Twenty years of Jedox, from an open-source cube to an AI planning platform

· Jedox

How Jedox got from a GPL-licensed OLAP server in Freiburg to the platform we implement today - the timeline, the four decisions that shaped the product, and what the analyst surveys actually say once you read the methodology.

Most people meet Jedox in the middle of its life. You get handed an instance, a model and a deadline, and the platform simply is what it is. But a lot of what looks arbitrary in Jedox makes sense once you know which decade it came from, and a few things that look modern are older than you would guess.

This is the long view: where the product came from, the decisions that changed its direction, and what independent reviewers actually say about it now. I have kept to sources you can check, and where Jedox's own pages contradict each other, I say so rather than picking the tidier version.

The timeline

Before the cube, 2002 to 2005

The Palo years, 2006 to 2013

Commercial, cloud and a rebuilt front end, 2014 to 2017

Four releases a year, 2018 to 2023

One major a year, 2024 to today

Four decisions that explain the product

Timelines show what happened. These are the turns that explain why Jedox behaves the way it does.

The spreadsheet was the starting point, not a compromise

Worksheet Server came in 2003, two years before the cube. The sequence matters: Jedox did not build a database and later bolt Excel onto it, it built a way to publish spreadsheets and later put a multidimensional engine underneath. That is why the Excel add-in has never felt like an afterthought, and why finance users take to the product faster than they take to most planning tools.

It is also why "get people off Excel" is the wrong pitch for a Jedox project. The honest pitch is that the spreadsheet stops being the system of record while staying the interface, which is a different promise and an easier one to keep.

The interesting recent movement is that Jedox has been carefully removing the spreadsheet from everything except data entry. Canvas took over reporting in 2021 and became its own report type in 2022.4, so a report is no longer a sheet inside a workbook. Dynatables took over analysis in 2024.1, then gained write-back in 2025.1. Twenty years on, the spreadsheet is still there, but it is no longer the only way in.

Open source paid for the distribution, then stopped paying

Palo was genuinely free software under GPL v2 and LGPL, and that is how a small company in Freiburg got an OLAP engine into more than a hundred countries. The licence changed at version 5.1, which Jedox dates to April 2014. Wikipedia's Palo entry dates the change to 2015, so if you need the year for something that matters, cite both.

What is worth noticing is that Jedox's own company timeline no longer mentions the open-source chapter at all. The 2005 entry describes Palo only as "its own in-memory OLAP server". The community forum threads where users argued about the end of the Community Edition are gone too, because the forum domain no longer resolves. This part of the history is quietly disappearing, which is a reason to write it down.

Cloud-only was an eight-year move, not a switch

Jedox Cloud launched in 2014. The last on-premises release was 2022.2 in May 2022. Between those two dates the company ran both, and the transition shows up in small places: for the final three shared versions the cloud and on-premises release dates differ by about six weeks, so any dated timeline has to say which table it is quoting.

A discrepancy worth knowing: the cloud lifecycle table gives 2022.2 an end of support of 24 May 2023, while the separate on-premises knowledge base says Jedox continues to support the final on-prem release with no published end date. If a client asks, quote both pages.

The operational consequence is the part consultants feel. On cloud, upgrades are not the customer's decision. Auto-Update is the mechanism: Cloud Console users get 14 days' notice, non-production instances update first with production following 7 or 14 days later, an instance can be rescheduled within 60 days, and on a critical failure Jedox states it addresses the issue rather than rolling back. Testing the updated instance remains the customer's responsibility. That single paragraph should shape how you plan regression testing, and it did not exist in the on-premises world.

Note also that cloud-only deployment is not cloud-only data. Integrator still reaches on-premises systems such as SAP ERP, usually over a VPN tunnel.

The AI story is four attempts, not one arc

This is the part most often flattened into a straight line, and it is more interesting than that.

2019 to 2020: machine learning, not language models. AIssisted Planning existed by early 2020, with an AI Transform exposing predictive forecasting inside Integrator. Time series prediction, driver-based prediction, classification. No conversation involved.

2023: an integration. In April 2023 Jedox announced a ChatGPT integration, which was a bridge rather than a product.

2024: the first generative features. 2024.1 shipped a generally available AI search in the Help Center, explicitly ChatGPT-based and carrying a warning not to share confidential information, plus natural language over Views flagged as preview.

2025: a step backwards, stated plainly. The 2025.1 notes record that the previous generally available version of JedoxAI was removed from the Start page, replaced by a beta customers had to contact Support to enable.

2026: named agents. 2026.1 ships a Knowledge Agent for product guidance and Integrator scripting, and a Reporting Agent for conversational analysis over the cube. Jedox's FAQ describes JedoxAI as generally available, and the agents have been extended monthly since May. There is also a separately licensed Jedox MCP add-on that exposes Jedox capabilities to MCP-compatible clients, though I could not find a dated announcement for it, so I would not claim a launch date.

I wrote up what 2026.1 contains, and what the Reporting Agent needs from your model before it produces anything useful, in a separate post.

The pattern across all four: Jedox ships AI early, pulls it back when it does not hold up, and ships it again in a better shape. Being willing to withdraw a generally available feature is unusual, and I would rather work with a vendor that does that than one that leaves a bad feature in place.

What the independent reviewers say

Analyst coverage is easy to quote badly. Here is what I think holds up, including the parts that are not flattering.

BARC's Planning Survey is a user survey, and Jedox's own results in it have moved in a genuinely interesting way:

Edition Result Peer groups
Planning Survey 17 (2017) 10 category wins, 1,687 respondents Development-oriented planning solutions
Planning Survey 22 (2022) 18 top rankings, 84 leading positions Six
Planning Survey 24 (2024) 4 top rankings, 19 leading positions Five
Planning Survey 25 (2025) 5 top rankings, 24 leading positions Five
Planning Survey 26 (2026) 18 top rankings, 32 leading positions Five

Jedox calls the 2026 edition its strongest result ever, topping Data Integration and Project Length in all five of its peer groups. That is a real improvement on two flat years.

The methodology caveat has to travel with those numbers, though, or the table misleads. Planning Survey 25 was the eleventh edition with 1,187 respondents, 19 products and 33 KPIs. Planning Survey 26 is the twelfth, with 804 responses of which 743 were analysed, 20 products and 26 KPIs. So the sample fell by roughly a third and the KPI set shrank in the same year the top-ranking count tripled. Peer groups are also rebuilt each edition based on specialisation, usage scenario and geographic presence, and a product needs at least 20 user reviews to be analysed at all. The counts are not comparable across years, and anyone presenting them as a trend line is overreading them.

Gartner placements are worth dating precisely, because Jedox has appeared in several different markets. It debuted in the Magic Quadrant for Strategic CPM Solutions in 2016, appeared in the cloud CPM quadrant in 2017 and the Cloud FP&A quadrants in 2018 and 2020. Leader status begins only with the Magic Quadrant for Financial Planning Software published in December 2023, repeated in November 2024 and December 2025, which makes it three consecutive years. Jedox was also included in the 2024 Magic Quadrant for Financial Close and Consolidation Solutions, but not as a Leader. That split matches what I see in practice: strong and improving in planning, earlier-stage in close and consolidation.

Other reviewers are more mixed, and that is useful. Nucleus Research placed Jedox among Core Providers in its 2025 CPM Value Matrix and moved it to Expert Vendors in 2026, which is movement in the right direction but not the Leaders group. Dresner named Jedox an Overall Leader for customer experience in 2021 and 2022, and has not featured it since. On Gartner Peer Insights, Jedox was a 2021 Customers' Choice at 4.6 stars from 63 reviews and now sits near 4.4 across roughly 222, which is the normal shape as a review base widens.

And the criticism is consistent enough to take seriously. BARC's aggregated review page for Jedox scores customer satisfaction 8.0 out of 10 and functionality 7.4, with reported problems led by slow performance at 35 percent and increasing resource requirements at 25 percent. The stated weaknesses are complexity when building intricate applications without prebuilt content, a history of departmental rather than enterprise-wide scenarios, and performance under high user counts and large data volumes. G2 and PeerSpot reviewers say the same two things: Excel familiarity and flexibility are the strengths, performance and complexity are the weaknesses.

I would not argue with any of that. Flexibility and performance are the same coin in a rule-driven OLAP model, which is most of why I write what I do about cube design: the platform will let you build something intricate, and it will not stop you building something slow.

Three numbering traps

Twenty years of renaming leaves potholes. These three cost real time.

There is no Jedox 8. The product went 7.0, 7.1, then 2018.1. Anyone searching for a Jedox 8, 9 or 10 will find nothing, because numbering moved to years.

AIssisted Planning versions look like platform versions and are not. It ships on its own train as 24.x, 25.x and 26.x. AIssisted Planning 25.1.0 is dated September 2024, roughly seven months before platform 2025.1 in April 2025. The Microsoft 365 add-in has its own numbering too, running from 2.2.0 in late 2022 to 3.6.0 in July 2026.

The version is no longer the whole answer. Since 2024 there is one major a year, and features arrive continuously in patches inside it. Both the 2025.1 and 2026.1 release notes are living documents with monthly dated sections. Two tenants on the same version can differ by months of functionality, and at least one 2026.1 breaking change is tied to a specific patch rather than to the April release. Record the patch level in your solution documentation.

Where that leaves us

Jedox today is a Kubernetes-based cloud platform, largely on Azure, licensed per named user on annual-billed subscriptions with tiered packages and connector bundles, serving a customer base Jedox puts above 2,900. The positioning language has moved from self-service BI in 2014, through enterprise planning and BI, to xP&A in 2021, to integrated planning built for Finance today.

What has not moved in twenty years is the actual idea: a multidimensional database you can write back into, driven from a spreadsheet interface, aimed at business users rather than IT specialists. Every major turn in this timeline is a change in packaging, delivery or front end around that core.

For anyone scoping work on it now, that history has three practical consequences. Flexibility is the product's oldest promise and its most common failure mode, so design for it deliberately. Upgrades belong to the vendor, so build your regression plan around a date you do not control. And the AI features are on their fourth iteration, which means they are worth testing properly but not worth building a business case on until you have seen them against your own model.

Back to all posts