Frequently asked questions
Q: What are the steps to integrate with the web store?
A: The integration to an online store includes the following steps: - Data integration - Data preparation - Setting up the environment for the Suite - Installation - Model release - Data serving
Q: What data do we send?
A: Minimum viable dataset includes: - Transaction data - for example: invoices and invoice items, with amount, quantity, customer and date they refer to - Product information - ID, name, description, category and super-categories Although, for our Studios to work without issues we have defined a data model needed, where mandatory, nice to have and optional data is listed, you can check it here: [URL]
Q: On the basis of which data do we suggest products (are they based on the features, model silhouettes, product name, sizes or something else) in order to assess how good the input data on the products are?
A: Currently, recommendations are determined on the basis of similarity between customers (similarity in activities and preferences extracted from transactions), and on the basis of similarity between products (description, categories, etc.)
Q: Are all the functionalities performed within your platform (e.g. sending a message is initiated within your platform) and in that case, can we keep the integrations we have with other companies?
A: All the functionalities are operationable within our platform; integrations with other systems/parties can remain unhindered. Besides, the platform is modular, so the client can choose what to use from our platform and what from other systems, since we also support API integration and communication with external systems (i.e. CRM).
Q: How much history is needed?
A: It would be nice to capture the periodicity, thus at least a couple of sales cycles
Q: Recommendations in real time on the online shop, how does that part work?
A: The data is sent in real time through Interactions api or EventSDK, and then with appropriate processing, we use it as input for the ML model. The recommendation results themselves are served through the Touchpoint studio.
Q: How can the existing loyalty platform and the new tool be integrated integrated?
A: We can easily integrate through the API. Depending on the use case, we support pulling the data from the existing loyalty platform, and sending outputs that the loyalty platform could consume.
Q: Does your tool connect directly to the customer messaging platform or databases are sent manually?
A: The tool can be integrated with existing messaging platform via API; there is also a possibility to send messages directly from the tool.
Q: Is data processed by you tool stored on one of your servers or is the tool using our existing database to connect to and collect the data?
A: Data is most often exported and stored in the cloud (object storage, S3); if that is not an option for the client, then we also support connecting to the existing database, where the access should be provided.
Q: Due to the legal regulations on the protection of customers' personal data, is it necessary to change the existing application database, so that we remain covered in this matter?
A: Most of outputs and insights generated within out tool are based on the customer ID (hashed ID). We are fully GDPR-compliant, given that we work with anonymized data. Data on the identity and link to the ID with the specific customer are and stay on the client's side and are not subject to our processing.
Q: How is the propensity to purchase calculated?
A: If you refer to predicting whether the customer will make a purchase in a future period - the propensity to purchase model is available. In addition to this, it is possible to predict the future monetary value of the customer - customer lifetime value (CLV).
Q: What are the main challenges when implementing the entire system?
A: Most often, the biggest challenge turns out to be collecting a sufficient amount of good quality and representative data satisfying our data model.
Q: Can you come up with an explanation as to why something was recommended?
A: The reasoning for recommendations given directly depends on the selected algorithm/strategy. For some more traditional models, it is much easier to explain how the model geenrated the output, compared to some other, more complex models. For these more complex models, we can discuss in more detail the idea and way of functioning.
Q: How cannn we integrate the data coming from CallCenter (data base access provided)?
A: We can connect to the database and collect it, the store to our object storage. Otherwise, we can provide you the guidelines for creating a script to export the data to our object storage.
Q: How does the realization of the integration process look like?
A: Integrations involves installing the platform and all corresponding services, data integration, data quality control, scheduling the automated pipelines for ETL and ML models and ingesting the results to our databases in order to serve it and consume it within the GUI or in some cases within other systems.
Q: What are the critical points, risks for realization?
A: Critical points relate to the integration itself in terms of the automation of data arrival and data quality.
Q: Operational usage after the introduction, how does that part work?
A: The tool can be used in several ways, by integrating through the API into external systems, as well as using it through a graphical interface. Use guide is provided withint he business documentation shared in the onboarding phase. The API is created according to the Open API 3.0 standard and there is a detailed documentation for every service/module.