; failing to get status information does not mean the socket is available; a delayed energy record does not mean zero consumption.
Storage appropriate to the data type
Keeping all data in a single storage structure makes it hard to meet different usage needs. eBattery's data layer is organized according to the access and consistency characteristics that each data type requires.
| Data field | Technical priority |
|---|---|
| Account, order and transaction records | Preserving relationships and transaction integrity |
| Station and location data | Making proximity, region and route queries efficient |
| Live status information | Freshness, timeout and fast access |
| Photos and other media content | Controlled storage and appropriate distribution |
| Message and event records | Ordering, delivery status and duplicate check |
| Analytics data | Batch evaluation without straining the operations system |
A cache provides fast access to frequently used information, but not all data is kept in the cache for the same length of time. A station’s address and a socket’s real-time usage status have different refresh needs.
For content obtained from map providers, contractual usage rules also apply. For example, Google Places policies set limits and certain exceptions for storing and caching content, and attribution is part of the use. For this reason, eBattery’s data policy tracks the information it produces itself and licensed third-party content separately.
Separating operational data from analytical data
The operational system focuses on completing what the user is doing at that moment. The analytics system, in turn, evaluates the use and quality of the service over time.
Searching for a station on the map, reviewing a route recommendation, attempting to start charging, or completing a quest can be measured as different product events. The context needed for these events is kept; private messages or detailed location history are not carelessly passed into general analytics records.
Analytics is used in three main areas:
Product experience: Understanding which steps users struggle with and which functions they benefit from.
Service quality: Identifying integration errors, data delays and transactions that could not be completed.
Infrastructure planning: Assessing aggregated demand signals under appropriate legal and technical conditions.
The measurements for these areas are defined separately. Looking at a screen and successfully completing a service are not counted as the same success indicator.
Regional demand analysis
eBattery's map and route usage can provide useful signals for charging infrastructure planning. Repeated searches in certain areas, the lack of suitable alternatives, or users having to deviate significantly from their route support a closer look at capacity needs.
These data points alone do not produce a definitive investment decision. The platform's user distribution, seasonality, purpose of visit, the operating status of existing stations and the scope of the data are assessed together.
For example, low usage in an area with heavy search activity does not necessarily mean low demand. Access problems, price, faults or an incompatible socket can also play a part.
In regional analyses, geographic and temporal aggregation, suppression of small groups and access limits are applied to prevent individual movements from being disclosed. Removing a name from data or encoding a user ID is not used on the assumption that it alone provides anonymity.
Data lifecycle
Collecting, actively using, archiving and deleting data are parts of the same life cycle. A purpose and a retention approach are defined for each data category.
A road alert on the map disappearing from view after 10 minutes does not mean that all technical records related to that report are deleted at the same time. The retention period for map visibility and for records needed for security, objection or abuse review may differ. This distinction is explained in the relevant notices.
Similarly, the financial record of a completed charging transaction and the temporary location data collected during the transaction do not have to be kept for the same period. Once the purpose ends, unnecessary retention of the data is prevented.
Web3 data bridge
For station projects on the Web3 side, the data flow is built on verifying operational records and being able to track appropriate summaries onchain.
IoT and device data provide information about energy transfer and technical operation. However, the amount of energy transferred alone does not show collected revenue or distributable profit. For revenue, completed sessions and collections are assessed together; for expenses, the relevant operating and accounting records are assessed together.
The Oracle layer supports bringing data from defined sources onto the chain. The reliability of the source, measurement accuracy and reconciliation of financial records are ensured separately. Writing a piece of data to the blockchain does not by itself prove that its source in the physical world is correct.
Public project reports are presented with the period, calculation method, data scope and verification information. Personal information and details in commercial documents that need to be protected are separated from the financial indicators required for investor transparency.
Private messages, identity documents and detailed personal movement records are not transferred to the public chain. Technical digests to be used for verification are also assessed for re-identification risk.
12.3 Scalability and Performance
In eBattery, performance is evaluated by how quickly and reliably the user can complete their task. A server responding quickly matters; but if the user cannot see a meaningful map result or cannot tell whether charging has started, the experience is not considered complete.
For this reason, the performance approach considers infrastructure measurements and user experience together.
Scaling with workload
Map queries, messaging, media uploads, route calculation and payment transactions consume different resources. The capacity needs of these workloads are monitored separately.
As map usage grows, geographic query and cache capacity may become important. At a busy community event, message delivery and connection counts may stand out. During holiday periods, route calculation and operator data updates may create more load.
The architecture is arranged to reduce these areas slowing each other down unnecessarily. Long-running calculations and background jobs are separated from the processes that handle the user's short interactions.
Capacity increases are not tied to processor usage alone. Queue length, response times, connection counts and error rates are also evaluated. This creates indicators that allow intervention before congestion reaches the user.
Map performance
Queries focused on the area visible on the map are used. Stations that are not on the user's screen do not need to be downloaded again on every movement.
Proximity queries, suitable geographic indexing and display density are handled together. In wide areas, grouping markers, increasing detail as the user zooms in and updating only the data that has changed provide both speed and readability.
Live status information is updated separately from station information that changes more slowly. This avoids re-fetching all of a station’s media and description content when only the status of one socket changes.
Mobile performance evaluation also includes data consumption and battery usage. The frequency of location updates is linked to the screen’s context and the needs of the relevant feature.
AI and route calculation performance
In the AI experience, the first response appearing and a usable result being fully completed are different measurements. Streaming lets the user see that the system is responding, but the readiness of a verified route card is tracked separately.
In route planning, finding candidate stations, checking technical suitability, calculating energy and evaluating alternatives are separate stages. Independent operations are run in parallel where appropriate, and unnecessary repetition is reduced.
Performance improvement is not achieved by removing mandatory compatibility checks. Producing a fast result alone, without checking the energy safety margin, socket compatibility or access condition, does not fulfill the purpose of the product.
When an external model or route provider is delayed, the system is managed with time limits and suitable alternative flows. A result whose freshness has not been verified is not presented as definitive information just to respond quickly.
Messaging and notification performance
In messaging, sending, reaching the server, delivery to the recipient and, where supported, read status are handled separately. Users can tell what stage their message is at.
Messages resent after a connection interruption are deduplicated. Ordering within the conversation is preserved and missing records are resynchronized.
A push notification is not the only channel where a message is stored or verified. Even if the phone’s notification service is delayed, the message must be accessible within the app. The same approach also applies to charging results and order updates.
Measurable service targets
Performance targets are tied to service indicators measured under defined device and network conditions. Alongside averages, percentiles that represent users experiencing high latency are also monitored.
For example, the p95 response time is the time below which 95 percent of measured transactions fall. This measurement allows slow experiences, which are not visible when looking only at the average, to be assessed.
| Field | Key indicators monitored |
|---|---|
| Mobile app | Launch experience, crash and freeze rates, memory and battery usage |
| Map | First usable view, query time and age of status data |
| AI | First response time, verified result time and failed transaction rate |
| Messaging | Delivery delay, retries and synchronization after reconnection |
| Charging | Command acceptance, session start and completion of the result record |
| Payment | Transaction success, duplicate checking and reconciliation accuracy |
| Background jobs | Queue wait time and number of unfinished jobs |
Service level targets are set based on these indicators. The measurement scope, the network conditions used, and the effects of external providers are separated in the records.
The overall experience for the user is the basis. Knowing that a delay originates from an external provider does not remove the need to manage the problem visibly and to consider alternatives.
Resilience to outages and recovery
Separating failure domains in critical services reduces the spread of a problem in one component to the whole application. Unlimited requests are not sent to an unhealthy external connection; controlled retry and temporary suspension mechanisms are applied.
Which functions can continue during an outage is also determined by service. A previously saved favorites list can be displayed, while a payment or charge command that requires current authorization is not treated as completed offline in the same way.
Backup and service recovery are separate matters. Besides taking backups, restorability, record integrity and re-matching with transaction systems are tested.
For each critical service, a target recovery time and an acceptable data loss limit are defined. In areas such as payment and charging records, recovery also includes reconciliation with the records held by external providers.
Controlled release and continuous monitoring
New releases are evaluated for performance as well as for function and security checks. Scenarios such as heavy map usage, concurrent messaging, bulk notifications and a large number of charging events are included in load tests.
Releases can be monitored through controlled user groups or staged rollout. Operational flows exist to restrict a feature or roll back to a suitable version when a critical error is detected.
Database changes are also part of this process. Transitions are planned so that old and new app versions can work together. The integrity of user data is a release criterion that ranks ahead of release speed.
12.4 Security, KVKK and Privacy
eBattery’s security approach aims to protect the accuracy of transactions and the accessibility of services together with the privacy of personal data.
A compromised account, a user’s location being shared with the wrong people, and a charging command sent without authorization are different risks. Security measures are arranged according to these different areas of use.
Verification frameworks such as OWASP MASVS are used as a reference in mobile security assessments. This framework covers areas such as on-device data storage, authentication, network communication, and interaction with the mobile platform together. eBattery’s approach is to relate these topics to the product’s real transaction flows.
Authentication and authorization
Verifying who a user is and being authorized to perform a specific transaction are separate controls.
Even if a user is signed in, they cannot view another user's order, private group, device or charging session. Access control is enforced on the server for every resource and action.
Operations staff, vendors, service providers and administrators have different scopes of authority. Permissions are limited to the level the job requires. For critical administrative actions, additional verification, logging and, where necessary, a second approval are applied.
Revoking sessions, evaluating suspicious sign-ins and account recovery are parts of the same security model. The account recovery process is not turned into an easy side door that bypasses normal sign-in checks.
In Web3 functions, the app session and the authority for wallet transactions are also treated separately. A user signing in to their Web2 account does not mean giving unlimited approval to all on-chain transactions that have financial consequences.
Communication, data storage and key security
Communication between the app, the server and external services runs over secure channels. Stored sensitive data is protected with access permissions and encryption policies.
Encryption keys and service secrets are managed separately from application data. Key rotation, access revocation and usage logs are included in operational processes.
Session information kept on the mobile device is managed with the operating system’s appropriate secure storage facilities. Passwords, access keys, payment information and unnecessary personal content are prevented from being written to error logs.
For payment card transactions, the secure collection and tokenization functions of the authorized payment provider are used. Card security codes are not recorded in eBattery's general application logs. The scope of payment information is assessed separately according to the service model used.
AI security and data access
Content sent to language models is limited to the information the task requires. For a route suggestion, the user’s entire message history or shopping records do not need to be sent to the model.
Content taken from external sources and text written by users is not accepted as a trusted system instruction. For example, a directive placed inside a station description does not authorize the AI to make a payment on the user’s behalf or to disclose other data.
The tools, transaction types and data fields the model can access are limited. Input and output validation, authorization checks and required user confirmations are enforced at the application layer. This approach is consistent with OWASP's security assessments of AI systems that are given excessive agency.
The terms under which model providers store data, reuse it and pass it to sub-service providers are part of service selection. A working technical connection is not considered sufficient on its own for the data processing terms to be appropriate.
Purpose-limited data processing
eBattery links personal data to specific service purposes. The KVKK principles of accuracy, being up to date, relevance to the purpose, proportionality and retention for as long as necessary form the basis of data management.
Account creation, route planning, social location sharing, order delivery and marketing communication are different processing purposes. The data needed and the appropriate legal basis are assessed separately for each purpose.
The privacy notice explains which of the user’s data is processed and why. Where explicit consent is required, consent is obtained separately. Displaying the privacy notice does not mean consent is presumed for all data processing; the Authority’s regulation also requires that the privacy notice and explicit consent be handled separately.
A location or notification permission obtained from the operating system is also not treated as replacing all legal processing conditions on its own.
Location privacy and social visibility
Location supports many of eBattery's features; however, not all of these features need the same level of visibility.
The location used to find nearby stations and the location shown to other users on the social map are separate operations. A user can keep their social visibility off while searching for nearby stations.
Social sharing is enabled by the user’s choice. When visibility is turned off, new sharing stops and the location record in the relevant app view is updated. The sharing duration, visibility scope and permission status are presented clearly in the app.
Sensitive location histories are not added to the public community profile. The private address of a home charger is not automatically made public just because the device is linked to the account.
Notification previews and shared cards are also covered by privacy. When a route card is shared, appropriate visibility controls apply to information such as the personal starting point.
Retention, deletion and user rights
Users can exercise their rights, such as obtaining information about the processing of their data, requesting correction of inaccurate information, and requesting deletion or destruction when the legal conditions are met. The right to object to a result that arises against them solely from automated processing is also part of the assessment processes. Applications under KVKK are concluded as soon as possible depending on their nature, and within thirty days at the latest.
Deletion is not treated as merely making a record invisible on the profile screen. Records in active data systems, related search indexes, caches and service providers are assessed within the data lifecycle.
Financial or transaction records subject to a legal retention obligation are kept for their own purposes and periods. This does not justify retaining all of an account's data indefinitely. For data whose processing conditions have ended, deletion, destruction or appropriate anonymization processes are applied.
The recovery design also ensures that deletion requests already completed are not undone when backups are restored.
International data flows
eBattery's global service goal may give rise to international data flows through cloud infrastructure, map services, notification systems and AI providers.
In these flows, it is not only the server where the data is stored that is considered. Support access from abroad, logs, backups and sub-service providers are also assessed.
Transfers abroad under KVKK are handled with an applicable adequacy decision or appropriate safeguard mechanisms. Standard contracts are one of the safeguards that can be used under the relevant conditions; the scope of the contract, the roles of the parties and the notification obligations are part of the transfer process.
Data protection obligations applicable in the other markets where it operates are also included in service design. The shared technical infrastructure is not operated on the assumption that the same legal processing conditions apply in every country.
Security incident management
For security incidents, the stages of detection, assessment, containment, remediation and safe restoration of the service are defined.
When examining the scope of an incident, the accounts, data types and transactions affected are identified. Response records are preserved; after the fix, changes that would prevent the same problem from recurring are evaluated.
Notification obligations apply in the case of personal data breaches. In Türkiye, the rules set by the Board provide that the Board is notified without delay and within 72 hours at the latest after the breach becomes known, and that the affected persons are informed as soon as reasonably possible after they are identified.
Security is not work that is considered complete once a new feature goes live. Code changes, external provider updates, access permissions and new data flows are assessed regularly. Findings from technical tests and operations reviews are addressed according to their priority, and the relevant controls are verified again.
