Algolytics
Algolytics is a Polish end-to-end data science platform combining low-code MLOps, AutoML, a real-time feature store and address data quality, built for banks, insurers, telecoms, e-commerce and logistics teams automating credit scoring and fraud decisions.
What is Algolytics?
Algolytics is the data science and AI platform published by Algolytics Technologies Sp. z o.o., a Polish company based in Warsaw. It is not a single application but a suite of five products covering the whole chain: acquiring and cleaning data, building models, then deploying and running them. The publisher describes a comprehensive platform that supports data science teams in effectively implementing machine learning and artificial intelligence solutions that automate processes and help make accurate business decisions.
Scoring.One is the low-code MLOps layer, sold under the line Deploy ML faster. Less code. No DevOps. It imports models written as PMML, Java, Python, Groovy or R, and adds an integrated data store, native REST and SOAP APIs, message queues and streaming including Kafka, a data quality layer, versioning and A/B testing of scenarios, and parallel execution of several models inside a single request. Scenarios are drawn in a visual editor with pre-built connectors and decision rules, promoted from DEV through TEST and UAT to PROD, and each one is exposed as an API endpoint. Its reactive, non-blocking JVM engine built on Vert.x is claimed to serve 3.5x more requests per second than MLflow Server, with response times ten times faster.
Automatic Business Modeler is the AutoML brick, in three steps: import and configure, build, then use the model, with classification or approximation, sub- and oversampling, exploratory analysis, variable transformation and selection, and criteria such as ROC Area, Lift, Accuracy, Captured Response, Precision, Recall and Profit. AdvancedMiner is the full data science workbench: logistic regression, MARS, trees, random forests, XGBoost, bivariate probit, SOM, multilayer perceptron, ensembles, k-means and Kohonen segmentation, neural networks, time series, survival analysis and graph models, plus a battery of statistical tests, on a client-server architecture with the proprietary GDBase analytical database, JDBC/ODBC connectivity and scripts mixing Python, SQL, Java, Groovy and R. Event Engine is the feature store, refreshing aggregates and profiles on every event, with scoring under 5 ms and up to 20,000 events per second. Data Quality handles address standardisation, deduplication and golden records, geocoding and reverse geocoding, enrichment and autocompletion.
Four packaged business solutions sit on top: credit scoring, fraud detection, location intelligence and customer intelligence. One scope limit matters: the address databases and spatial data cover Poland.
What it does
- Deploy a predictive model to production in days instead of weeks, with no DevOps work.
- Score in real time, with claimed processing under 2 ms and thousands of requests per second.
- Build a machine learning model without programming, through the Quick, Gold or Advanced modes of ABM.
- Orchestrate a visual low-code decision workflow, from data access through to calls to external systems.
- Standardise, deduplicate, geocode and enrich address data.
- Autocomplete an address form in real time.
- Aggregate streaming events and maintain real-time customer profiles in a feature store.
When to use Algolytics / When not to
A quick filter to help you decide if Algolytics is the right fit.
When to use Algolytics
- Data science and machine learning teams that must move models from prototype to production without building their own MLOps layer.
- Banks, lenders, insurers and telcos automating a real-time risk decision, where the pre-built models claim ROC up to 85% and a decision in under 5 seconds.
- E-commerce and logistics teams dealing with dirty customer addresses, who need standardisation, deduplication, geocoding and address autocompletion.
- Business analysts with no programming or advanced statistics background, since ABM builds a model through a guided AutoML workflow.
- Organisations with no DevOps team, or that want a European vendor and Polish national address references such as TERYT, EMUiA and PNA.
When not to use Algolytics
- Teams that need geocoding or location intelligence outside Poland: the address databases and the spatial data cover Poland.
- Anyone shopping for a large language model or generative AI, since this is classical predictive machine learning (regression, trees, XGBoost, neural networks, time series, graph models).
- Buyers who want a displayed price: only Data Quality publishes a rate card, while Scoring.One, the ABM full version, AdvancedMiner and Event Engine go through a sales conversation.
- Pure self-service buyers: four of the five products have no online sign-up, only a demo request form.
- Procurement and compliance teams that need documented artefacts, as the site publishes no terms and conditions, no DPA, no product-level subprocessor list and no certification such as ISO 27001 or SOC 2.
How to use Algolytics
A typical end-to-end flow, from setup to results.
- Pick the product that matches the need: Scoring.One for MLOps, ABM for AutoML, AdvancedMiner for data science, Event Engine for a feature store, Data Quality for addresses.
- For Data Quality, create a free account on the web application at app.dataquality.pl/rejestracja, with 1,000 records offered to get started.
- Or request API access and run your first tests from the developer portal at developer.algolytics.pl/signup, using the web app and API documentation on GitBook.
- For Scoring.One, start the self-service free trial at trial.scoring.one/user/registration.
- Design the decision scenario in the visual low-code editor, wiring data access, rules and models together.
- Test the scenario in real time, then promote it from DEV to PROD; each scenario becomes an API endpoint.
- For ABM, request the demo through the form on the product page, then log in to the application at e-abm.pl.
- Follow the three-step ABM workflow: import a CSV file or an external database and set the parameters, let the tool build the model, then use it for offline scoring, scoring code or export to Scoring.One.
- For AdvancedMiner, download the free version after filling in the form; the full version is licensed through the contact form, and Event Engine demo access is requested by form as well.
- For a general commercial demo, use the contact page form to talk to the publisher's experts and see the Algolytics Platform in action; product documentation is centralised at algolytics-technologies.gitbook.io/algolytics.
Pros & Cons
Pros
- A complete chain from a single vendor: data preparation, AutoML, advanced data science, feature store and MLOps.
- Performance that is quantified and benchmarked: under 2 ms of processing, thousands of requests per second, 20,000 events per second, and 3.5x MLflow Server with ten times lower latency.
- Low-code taken seriously: models can be changed without mobilising IT, and deployment requires no Docker or Kubernetes.
- A public and detailed rate card for Data Quality, including a worked calculation example.
- A permanently free version of AdvancedMiner, whose usage period is stated as unlimited, alongside a self-service free trial of Scoring.One.
- An API documented on an Azure API Management developer portal, with product documentation on GitBook.
- A European publisher with the website served from Poland, a detailed GDPR-structured privacy policy, genuine specialisation in Polish address data (TERYT, EMUiA, PNA) backed by a published comparison report on 2,500 addresses, and named customer references in banking, telecoms, insurance, e-commerce and logistics.
Cons
- No terms and conditions are published: the full sitemap of 63 URLs contains none, and the privacy policy, which carries neither an effective date nor a version number, is the only legal page.
- Public prices cover Data Quality alone, while Scoring.One, the ABM full version, AdvancedMiner and Event Engine display no amount at all; the pricing page itself is linked from no menu and is reachable only through the sitemap or a search engine.
- No DPA is published or announced as available on request, no product-level subprocessor list exists, and no security certification such as ISO 27001 or SOC 2 is mentioned.
- No documented option to exclude customer data from model training, and nothing about how the publisher may use customer data.
- No hosting country and no hosting region is announced for customer data.
- The geodata, the address databases and the location intelligence cover Poland only.
- Uneven content and small errors: the detailed case studies, the careers page and the subsidised projects page exist only in Polish, an English navigation button is still in Polish (Dowiedz sie wiecej), the privacy policy misspells one of its own addresses (gdrp@algolytics.pl in section I against gdpr@algolytics.pl in section V), and no support email is published, only the general info@algolytics.com and info@algolytics.pl.
Pricing & Plans
A free plan exists: the AdvancedMiner Demo version has a usage period stated as unlimited, the ABM Demo is free for 14 days, Data Quality offers 1,000 records at no charge, and Scoring.One has a self-service free trial. The lowest published price point is 0.0075 PLN (0.75 gr) net per API request, and it applies only to the address autocompletion service of the Data Quality product, in the 1,001 to 100,000 records per month tier. Data Quality is in fact the only product with a public rate card: prices are net amounts in grosze (1 gr = 0.01 PLN) per single API request, charged on cumulative volume tiers. In that same 1,001 to 100,000 tier the seven services are priced at 7.5 gr for standardisation, geocoding and TERYT identifiers, 7 gr for building and resident data, 7 gr for surroundings data, 1.5 gr for financial risk, 0.75 gr for autocompletion, 1.5 gr for route calculation and 7.5 gr for reverse geocoding; the same seven fall to 5.25 / 4.9 / 4.9 / 1.05 / 0.53 / 1.05 / 5.25 gr between 100,001 and 500,000 records, and to 3.15 / 2.94 / 2.94 / 0.63 / 0.32 / 0.63 / 3.15 gr above 500,000. The first 1,000 records are billed at 0 gr for all seven services, although the page is internally inconsistent about whether that allowance is monthly or one-time. The publisher states that the amounts are net per single API request and that it does not charge for data that could not be recognised, corrected or geocoded, and its own worked example prices the geocoding of 200,000 records at 0 PLN for the first 1,000, 7,425 PLN for the next 99,000 and 5,250 PLN for the last 100,000. Buyers should note that one autocompletion request equals one alphanumeric character sent, so a complete address form generates roughly 12 to 15 requests, about 0.19 to 0.23 PLN per form at the 200,000 requests per month tier, and that surroundings data is available only in the web application. No amount is published for Scoring.One, for Event Engine, for the ABM full version, described as a subscription fee and as pay as you go, or for the AdvancedMiner licence.
- free
- 14 days of access
- all features
- limit of 100 models per month
- 500 MB of storage
- 24/7 access
- basic support.
- subscription
- also described as pay as you go
- all features
- limit of 1
- 000 models per month
- 10 GB of storage
- 24/7 access and a dedicated account manager
- price on request.
- unlimited usage period
- CSV
- TXT and MS Excel formats
- graphical interface
- SNA module
- custom scripts and processes
- maximum 1 GB of RAM
- one task at a time
- licence
- CSV
- TXT and MS Excel formats plus JDBC and ODBC databases
- graphical interface
- SNA module
- custom scripts
- unlimited RAM
- unlimited concurrent tasks
Data, GDPR & hosting
A consolidated view of how Algolytics handles your data.
GDPR overview
GDPR implementation is concrete and documented. The privacy policy is built on Regulation 2016/679 and on the Polish act of 29 August 1997, and it names its legal bases article by article. Rights are listed: access, rectification, erasure in certain situations, restriction, objection, portability where the data is necessary for the service, and a complaint to the President of the UODO (2 Stawki Street, 00-193 Warsaw). Rights are exercised by email at gdpr@algolytics.pl. Personal data is in principle not transferred outside the EEA, although ICT subcontractors may trigger such transfers; for the United States the publisher declares it works only with companies participating in the Data Privacy Framework. Technical and organisational measures are declared. Two gaps: the policy carries no effective date and no version number, and no DPO is named. No Article 27 representative is designated, which is expected for an EU-established company.
Who owns the data?
No terms and conditions are published, so the site never states who owns the data or the models loaded into the platform, nor what the publisher may do with them. The only legal document available is the privacy policy, and it covers the personal data of people using the publisher's websites and services rather than the customer data processed inside the platform. It names Algolytics Technologies Sp. z o.o., Przeskok Street no 2, 00-032 Warsaw (NIP 701-080-13-66, REGON 369456263), as the controller of that personal data, and points data protection requests to gdpr@algolytics.pl. Ownership of customer datasets, models and scoring outputs therefore has to be settled in the contract, because nothing public settles it.
Reuse rights
Because no terms and conditions are published, the site says nothing about whether a customer may reuse, export or redistribute the data and the models processed on the platform, and nothing about what the publisher may do with them. The privacy policy only describes how the publisher handles personal data collected from website visitors and business contacts. It invokes four legal bases: consent (art. 6(1)(a)), performance of a contract (art. 6(1)(b)), legal obligations such as accounting and tax duties (art. 6(1)(c)) and legitimate interest (art. 6(1)(f)) covering promotion of the services, communication, audience analysis, abuse prevention, statistics and the defence of claims. Data supplied is described as necessary to conclude contracts; for a contact designated by their employer it covers surname, first name, role, email and telephone. Named recipients include lawyers and legal advisers, public authorities, organisational and IT providers, an auditor, customer service providers, marketing campaign partners, PayU S.A. (Poznan) for payments and an online recruitment portal. Third-party cookies come from Google Ireland Ltd, Consentmanager AB, LinkedIn Inc., Meta Platforms Ireland Limited and Hubspot Ltd, in three families: necessary (legitimate interest), statistics and marketing (both by consent). The publisher states that personal data is not used for automated decision-making or profiling producing legal or similar effects.
Data retention & training
Hosting summary
The site announces no hosting country and no hosting region for customer data. There is no trust or security page, and no mention of a data centre, a cloud region or a hosting provider anywhere. The only related statement comes from the privacy policy, and it concerns personal data collected from website visitors and business contacts: such data is in principle not transferred outside the EEA, but the publisher writes that its cooperation with subcontractors in ICT services and infrastructure may result in transfers outside the EEA. For the United States it declares that it works only with companies participating in the Data Privacy Framework. One network fact, observed rather than declared: the website resolves to 188.210.221.84, an address located in Poland (Torun, AS50599 DATASPACE P.S.A.). That describes where the website is served, not necessarily where platform customer data is processed or stored. Buyers with a data residency requirement should have the hosting location, the subcontractor list and the transfer mechanism written into the contract, since the site commits to none of them.
Things to keep in mind
Risks and trade-offs to weigh before adopting Algolytics.
- Automated risk decisions land on real people: a credit or fraud score returned in seconds can refuse a loan or block an order, so the pre-built models, the 400+ ready-made anti-fraud rules and the behavioural signals (device fingerprinting, network and IP, email and phone, location, keyboard and mouse behaviour, social networks) call for human review, explainability and bias monitoring on the customer's side.
- No terms and conditions are published, so the contract governing use of the platform, including anything about ownership of the data and models you upload, cannot be read before a sales conversation.
- The privacy policy carries no effective date and no version number, so changes to it cannot be tracked, and its scope is site visitors and business contacts rather than, explicitly, the data customers process inside the platform.
- Nothing is documented about hosting: no country or region is announced for customer data, the policy admits ICT subcontractors outside the EEA, and no DPA, no product subprocessor list and no security certification is displayed.
- No opt-out from model training is documented, and the site says nothing about whether the publisher may use customer data, a point worth settling in writing before uploading sensitive datasets.
- Data Quality billing can surprise: prices are net and expressed in grosze, and one autocompletion request equals one alphanumeric character typed, so a single address form burns 12 to 15 requests. The pricing page is linked from no menu, and it contradicts itself on whether the free 1,000-record allowance is monthly or one-time.
- Peripheral signals to weigh: geodata and location intelligence stop at the Polish border, the detailed content exists only in Polish, a Polish label survives in the English navigation, the privacy policy misspells one of its own contact addresses (gdrp@algolytics.pl against gdpr@algolytics.pl), and no incorporation date is published for the company, whose earliest archived capture of the domain is a parking page.
Setup & Integrations
Technical difficulty
Effort varies by product. Data Quality is the easiest: create an account on the web application, or sign up on the developer portal and call the API immediately. Scoring.One offers a self-service trial, then visual low-code scenario design and one-click deployment, with no Docker, Kubernetes or infrastructure provisioning. ABM targets analysts with no programming or advanced statistics. AdvancedMiner is an expert tool: client-server architecture to install, scripting in Python, SQL, Java, Groovy and R. Event Engine is an engineering project of event integration, endpoints and aggregates. Four products out of five require a request form before any access.
Deployment
Integrations
Supported languages
Behind Algolytics
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Algolytics.
Frequently asked questions
What is Algolytics and which products does it include?
Is there an API and public documentation?
Can I try Algolytics for free, and is there a permanently free version?
How much does Algolytics cost?
What does the platform integrate with?
What performance does the publisher claim?
Is customer data hosted in Europe?
In which countries does the geocoding work?
Does the site publish terms and conditions?
How do I contact the publisher?
Should you pick Algolytics?
Algolytics is a mature, deeply technical stack rather than a single app. Its real differentiator is vertical integration: one publisher covers data preparation, AutoML, advanced data science, a feature store and low-code MLOps, with figures it is willing to put on the record, from processing under 2 ms to 20,000 events per second and 3.5x more requests per second than MLflow Server. Add a genuine specialisation in Polish address data, a published comparison report on 2,500 addresses, and named references at banks, telecoms, insurers, retailers and logistics operators, and the engineering credibility is not in doubt. The company traces its lineage back to StatConsulting, whose name its YouTube channel still carries.
The brake is commercial and documentary opacity. There are no terms and conditions to read before signing. Public prices exist for one product out of five, on a page no menu links to. Nothing is committed in writing about where customer data is hosted, about a DPA, about subprocessors, or about whether customer data feeds model training. The privacy policy is thorough on visitors and business contacts, but it carries no effective date and does not clearly cover the data customers process inside the platform.
Expect a classic B2B purchase: demo, scoping, quotation. Only Data Quality and the Scoring.One trial can genuinely be adopted alone, and Data Quality is also where pricing is legible, from 0.0075 PLN net per autocompletion request in the 1,001 to 100,000 records per month tier. One geographical caveat matters: the machine learning platform is usable anywhere, but the address databases, geocoding and location intelligence cover Poland. For a European team building risk, fraud or customer decisions, it is a serious candidate, provided the contract fills the gaps the website leaves open.
- Choosing a selection results in a full page refresh.
- Opens in a new window.