How ISDIN unlocked self-service BI for 1,800 employees after modernizing their data stack

A culture built on one number

Founded in Barcelona in 1975 by two friends from local family-owned science and beauty businesses, ISDIN is today a Spanish skincare and dermatology company whose products reach consumers in more than forty countries. It is Spain’s leading sun care brand – a position built over nearly half a century of European photoprotection and dermatological research – and, in its own framing, “a global benchmark in dermatology.” Its mission – “to inspire everyone we touch to enjoy a healthy, happy and beautiful life” – is well known internally.

Less visible, but no less important, is the operating principle that has sat underneath the company for fifteen years: everyone, from a sales representative in Mexico to the chief executive in Barcelona, looks at the same numbers.

That principle, set down by ISDIN’s founder-led leadership long before “data democratization” became a slogan, has scaled with the business from six hundred employees to more than 1,800. It is operationalized through an internal program called iZoom: a set of workbooks – iZoom Sales, iZoom Expenses, iZoom Supply Chain, iZoom Community, iZoom Finance, iZoom Markets – all drawing from the same warehouse, all governed centrally, all taught as part of the company’s onboarding curriculum.

“Our official motto is to inspire everyone we touch to enjoy a healthy, happy, and beautiful life. Fifteen years ago, our CEO took the decision to make sure that everyone, from every business area, looks at the same data. That principle is still load-bearing today.” –  Ignasi Oliver, Data Manager, ISDIN

Beneath the principle sits an unusual asset for a cosmetics company. Every ISDIN product carries a QR code; scanning it in the ISDIN app or on the website earns the registered consumer points redeemable for products or experiences.

The result is a stream of customer-level, product-level behavioral data – purchase patterns, redemption choices, dormancy signals – flowing into the same warehouse that powers iZoom. ISDIN sells skin care; what it captures is consumer behavior at meaningful resolution, and that data feeds the loyalty and community strategies that drive a material share of revenue.

“The community strategy is one of the main motors of the business. People scan the QR code with the ISDIN app or via the website, they get points, and they redeem for experiences or products. We extract very insightful KPIs from it, and the community team works on them every day.” –  Ignasi Oliver

Running all of this is ISDIN’s Digital Skin function – the company’s internal name for its global technology organization, led by Rod Menchaca Betancur, Global Director of Digital Skin. Inside Digital Skin sits a dedicated data organization built around three specialist teams: Analytics, Data Engineering, and Data Science and AI, each drawing from the same BigQuery warehouse but solving different shapes of problem.

Day to day, those three teams are managed by Ignasi Oliver, ISDIN’s Data Manager. Analytics and Engineering power iZoom; Data Science and AI is the team most constrained by the old stack, and the one this story ends with.

Where the old stack broke

By 2023, the stack that had carried iZoom for a decade was beginning to give. QlikView and Qlik Sense held both the data and the visualization, on premise, in a single layer. Dashboards that had once felt fast were now taking ten seconds to load; trend views, fifteen. The data team’s credibility – the thing they had spent fifteen years building – started to fray at the seams.

The deeper problem was strategic. Data sat trapped inside Qlik, out of reach of the data-science and AI projects that needed cloud hyperscalers to run. The on-premise infrastructure had been optimized about as far as it could be.

ISDIN wanted to decouple storage from visualization – to put its data into a modern warehouse and pick its front end separately – so that neither choice would lock in the other, and so that the AI ambitions sitting behind the analytics work could finally be unblocked.

“If I’m analyzing the performance of the sales of the Mexico team, and I want to look at a specific product over the last five years, it took maybe ten clicks, and you had to wait fifteen seconds per click. Twelve minutes just for loading – people didn’t even bother.” –  Ignasi Oliver

The downstream effects were the kind that quietly hollow out an analytics function. Business users stopped attempting deeper analysis and instead routed their questions back to the data team as SQL requests.

The analysts, in turn, were pulled out of model-building and into a ticket queue. “We became task-oriented,” Mr. Oliver says, “instead of strategic.”

Propelling Tech, the implementation partner ISDIN worked with through the migration, puts the user-side cost more bluntly:

“Adoption had dropped. The Qlik apps had become so slow that the team had to fall back on daily scheduled snapshots just to give users something to look at.” –  Emili Bonilla, Partner at Propelling Tech

“We wanted to start using the data we had on Qlik for other projects – data science, AI projects that required the data to be in a hyperscaler like BigQuery or Snowflake. The fact that the data was, to put it some way, locked on Qlik was a pain that would not have allowed us, in the future, to be agnostic on the storage and engineering side of the warehouse from the visualization platform.” –  Ignasi Oliver

Two non-negotiables

When ISDIN began evaluating cloud-native BI in 2020, it did so with two requirements that were unusually firm. The first was warehouse agnosticism: whatever the company chose, it would have to work as well on Snowflake as on BigQuery, with no obligation to buy into a vendor’s full data ecosystem. 

The second was concurrency. At 08:00 each weekday, ISDIN’s sales managers across multiple markets open iZoom to check the previous day’s performance; the platform would have to absorb 150 or more simultaneous users without flinching.

Propelling Tech, who advised ISDIN through the evaluation, points to the buying culture that made the project work in the first place:

“ISDIN has always been willing to innovate and to be ahead of its competitors. With our advisory, one of their internal Qlik developers ran a quick test on Astrato and cleared the two limitations they cared most about – performance and self-service – in a very short time. They decided on facts, not on market perception.” –  Emili Bonilla, Partner at Propelling Tech

Three platforms made the shortlist: Tableau, Power BI and Astrato. Tableau and Power BI both performed adequately on raw speed. But both came with a commercial logic that worked against the architecture ISDIN was trying to build.

“With Tableau and Power BI, the message was: to extract the maximum benefit, it’s much more recommended to also be in their full data analytics environment. That was the opposite of what we had decided two years previously. The assurance that the Astrato team gave us – that if next year we have a project on Snowflake, we can still share the same workbooks and the same models we built in Astrato – that gave us much more confidence that this was the way to scale.” –  Ignasi Oliver

Independence, in the end, was the decisive factor. Astrato’s warehouse-native architecture – querying data where it lives rather than pulling it into a separate engine – meant ISDIN could keep its data layer in BigQuery, build its semantic logic in dbt, and treat its visualization tool as a window onto the warehouse rather than a parallel copy of it. The logic held all the way up the function: Mr. Menchaca Betancur frames the choice in starker, executive terms.

“We moved to BigQuery to modernize our data infrastructure. We needed a powerful analytics layer that would be business-friendly but also future-proof. Astrato gave us everything we wanted.” –  Rod Menchaca Betancur, Global Director of Digital Skin, ISDIN

Emili Bonilla, who has led similar engagements across most of the major BI platforms, is more catholic about where each tool fits – and what made ISDIN’s match a clear one:

“Power BI can be the right answer when the use case is simple dashboards on modest data volumes. But where customers need to do more – and they often need to do more than they realize – Astrato’s combination of cloud compute, security, advanced simulation, LLM integration and speed of development means it sits as either a replacement for, or a complement to, the traditional BI platforms.” –  Emili Bonilla, Partner at Propelling Tech

Migrating without breaking the muscle memory

The migration of the data layer had started before the visualization decision was made – ISDIN had been shifting business logic out of Qlik and into BigQuery for several years.

Once the warehouse was ready, the visualization migration was sequenced into a six-month plan: a silver-layer rebuild that was fully agnostic to whatever sat on top, a soft-landing phase with a selected group of users, and then global rollout.

“It was critical to create a reliable silver layer that was fully agnostic to the visualization layer. We used an accelerator that migrated the Qlik code and optimized it under a blueprint ISDIN defined – naming conventions, development guidelines. Once the silver layer was ready, Astrato’s low-code approach let us build the initial app version in less than two weeks.” –  Emili Bonilla, Partner at Propelling Tech

ISDIN chose continuity over reinvention. Rather than redesign its dashboards from scratch, the team lifted the familiar Qlik structure across to Astrato – a general view with tables, a trends tab with time-series graphs, a movers tab with composition charts. Users opened the new tool and saw something they already knew how to use.

“Most of the iZooms have the same basic structure – a general view with tables, a trends tab where you can interact with the data graphically, and a movers tab. The biggest improvement is that we have decreased the usage of Excel. Before, people downloaded data and used VLOOKUPs between files to extract insights. Now they can do all of this inside the BI platform – the intended use, working fantastic.” –  Ignasi Oliver

The self-service layer was a genuinely new thing. Through Astrato’s personalization features, users can now reorder columns, save bookmarks, and convert a table view into a bar or scatter chart without an analyst in the loop. VLOOKUP in Excel – for years the unofficial second BI tool at ISDIN – quietly faded out of daily use.

Propelling Tech’s team made one deliberate design call worth noting: they migrated only the key features rather than try to port everything. The result was applications that were more performant and easier to use than their Qlik predecessors. The exception was a small set of very advanced apps with expressions running over a hundred lines of code, which had to be redesigned rather than migrated as-is – a slower path, but one that left the resulting estate cleaner to maintain.

A near-invisible transition

Adoption was, by Mr. Oliver’s own description, almost a non-event – and the reasons travel beyond the specifics of a Qlik shop. 

ISDIN made two design choices that lowered the cost of the migration regardless of where any individual user was coming from.

They preserved the familiar dashboard architecture so business users opened the new tool and saw something they already knew how to read. And they leaned on Astrato’s in-product AI assistant and documentation to absorb the long tail of “how do I…” questions that would otherwise have queued up in an analyst’s inbox.

Analyst onboarding worked across paradigms. Analysts with Qlik backgrounds found Astrato’s associative filtering and dynamic interactions immediately familiar and were productive within days.

Analysts arriving from more modern cloud-native tools – Looker, for instance – needed longer to recalibrate, but Mr. Oliver characterizes the learning curve as “not steep” rather than disruptive. The most notable user-side friction was the need to click “apply” after selecting filters in a table – small enough, in Mr. Oliver’s phrasing, to file under anecdote rather than concern.

“There are a lot of similarities in user experience, so people were very, very used to it. The main difference they noticed is, ‘oh, sometimes I forget I have to click apply when I select directly in the tables.’ That is more anecdote than concern. The chatbot helper available in Astrato has also been useful – it’s good to have the documentation available through a chatbot, not only direct contact with the Astrato team.” –  Ignasi Oliver

What changed

The clearest gain is the one that does not appear in a dashboard. ISDIN’s analytics team has stepped out of the ticket queue. Freed from running ad-hoc SQL for the business, analysts are back to building – expanding metrics, refining dimensions, layering on dbt models – while output has risen without a single new hire.

“The biggest internal benefit Astrato has implied for us is that we are now focused on increasing the metrics and dimensions we make available, so business teams can be self-sufficient and use self-service. We have been able to increase the workload in terms of evolution of the data models without actually increasing headcount or internal resources.” –  Ignasi Oliver

Performance, the original trigger, is in another league. Dashboards that took ten or fifteen seconds now answer in a beat. The credibility the data team feared it was losing has, in Mr. Oliver’s words, been recovered.

“Our analytics team works now in a very orderly, very focused way at each stage of development – data layers, SQL, dbt models. They don’t have to keep in mind all the business logic on top of Qlik anymore. They prepare the data; Astrato performs the relation we want with it.” –  Ignasi Oliver

From outside the data team, Emili Bonilla points to the user-side recovery as the more visible shift – and as the moment ISDIN materially expanded its commitment to the platform:

“Adoption has increased drastically. Users came back, usage has kept rising, and ISDIN moved to an unlimited-users subscription – so the whole organization now has access.” –  Emili Bonilla, Partner at Propelling Tech

There is a quieter discipline ISDIN has built on top of the new stack. Each iZoom now carries a “last updated” marker, per data source, in plain sight. A salesperson who cannot find a new account in iZoom can see that the CRM feed was refreshed three hours ago and the account was added two – so the answer is wait, not investigate. Real-time, in Mr. Oliver’s framing, is not a marketing word but an operating contract between the warehouse and the visualization layer.

“We are very much in a D-minus-one world. Users know exactly when each source was last updated, so they can answer their own freshness questions. If we did not have a visualization platform that could pull from the warehouse in real time, this would be an issue. Previously, it was. Now it is not.” –  Ignasi Oliver

What’s next: from describing the past to predicting the future

ISDIN is already turning iZoom from a system that describes the past into one that anticipates the future. 

Mr. Oliver’s Data Science and AI team – the one most constrained by the old stack, and the most freed by the new one – has built XGBoost and deep-learning models that produce predictive metrics: revenue forecasts by affiliate, churn-rate predictions for the QR-code community, projected sales by business unit. 

Until now, those models lived in the data-science layer. The 2026 initiative is to surface their outputs as first-class columns inside the iZoom workbooks themselves – dashboards, automated reports, and the self-service dictionary that lets business users build their own views on top.

“What we have done for all these years is show what has happened. The big change for this year is incorporating predictive KPIs into the same iZooms – what we have today, plus what our best model predicts will happen, considering all the information we have up to yesterday. Now that we have the visualization platform and the data warehouse ready, this is the moment.” –  Ignasi Oliver

Sitting behind that is a more radical thesis: that AI will reshape what a BI tool is supposed to be. Mr. Oliver argues, with some force, that business users are arriving at iZoom with a new expectation – one built up over two years of ChatGPT and similar tools. They no longer want to learn a dashboard. They want to ask a question.

“The main shift will come from the assumption that, in order to get answers to a business question, I have to look into a dashboard. People have been using AI solutions in text form for a few years now – chatbots, ChatGPT, Claude. They are starting to expect to get business answers the same way.” –  Ignasi Oliver

For an AI layer inside BI to actually work, Mr. Oliver argues, it has to do three things at once. The first is the obvious one: be accurate. The second is to understand context – who is asking, with what authority, and what they actually need.

“It is not the same if I, as a data manager, ask ‘how have the sales in Mexico been’ as if the country manager of Mexico asks the exact same question. The expected answer is quite different. The country manager expects a summarized overview – not the detail of every product.” –  Ignasi Oliver

The third is the hardest: intent. The same question can be the start of a routine review or a coded check on a supply problem nobody has named out loud yet. Mr. Oliver thinks BI tools will, in the near future, be expected to read which it is.

“If the country manager is asking how the sales are because of a regular review, that is one thing. If the real question is ‘there has been a disruption in shipment, has it shown up in the numbers yet?’ – that is another. These expectations are, a lot of the time, only in the mind of the person asking. In the very near future, that is what the BI layer will be expected to capture.” –  Ignasi Oliver

Combine all of that – predictive metrics surfaced inside the dashboards, an AI layer that understands persona and intent, and a warehouse-native architecture that doesn’t care which model, or which hyperscaler, is sitting behind it – and the result is the BI tool ISDIN is quietly building toward. 

Mr. Oliver, asked to summarize it, was characteristically restrained: “Once we are able to mix this with the AI layer, I think it will make a big change for the better.”

‍

“Cloud-native, adaptable, agile, real-time – and, most importantly, agnostic. That is what a modern BI layer should be – and what Astrato has proven to be.” –  Ignasi Oliver, Data Manager, ISDIN

‍

‍