Sift Lab
Sift Lab is a Swedish AI platform that turns a retailer's first-party customer data into predictions, segments and ready-to-activate audiences. It layers predictive analytics, an agentic assistant and a recommendation engine on top of an existing data warehouse.
What is Sift Lab?
Sift Lab is a customer intelligence platform for retailers and e-commerce companies, published by the Swedish company Sift Lab AB. It grew out of academic research: the platform is founded on work by Martin Rosvall, professor of computational physics at Umeå University, and the company still maintains a research collaboration there. It was founded in 2015 by Jakob Sjölander, Christian Persson and Martin Rosvall, and is now led by chief executive Linda Hellström.
Technically, the platform describes itself as AI-native rather than as another business intelligence tool. At its core sits a proprietary distributed columnar database built on DuckDB, tuned for data density and snapshot optimisation. Rather than replacing a customer's warehouse, it sits directly on top of existing infrastructure such as Snowflake, BigQuery or Databricks and unifies customer, product and channel data into a single semantic layer. New data streams in continuously and becomes queryable within milliseconds, so dashboards, segments and predictions run on current data.
On top of that core, the product is organised into eight modules: all-in-one analytics, the Sift Sense AI agent, AI audiences, customer segmentations, on-site recommendations, email recommendations, an AI-powered customer data platform and retail media. The modelling combines time-series methods with a transformer and LLM-based recommendation engine using multimodal embeddings and GPU acceleration, predicting churn, next-best-action, lifetime value and product preference.
Sift Sense is the agentic layer. It reasons across the semantic model, answers ad hoc questions about customers, products and campaigns, and can build an audience and trigger it into a marketing channel on request. Answers are presented as traceable, and the agent runs on Google language models with the customer deciding what is shared.
Sift Lab claims sub-second queries roughly 10 to 100 times faster than traditional BI tools, costs up to 36 times lower than warehouse-centric architectures, more than 100 ready-to-deploy use cases, and an average 200% improvement in campaign performance in the first year. None of these figures is accompanied by a published methodology. Access is through a web application and a documented recommendations REST API; the company reports serving more than 30 of the largest Nordic retailers.
What it does
- Predict churn, next-best-action, customer lifetime value and purchase propensity from first-party data
- Build AI-generated customer segments such as RFM, personas, life cycle and cohorts
- Create predictive audiences and push them to email, SMS, Meta, Google and TikTok
- Ask questions about customers, products and campaigns in natural language through the Sift Sense agent
- Serve personalised product recommendations on-site and inside email campaigns
- Explore sales, customer, product and channel performance in pre-built interactive dashboards
- Package and monetise first-party audience insight for brand suppliers through retail media
When to use Sift Lab / When not to
A quick filter to help you decide if Sift Lab is the right fit.
When to use Sift Lab
- Retail and e-commerce CRM teams that want to build and activate predictive audiences without waiting on a data science team
- Marketing and growth managers running campaigns across email, SMS, Meta, Google and TikTok who need better targeting than rule-based lists
- Data and analytics teams that already run Snowflake, BigQuery or Databricks and want a customer-facing layer on top rather than another warehouse
- Merchandising and category buyers who need product, returns and assortment insight tied to actual customer behaviour
- Retailers building a retail media business who want to package predictive audiences and insight for their brand suppliers
When not to use Sift Lab
- Small businesses or solo operators looking for a self-service tool, since there is no public pricing, no free plan and no trial
- Organisations without an existing customer data foundation, as the platform is designed to sit on top of a warehouse or structured imports
- Teams outside retail and e-commerce, because the dashboards, segmentations and models are built around purchase and product behaviour
- Buyers who need a mobile application, as no iOS or Android app exists
- Procurement processes that require published security certifications or a standard data processing agreement up front, since neither is available on the site
How to use Sift Lab
A typical end-to-end flow, from setup to results.
- Request an introduction through the Book demo form, which is the only public entry point, as there is no self-service sign-up
- Agree commercial terms with the sales team, since pricing is quoted rather than published
- Connect the platform to your existing data layer, typically Snowflake, BigQuery or Databricks, or import files directly
- Map your data to the published specification covering interactions, products and users
- Let the semantic layer unify customer, product and channel data into a single model
- Open the pre-configured dashboards for sales, customer, product, marketing and supplier or inventory analysis
- Filter by clicking any element in a dashboard, or build your own dashboards and reports
- Generate segments and predictive audiences, using RFM, personas, life cycle or cohort logic
- Export an audience in one click to the marketing channel of your choice, or automate it in a recurring journey
- Generate an API key from the Account menu to pull recommendations programmatically into your site or email campaigns
Pros & Cons
Pros
- Unusually transparent technical documentation, describing the database architecture, semantic layer and modelling approach in public
- Sits on top of an existing warehouse instead of replacing it, which limits migration work
- A genuine, documented recommendations REST API with key-based authentication, parameters and request examples
- Verifiable academic roots, with a named founding professor and a continuing collaboration with Umeå University
- A long list of named Nordic retail references, including Nordic Nest, Rusta, Byggmax, Bokus and NA-KD
- An established company, trading since 2015, active, with filed accounts and PwC as auditor
- GDPR compliance claimed by a publisher established inside the European Union
Cons
- No public pricing at all, no free plan and no trial, so the cost cannot be assessed without contacting sales
- The privacy policy dates from May 2021 and still invokes the Privacy Shield, invalidated in 2020, alongside a security measure promised for the first half of 2025
- No terms and conditions, no legal notice and no postal address anywhere on the site
- No data processing agreement and no subprocessor list published, which complicates procurement review
- No security certification is claimed, neither ISO 27001 nor SOC 2
- Performance claims such as 10 to 100 times faster, 36 times cheaper and 200% campaign uplift come without any published methodology
- Visible site neglect: the Use Case Library still contains Lorem Ipsum placeholder entries and the Careers link in the menu is broken on every page
Pricing & Plans
Sift Lab publishes no pricing. There is no pricing page on the site, none appears in its sitemap, and no amount, currency or billing unit is stated anywhere on the site or in the product documentation. No permanent free plan and no free trial are offered or mentioned. The only routes available to a prospective customer are the Book demo, Talk to sales and Get started calls to action, all of which lead to the same contact form. Pricing is therefore quoted commercially on a contract basis, and any budget estimate requires direct contact with the vendor.
Data, GDPR & hosting
A consolidated view of how Sift Lab handles your data.
GDPR overview
GDPR compliance is claimed explicitly. The privacy policy lists compliance among its data protection mechanisms, and the Sift Sense page describes the agent as secure and GDPR compliant, with the customer choosing what is shared with the underlying Google language models. Stated legal bases are contract performance, legitimate interest and consent. Data subject rights are set out: access through a free annual register extract, rectification, erasure, restriction and withdrawal of consent, with the Swedish supervisory authority named as recourse. As the publisher is established in Sweden, no Article 27 representative is required and none is named. Two caveats matter: the policy is dated 17 May 2021 and has not been revised since, and it still relies on the EU-US Privacy Shield, invalidated in July 2020.
Who owns the data?
The only public document is the website privacy policy, and it names Sift Lab AB as the controller for personal data collected through the site, apps and services it administers. It covers visitors and prospects: names, contact details, IP address, location and browsing behaviour, which the company may enrich from Swedish public registers such as SPAR. Sift Lab states it may share this data with companies that process it on its behalf, but it names none of them. Crucially, no public contract governs the customer data loaded into the platform itself, and no data processing agreement is published, so ownership and permitted use of that operational data are settled privately in the commercial contract.
Reuse rights
Sift Lab publishes no terms and conditions, so there is no public statement on whether a customer may reuse the data or outputs produced by the platform. What the privacy policy does describe is the company's own use of website and prospect data: fulfilling service orders, customer service, analysis and market research, segmentation and profiling, and marketing by SMS, email, app, push, social media or post. It states plainly that automatic profiling is performed on web activity and that the resulting profile may be used for marketing. Consent can be withdrawn and marketing opted out of at any time. For data held inside the platform, reuse rights are not addressed publicly and would need to be confirmed in the commercial agreement.
Data retention & training
Hosting summary
Sift Lab does not disclose where customer data is hosted. No hosting country, no region, no cloud provider and no data centre is named anywhere on the site or in the product documentation, and there is no trust or security page. The one adjacent statement in the privacy policy is a transfer clause about partners: the company says it only works with partners who process personal data within the EU or EEA, or with companies maintaining an equivalent level of protection, citing the Privacy Shield framework. That is a statement about processors, not a declaration of hosting location, and the framework it invokes was invalidated in July 2020. On security measures, the policy claims encryption in transit and at rest using industry-standard protocols, role-based access control, data minimisation and a structured incident response plan, with multifactor authentication described as being implemented. Any buyer with a data residency requirement will need to establish the hosting jurisdiction directly with the vendor, as it cannot be determined from public sources.
Things to keep in mind
Risks and trade-offs to weigh before adopting Sift Lab.
- The platform profiles individual shoppers to predict churn, spend and product preference, so retailers must ensure their own legal basis and transparency towards customers, since Sift Lab's public policy covers only its website
- No data processing agreement or subprocessor list is published, leaving a material gap for anyone whose compliance process requires them before customer data is loaded
- The privacy policy has not been revised since May 2021 and still relies on the Privacy Shield, invalidated in 2020, so the stated safeguards for transfers should not be taken at face value
- Automated segmentation and next-best-action can harden into self-fulfilling targeting, quietly excluding customers the model scores as low value unless humans review the segments
- Trusting an agent's natural-language answers without checking the underlying figures invites confident but unverified decisions, and the traceability the vendor offers is only useful if someone actually uses it
- Performance claims of 10 to 100 times faster, 36 times cheaper and 200% campaign uplift carry no published methodology and should be validated during a pilot rather than assumed
- Because pricing is entirely private and the tool becomes embedded in the data and marketing stack, switching costs can build up well before the commercial terms are ever benchmarked
Setup & Integrations
Technical difficulty
Moderate, and not self-service. Onboarding runs through a sales conversation, then a technical connection to your existing data layer such as Snowflake, BigQuery or Databricks, or a file import mapped to the published data specification covering interactions, products and users. That first step needs data engineering involvement. Day-to-day use is designed to be code-free, with pre-configured dashboards and one-click audience exports, while SQL and the semantic layer remain available for deeper work. Pulling recommendations into a website or email campaign requires generating an API key and a straightforward REST integration.
Deployment
Integrations
Supported languages
Behind Sift Lab
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does Sift Lab actually do?
How much does Sift Lab cost?
Is there a free trial or a free plan?
Do I have to replace my existing data warehouse?
Does Sift Lab offer an API?
Which marketing channels can audiences be activated in?
What is Sift Sense?
Is there a mobile app?
Who is behind Sift Lab?
Is Sift Lab GDPR compliant?
Should you pick Sift Lab?
Sift Lab is one of the more technically credible customer intelligence platforms in the Nordic retail market. Its architecture is described publicly and in unusual detail, its academic roots at Umeå University are verifiable, its recommendations API is genuinely documented, and its customer list names well-known Nordic retailers. For a retailer that already runs Snowflake, BigQuery or Databricks and wants a predictive layer on top rather than another warehouse, the positioning is coherent and the promise is specific.
The reservations are mostly commercial and contractual rather than technical. Nothing about the cost is public: there is no pricing page, no free plan and no trial, so a buyer cannot form even a rough budget without entering a sales conversation. The legal documentation has fallen behind the product, with a privacy policy dated May 2021 that still relies on the Privacy Shield framework invalidated in 2020, no terms and conditions, no published data processing agreement and no subprocessor list. For a tool whose entire value rests on processing customer data, that gap will matter to any serious procurement review. No security certification is claimed either.
A few details suggest a small team stretched thin: the Use Case Library page is still filled with Lorem Ipsum placeholders while the site advertises more than a hundred use cases, and the Careers link in the main menu is broken on every page. The headline performance figures are striking but come without any published methodology, so they should be treated as vendor claims to test during a pilot rather than as established results. Sift Lab is best approached as a contract purchase where the technical fit is strong and the contractual detail has to be established directly with the vendor.
- Choosing a selection results in a full page refresh.
- Opens in a new window.