Founded 2020 | HQ Heidelberg, Germany | 40 employees (approx.) | <€2M revenues (approx.)
paretos is a German company offering a Decision Intelligence platform to help organizations make operational decisions without the overhead of building data models. With business users as the platform’s primary operators, paretos believes it is closer to the point of business decision-making than traditional business intelligence and analytics or popular task and process mining technologies.
Founded in 2020 by Thorsten Heilig (CEO) and Fabian Rang (CTO), Heidelberg-based paretos launched its platform in mid-2021, raising its first €3.5 million seed round in September 2021 with LEA Partners. September 2022 saw the company raise a further seed round of €6.5 million, this time led by UVC Partners, with participation from Interface Capital, angel investor and ex-Vodafone CEO Hannes Ametsreiter, and returning investor LEA Partners. With around 40 employees, paretos currently has approximately 20 paying customers and is generating revenues below €2 million.
paretos positions its platform as one designed for “Decision Intelligence,” or helping organizations make operational decisions without the overhead of building their data models. Decision Intelligence is hot right now; in the broadest sense it is a mashup of AI, ML, decision theory and mapping, BI, etc. that claims to augment – or in some cases automate – complex business decision-making.
paretos’ approach is based on its “decisionOS” – the layer that allows business users to interact with the underlying technology – which enables the connection between existing data sources from within the organization and can design and ultimately predict outcomes based upon that information (see Figure 1). This relies upon customers learning to trust the platform to perform those predictions and make decisions based on them, aided by insights that explain which elements have influenced the decision and how. “Socrates” is what paretos calls the range of algorithms they provide through this proprietary optimization layer, designed to interpret data and select models accordingly. paretos believes this method offers a significant amount of optimization per iteration, decreasing the time to value on projects.
paretos is a SaaS-based AI service that runs an optimization algorithm, or to be more accurate, a set of optimization algorithms. Although further optimized by paretos, Socrates is derived from research undertaken by Munich University of Applied Science (the CTO’s alma mater). This optimization algorithm is self-learning and continuously evaluates decision results, optimizing those results further through a human feedback loop. The algorithm is fed by data from the client’s business applications, and paretos comes with pre-configured integrations ranging from Google Ads to Snowflake. It’s critical at this point that the data is analyzed and cleaned, as with any AI system, the quality of the outcomes will depend on the quality of the data. paretos provides tools to help with this work and gives feedback and recommendations to get started.
Once the initial data-wrangling stage of the process is complete, an iterative optimization cycle – again familiar to most projects of this type – determines whether paretos has met a required standard before moving to review and evaluation. In theory the paretos interface can then be used to identify input/output relationships, predict future outcomes, and receive recommendations with optimal scenarios for the given decision. paretos algorithms run millions of scenarios for each given decision and provide only the optimal ones. This is different from other solutions because users can optimize for different target KPI instead of only one. Understanding tradeoffs between suggested courses of action, users can either decide the correct business decisions to make or automate that decision-making process.
By specifically targeting business users as the primary operators of the platform, paretos believes that it places itself closer to the point of business decision-making than traditional business intelligence and analytics or popular task and process mining technologies. Able to use current organizational data closer to real-time rather than fully retrospective analysis, paretos believes that its customers benefit from being able to make more impactful decisions in areas where those results can be immediately felt.
For this reason, paretos has designed a 3-month onboarding process for new customers to the platform to build that trust in the decision-making. It begins from an initial sales engagement – paretos says currently its early adopting customers tend to have an initial use case up front – and an assessment of existing data sources. This first stage is typically the most problematic for all projects of this type, and paretos says that its experience matches here, with the company offering an auditory data quality tool that can provide scoring of the suitability based on several parameters (completeness, cleanliness, diversity, quality, and freshness) and recommendations for how to improve those metrics. paretos says that it has over 100 existing data connectors to aid this part of the onboarding process.
At its current phase of life, paretos is working mostly reactively, allowing it to be challenged across a broad range of use cases. That supports paretos’ current overall approach to growth, which follows the well-established “land and expand” model popular with start-ups looking to scale. Similarly, the company is still at the stage of finding where an eventual vertical and/or horizontal position of comfort might lie. It’s a short-term strategy that demands not only flexible technology but also an energetic workforce and broad domain experience to succeed.
This approach has brought paretos a broad scope of current customer projects in industries including consulting, commerce, and logistics. Within these industries, paretos is addressing use cases such as optimized forecasting of vehicles and workforce (logistics), dynamic pricing and ad spend optimization (commerce), and personnel evaluation improvements (consulting). A broad range of experience is garnered here, which should help paretos make pragmatic decisions about where it sees its future go-to-market targets lying.
To that end, one of the critical elements of paretos’ technical roadmap over the next 12 months is the industrialization of key use cases – for both data input and outputs – gained from these existing projects, with a target of 10 by the end of 2023. Alongside this, paretos will tune the base models from those same experiences to improve initial results and inform Socrates’ ability to select models/optimizations through iterations. Both will help support the overarching desire to employ the platform for more valuable, causal use cases in the medium-term, for example: “What is happening, why is it happening, and how can I best act to address it?”
Historically, there has been a reluctance to trust “black box” technology to inform business decisions; for the most part, this has been for very sound reasons. Understanding how decisions are made and where you can examine the workings has underpinned the faith in them. paretos suggests that by taking your own operational data and allowing paretos to sprinkle its proprietary data science magic upon it, it will provide actionable insights across various use cases and industries, with pointers to how those insights were arrived at.
Organizations looking toward automated or semi-automated optimization now find themselves in a bind; the skills required to build in-house models on top of AI/ML platforms are scarce and, therefore, expensive on an ongoing basis. This is likely to push those without high-functioning data science teams (the overwhelming majority) toward a firm like paretos as a way of circumventing the problem. paretos’ fast start in building experience with deploying its technology in real work or real problems has allowed it to be bold with respect to where it believes it can add value. The next stage for the company is to decide where it should focus, and to add that value as it drives toward its self-directed panacea of identifying causality, which will help it direct its limited resources as a start-up to the best end.
Advice to Buyers
Organizations that know they will never be able to manage in-house data science teams face two broad options if they want to use that stratum of technology to inform their decision-making. One is to commission bespoke development from a third party with ongoing operational costs. The other is adopting a SaaS alternative where those costs are baked into a subscription. paretos provides a SaaS option, promising to help wrangle your operational data and iteratively improve its models using your data to inform decisions for your organizational context. Organizations that have already invested in finding suitable use cases to test could find paretos’ approach attractive, especially if they are able to guarantee their own data quality up-front.
- Strong momentum with diverse customers and industries
- Broad range of investors
- Looking toward identifying causality
- Having vertical/use case starter configurations
- Targeting organizations that lack data science teams or a desire to build them (which is most organizations)
- Insights that drive CxO needles within a few quarters; these are always easier to sell
- Has made headway in logistics, commerce, and consulting
- Strong initial revenue retention numbers
Attribution-NonCommercial-NoDerivatives 4.0 International
CC BY-NC-ND 4.0 license