UnityBase for B2B Startups: From MVP to Scaling
For the first client of a B2B service, a simple dashboard where they can create a request, attach a file, and check the status is sufficient. The second client will need separate queues for departments, and the third will require an exchange of requests with their own accounting system. A platform for an MVP should be evaluated against these subsequent requirements: will the team be able to add them while users are working with real data every day?
UnityBase is a Ukrainian low-code platform for developing information systems. It provides developers with a shared data model, tools for standard operations, and a server-side JavaScript environment for application rules. Let us examine how this works using a hypothetical request service example.
How to scope the MVP
For a pilot, one end-to-end process is enough. A client submits a request, an operator assigns an assignee, and the client receives the result in their dashboard. The team describes organisations, requests, and assignees in the UnityBase metadata. A shared description of entities, attributes, and relationships is used by tools for working with the database schema, APIs, and the administrative interface.
Standard forms can be used in the operator's dashboard. For the client, the team can create a separate interface with a short form and request history. Product rules are implemented in server-side JavaScript. For example, it is the application code that determines who can close a request, when a comment is required, and how to process a repeat request.
What hypothesis should the pilot test
Before detailing the model, agree with the client on exactly what should change in their workflow. For example, requests currently arrive by email, an operator transfers them to a spreadsheet, and the client calls to find out the status. In the pilot, you can measure how many requests clients submitted themselves, how many of those operators processed without clarifications, and how often people still had to ask about the result. These observations will suggest whether it is worth developing the dashboard and which part of it to improve first.
It is worth adding a scenario to the demonstration where the process does not go according to plan. For example, a client cancels a request after an assignee has already started working on it. The team must agree on whether the client can do this themselves, what the assignee sees, and how the record in the queue changes. Clarifying such rules on a small number of scenarios will help separate the universal behaviour of the service from agreements with an individual client.
It is convenient to divide requirements into three groups. The first includes standard operations, such as maintaining a directory of departments. The second includes rules that determine the outcome of work, such as status transitions and mandatory checks. The third includes external dependencies, such as notification delivery. This division allows you to assess where the team can use platform mechanisms, where custom code is needed, and what depends on third-party services.
How to segregate access between clients
A client's employee sees their own organisation's requests, an assignee sees requests assigned to them, and an operator distributes work within their authority. UnityBase provides tools for controlling access to operations and server-side restriction of individual record visibility (Row-Level Security, RLS).
In the pilot, create two organisations. Log in under the first client's account and check whether the second client's request is protected when accessed via a direct link, through the API, and through a link to an attachment. For custom server methods and integrations, check separately how access rules are applied. If the client requires an audit, also check whether it is possible to determine who changed the status or the assignee.
Select a role. In this example, the request was submitted by the "Alpha" organisation, and the operator assigned it to the assignee, Olena.
"Alpha" employee
Request card — Available. The employee sees the status and can open permitted attachments.
Actions — Can add clarifications. Assigning an assignee is available to the operator.
"Beta" employee
Request card — Unavailable. The record does not appear in search or exports.
Direct link — The server does not return the content of the request or its files.
Olena, assigned assignee
Request card — Available in her work queue via the current assignment.
Actions — Can update the status within her stage. After the request is transferred, access is determined by the new assignment.
What changes when an API is added to the dashboard
For the integration of a client's accounting system, a separate account with its own permissions will be needed. If the integration only creates requests and retrieves their statuses, its access should correspond specifically to these operations and the data of a specific organisation. Also, check how restrictions apply to lists, searches, and exports. For example, an export file must contain the same permitted set of requests that the operator sees in the dashboard.
A minimal integration check covers a successful request, a retry after a lost response, and an attempt to access another organisation's request. For retries, you can agree on an external request identifier within the organisation. Then, two import attempts with the same identifier should lead to the expected result without duplicating the operator's work. The exchange log should contain enough information for support to match an error with a client's request and explain the next steps.
Visibility rules should be captured in a small matrix. For each role, specify which records it can read, which operations it can perform, and under what conditions. For example, an assignee can update the status only for their current assignment, while an operator can transfer it to another department. Such a matrix provides a common foundation for development, acceptance, and verification after changes.
How to change the model after launch
Suppose that in the pilot, a request had one assignee, but now the client needs to transfer it between departments and keep a history. The team adds an "Assignment" entity with a link to the request. UnityBase tools help update the schema and perform the migration, while the team defines the data transfer rules. In particular, they decide how to migrate current assignments and what to do with requests that have not yet been assigned to anyone.
Before the transition, perform a migration on a representative copy of the database. Verify whether a corresponding record has been created for every request with a current assignee, whether attachments open, and whether access rights are observed. If an external system still reads a field with a single assignee, determine what value it will receive during the transition. Agree on rollback conditions and the method for handling requests that arrive during the switchover.
Switch between stages to track how request No. 104 moves from a single field to the assignment history.
1. Before the update
Request No. 104 — Olena is recorded in the "Assignee" field. The client's accounting system reads this field.
2. Trial migration
Request No. 104 — The migration creates a linked record "Current assignment — Olena". The old interface still reads the previous field.
Reconciliation — Olena sees the request, the attachment opens, and the integration returns the same status.
3. After the switchover
Transfer to another department — Olena's assignment is completed. A current assignment for the new assignee is created.
History — The operator sees both assignments. The new one is used for the current queue.
How to carry out the change in stages
The first step is to find out who uses the old assignee field. Besides the dashboard, this could include reports, notifications, and external integrations. For each consumer, determine when they will switch to the new model and what value they should receive until then. For example, a report on current requests might take the last active assignment, whereas a report on transfers between departments needs the entire available history.
If continuity is important to the client, you can first add the new structure, then populate it with existing data and reconcile the results, and enable reading from the new model after verification. For the transition period, determine where new assignments go, how the team will detect discrepancies, and which source to consider valid. The order depends on whether compatibility with the old API can be maintained and whether a short window for switching is acceptable.
After the trial run, record the duration of each stage and the list of exceptions. For example, an old record might refer to an assignee whose account has already been closed. For such cases, an agreed-upon decision regarding history and current access is needed. IQusion's experience in modernising complex systems shows why application operations, attachments, and exchange with related systems are checked separately after migration.
The rollback criterion must be observable. For example, if active requests disappear from the work queue after switching or the integration stops receiving statuses, the team reverts to the agreed-upon mode. Determine in advance how to reconcile operations performed after the switch so that user changes are preserved when restoring operations.
What to measure before scaling
For the test, reproduce a real combination of operations. For example, while the integration imports a batch of requests, operators open the queue and clients upload files. Measure the queue opening time, error frequency, and exchange latency. Compare the results with the client's requirements and repeat the test on a larger volume of data.
If the queue opens slowly, check queries and indexes. If duplicates appear after an exchange interruption, check the handling of retry requests. These measurements will help determine where changes to the application code, database, or infrastructure are needed.
Select a mode to see what to measure and what conclusion can be drawn. The team will obtain the values in their own environment.
Operators work with the queue
Measure — The time taken to open the queue and save a status change.
Purpose — This is the baseline result for comparison with performance during import.
Integration imports requests
Measure — Batch processing time, exchange queue lag, and the number of errors.
Purpose — The run shows exchange behaviour while operators are not creating additional load.
Operators and import work simultaneously
Compare — Queue opening time and exchange latency with the results of individual runs.
If there is degradation — Check for locks and load on the database and storage during the same time interval.
How to distinguish platform limitations from specific solution limitations
Describe the test along with the result. Record the data volume, size and frequency of import batches, number of concurrent operators, environment characteristics, and measurement period. For response time, it is useful to look at slow queries alongside typical ones. An average value can hide a situation where the queue works quickly during the day, but delays increase during batch exchange.
Run interactive work and imports separately, then repeat them simultaneously. If the delay is noticeable only in the joint run, check which resources the operations are competing for. The cause could be database locks, storage limitations, or the way the batch operation is executed. After making a change, repeat the same run to compare results under identical conditions.
Evaluate recovery after a failure separately. For example, stop a test import in the middle of a batch and check whether the exchange can be resumed, whether the operator sees already accepted requests, and whether the link to records in the client's accounting system is preserved. For a B2B service, this is part of readiness for growth: as the number of integrations increases, the team must be able to restore operations and explain the state of the exchange.
When to consider UnityBase
UnityBase should be included on a shortlist if the product's foundation consists of related data, workflows, multiple roles, and integrations. If the most work is required by a non-standard interface or specialised calculations, assess what portion of this work is covered by the platform's tools.
To make a decision, implement one process with two organisations, connect one external system, and perform one model change on an already populated database. Record the time spent on setup and development, access verification results, and migration duration. Add the costs of deployment, maintenance, and team training. Such a pilot will show whether UnityBase is suitable for both the first launch and the subsequent change that a client will request.
What to include in cost estimation
When comparing platforms, use an identical scope of work. For a request service, this could be a dashboard, an operator queue, isolation of two clients, one external exchange, and a structure change on a populated database. Note which parts were successfully implemented using standard components, which required custom code, and how much time the verifications took. This will make it clear where the advantage at the start is maintained after the first update.
For maintenance, consider development and testing environments, backups, performance monitoring, integration support, and time for releasing changes. Check whether another developer will be able to recreate the environment, find an access rule, and repeat the migration following the team's instructions. If these tasks are performed only by the prototype's author, the time for knowledge transfer also belongs to the product's cost.
What should be the output of the pilot
- Working process. The client submits a request and receives a result, and the operator can process at least one exception to the standard process.
- Access and exchange verification. Two organisations work with their own data, the integration recovers after an interruption, and a re-transmission yields an agreed-upon result.
- Reproducible change. The team updates the model on a copy of the production data, reconciles the result, and knows how to return to the previous mode.
- Measured costs and performance. There is factual data on team effort, performance under load, and maintenance requirements.
Based on these results, the team can decide which parts of the product to leave on platform mechanisms, where to invest in custom implementation, and what to check before connecting the next client. The evaluation of UnityBase should be tied to how well the platform helps the team execute these changes in the context of their product.