From the IT perspective (the short version)

Intended audience: The IT department
Last modified: July 29, 2022

The Lokad app is a webapp provided as SaaS (Software as a Service). The purpose of Lokad is to deliver predictive analytics in order to optimize the supply chain (better stocks, better prices, etc.). The Lokad app is intended as an analytical layer that operates alongside transactional systems (ERP, WMS, CRM, etc.). It comes with a monthly subscription flat fee that typically bundles the app itself with professional services. Those professional services, provided by Lokad’s engineers (Supply Chain Scientists), alleviate almost entirely the need for technical support from the IT department itself for this scope. The one key contribution expected from the IT department is the setup of a data pipeline pushing flat files (by SFTP or FTPS) to Lokad, and potentially reintegrating the results generated.

Technical overview

The Lokad app is multitenant. Each tenant (i.e. client account) has its own dedicated file system and its own dedicated codebase repository. The filesystem is accessible through FTPS, SFTP and a web interface. This filesystem is geared toward large flat files (up to 100 GB per file) and features data versioning (like Git). The codebase repository is used to host Envision scripts. Envision is a proprietary DSL (Domain Specific programming Language) developed by Lokad. This DSL is heavily specialized for predictive optimization use cases. Envision scripts are used to perform the core numerical analyses (including machine learning algorithms, solvers, ...) and to generate data rich dashboards.

The app is redeployed in full every Tuesday between 10:00 and 14:00 (time of Paris). The downtime is typically kept under 5min. Lokad takes full ownership of all the versioning concerns.

The IT department is not expected to ever acquire any specific competency with Lokad’s stack. However, if you are curious, there is a complete technical documentation

IT contribution overview

We expect the IT department to set up a data pipeline that pushes a short series of relevant flat file extractions toward Lokad by SFTP or FTPS. The extractions are performed over the transactional systems (ex: ERP). We have a strong preference for raw table extractions (no filter, no join, no transformation), which requires minimal effort. From an ETL perspective, we only require the ‘E’ (extract) part under its simplest form (plain copy). Format-wise, Lokad is compatible with every reasonably tabular flat file.

The data pipeline is expected to run at least on a daily basis, and to be fully automated. The amount of work for the IT department depends on the data extraction scope (which systems? which tables?). However, as a rule of thumb, the data pipeline setup typically requires about 15 to 45 man-days, even for large companies. Once the data pipeline is in place, Lokad typically requires only minimal monitoring from the IT department, which is typically done with 1 or 2 man-days per month.

Security overview

The app is hosted in Microsoft Azure data centers located in the EU. We do not process any personal data , as we do not need such data to operate. When establishing the data extraction scope, we exclude any column or field that would contain personal data.

For authentication, our preference goes to SAML. We strongly suggest having users access Lokad via a federated identity such as Azure Active Directory, Office 365 or Google Workspace. This eliminates all the password-related problems.

Upon request, security audits and penetration tests can be performed by our clients. Details depend on the negotiated agreements.