Every Business Has Data.
The Question Is Whether It's Working For You.
You've probably heard that you need a "data warehouse", a "lakehouse" or a "modern data stack". But what do these terms actually mean, and why should a company your size care? This guide explains the idea in plain terms, along with what it takes to build a data foundation your business can actually rely on.
Strip away the buzzwords and what you're left with is an engineering pattern that's been around since the early 1990s. At its core, it's about consolidating your different data sources: your ERP (SAP, Microsoft Dynamics, BMD), your CRM (Salesforce, HubSpot), accounting and bank exports, your webshop, and the Excel files that quietly hold half the company together. Data from the most important of these systems is pulled into a set of so-called Data Marts that complement each other and provide a unified, analytical view of your business: a single source of truth you can actually rely on. A historian reconstructing a single sale faces the same task. A merchant's ledger, a letter, a customs register and a city chronicle all mention it, but they name the merchant Bianchi, Weiß, Bianco and Weiss. Your company keeps the same kind of archive: the ERP is the ledger, the CRM and its e-mails are the correspondence, the bank's payment records are the register, and the Excel report someone built for the last board meeting is the chronicle. Before anyone can say what happened, the sources have to be reconciled.
A Data Mart is simply a clean, detailed collection of tables describing one aspect of your business, without that view being limited to a single source system. Think of a data warehouse as a modular system made up of these Marts, each of which can evolve as your business does. The Marts contain the business logic that people across your company already reference when they build reports, run ad-hoc analyses, or try to answer a business question, but with the advantage of being auditable, governable and clean. Reconciling also means knowing each source's conventions: to date that sale, a historian has to know that the Venetian year began on 1 March, while the letter follows an older calendar ten days behind. A Data Mart records the same kind of decisions: which convention each source follows and how the sources were aligned.
Because each Mart represents the single authoritative source for one area of your business, reporting becomes standardised inside departments and inconsistencies in reporting explainable when identified in a meeting. Business hypotheses can be tested against real data. AI and BI strategies become something you can actually pursue rather than just plan for. The barrier for new employees, in finance, operations, or anywhere else also drops significantly. Instead of requesting access to five different systems and chasing three colleagues for context, they can reference a unified definition of your KPIs and business transactions and start delivering real answers to real problems from day one. When two sources disagree, a historian does not silently pick one. The footnote explains the difference, and a source written to impress is taken with a grain of salt.
There's a technical upside too. By separating analytical workloads, which often require significant computational resources, from your operational databases, you ensure that reporting and exports never interfere with the systems running your day-to-day operations. Historians work the same way: they study transcriptions in the reading room, while the originals stay untouched in the archive.
The result is what historians call a critical edition: one record, with every source and every decision in the footnotes. A set of Data Marts gives your business the same. Together with the warehouse or lakehouse they live in, they form your data foundation: the layer every report, every analysis and every AI use case in your company builds on. The rest of this guide looks at where those Marts live, whether in a warehouse, a lake or a lakehouse, why this kind of foundation matters especially for SMEs, who needs to be in the room when Marts are designed, and what a sensible first step looks like.
adì 12 febraro 1594 m.v.
Corrado Bianchi da Norimberga
panni di lana b. 400 · L. 1.240
den 2. Februar 1595, a. St.
für Herrn Conrad Weiß
416 Ellen Wolltuch gekauft
13.II.1595
Corado Bianco, Nurimberga
1 balla, br. 400
Im Frühjahr 1595 brachte
der Weiss aus Venedig
über 1000 Ellen Tuch
Analysis happens on copies. Nobody annotates the originals, and nobody queues behind the researcher to use them.
1 Venetian year began 1 March (more veneto); letter dated old style.
2 Recorded as Bianchi, Weiß, Bianco, Weiss.
3 400 Venetian braccia ≈ 416 Nuremberg ells. Chronicle figure not adopted.
Warehouse, Lake, Lakehouse: What Is the Difference?
Data Marts need a home. Over the last thirty years, three answers have emerged to the question of where analytical data should live, and the names are used loosely enough that they are worth untangling.
A data warehouse is a database built for analysis rather than for running daily operations. Data arrives structured and is shaped into clean tables with fixed definitions before anyone reads it. It is the archive’s reading room: orderly, catalogued, reliable. Its strength is consistency; its limit is that everything has to fit into a table first.
The data lake came later, as companies started collecting what does not fit neatly into tables: PDFs and contracts, e-mails, machine and sensor logs, images, raw exports. A lake stores all of it cheaply, in its original form. It is the archive’s storeroom: nothing gets thrown away, but without a catalogue, finding anything, let alone trusting it, becomes hard. Many early lakes turned into what the industry calls data swamps.
A lakehouse combines the two. Data sits in inexpensive storage in open file formats such as Apache Iceberg or Delta Lake, and a layer on top adds what made warehouses reliable: tables with a defined structure, change history, access rules and fast queries. Structured ERP data and unstructured documents live side by side, and the same foundation serves reporting, analysis and AI.
For an SME, the label matters less than the principle. Warehouse or lakehouse, what you are building is your data foundation. Whether your Marts end up in a classic warehouse or a lakehouse depends on your data, your existing systems and where you are headed; for new builds, the lakehouse has become the default because it keeps options open without locking you into a single vendor. In both cases, what matters is the same: originals stay untouched, copies are reconciled, and every figure can be traced back to its source.
Why SMEs Need a Data Foundation
Large enterprises have had data teams and warehouses for decades. SMEs have mostly made do, and for a long time that worked: the business was small enough that a few people could hold the numbers in their heads.
Many still run primarily on spreadsheets and manual exports. Most others sit somewhere in the middle: a central database exists, but Finance defines revenue differently than Sales, nobody can agree on which dashboard to trust come Monday morning, and nobody can explain why the numbers do not align, even when the reason is perfectly valid from a business perspective.
That gap gets more expensive as a company grows. Every new system, site or product line adds another source, and the time spent reconciling numbers by hand grows with it. Knowledge that lives in a single spreadsheet in Controlling leaves the company with the person who built it. And when a bank, an investor or a potential buyer asks for figures, putting them together becomes a project of its own.
Your data foundation closes that gap without requiring a large in-house team. It is less a technology project than a discipline, the one historians call source criticism: understanding who produced a record, under which conventions, and what has changed since, before treating it as evidence. Each of the four questions below has a direct counterpart in your data.
adì 12 febraro 1594 m.v. · Corrado Bianchi da Norimberga
panni di lana b. 380 400 · L. 1.240
“b.” is a braccio, but which one?
In Venice a wool braccio is ≈ 68.3 cm; in Florence ≈ 58.4 cm. Same word, 17 % difference.
“Revenue” means something else in Finance and Sales
Invoiced and net of returns, or booked at order? Both are valid; the definition has to travel with the number.
380 struck out, 400 added in a later hand
Which figure did the merchant’s partners see at the year-end balance?
An order edited after last month’s export
Without change tracking, two correct reports show different totals, and nobody can say why.
Letter and Fondaco register agree on the buyer
Three independent records name the same merchant. The chronicle agrees on the name, not the amount.
ERP total vs. payments received
Validation compares independent systems before a figure is trusted in a board meeting.
“1594 m.v.” means February 1595
Rule: for dates more veneto, add one year to January and February.
A fiscal year starting 1 April
Business logic: which months belong to FY 2025, which exchange rate applies, which orders count.
Marts Are Designed With the People Who Know the Business
A Data Mart is only as good as the business logic inside it, and that logic rarely lives in the source systems. It lives in people: in Controlling, where someone knows that last year’s credit notes get booked in January; in Sales, where someone knows which three customer accounts are really one group; in Logistics, where someone knows that “shipped” in the ERP does not mean the truck has left.
A historian working on a sixteenth-century ledger does not guess what “b.” means either. They consult the specialist in Venetian measures, the palaeographer, the archivist who knows the collection. The critical edition carries their judgement in its footnotes.
That is why, at MoLo, we involve your domain experts directly in the design of every Mart, not only when collecting requirements and at the final sign-off. In practice, that means short working sessions with the people who own a process: we walk through real records from your systems together, agree on definitions in their language, and write down the rules and exceptions they apply every day. They check the first numbers against what they know to be true, and they own the definitions afterwards.
The effect goes beyond correct figures. When Finance and Sales have defined revenue together, a difference in the board meeting becomes a footnote rather than a debate. Knowledge that used to sit in one person’s head becomes documented and shared. And people trust, and actually use, reports whose logic they helped shape. In the end, that is what decides whether a data foundation pays off.
From Foundation to AI, One Mart at a Time
Almost every SME conversation about AI eventually reaches the same point: the model is ready, the data is not. An assistant that answers questions about your margins, a forecast of next quarter’s demand, a score that flags customers about to leave: all of them need a data foundation underneath: clean, connected, well-defined data. Without it, AI repeats the inconsistencies of your spreadsheets, only faster and more confidently.
The good news is that building that foundation no longer has to be a multi-year programme. Over the past decade, developer communities have optimised and open-sourced much of the tooling, from connectors that pull data out of common ERPs and CRMs to frameworks that version and test transformation logic. That makes a step-by-step approach realistic for SMEs.
A sensible start is small: pick one business question that matters, connect the two or three sources it depends on, and build the first Mart together with the people who own that question. Once it is trusted and used, the next Mart builds on it. Each step delivers value on its own, and the foundation grows with your business instead of being designed for a future that may never arrive.
The last illustration in this guide shows what that unification is worth. Once the sources are reconciled, the edition’s glossary becomes the legend of a map: the braccio becomes a length you can measure, four spellings become one merchant house in Nuremberg, and a single sale becomes a route over the Brenner that anyone can follow. Your data foundation does the same for your business. When revenue, customers, fiscal years and processes each mean one thing, the numbers from Finance, Sales and Operations finally fit on the same map, and questions that cut across departments, such as which customers, products or channels actually make money, can be answered in minutes instead of after a week of exports.
The technology for all of this exists and is more accessible than ever. What varies from company to company is the business logic: which sources matter, which conventions they follow, and which definitions your people actually work with. That cannot be bought off the shelf. It has to be worked out together with you.
That is where we come in.
Our goal is to help European SMEs turn years of collected data into real data assets that AI and BI strategies can actually be built and executed on. We believe that everything starts from the data and that clean, maintainable data infrastructure is not a trend or a luxury. It is a differentiator.
If this resonated with you, or you would like to talk through what a first Mart could look like in your company, we would love to hear from you. Reach out and let us talk data.
Book a 30-minute call →A reconciled record tells you what happened. What it means for your business, and where that picture is still incomplete, is the next layer: shared definitions and context that sit on top of your data foundation.
A reliable data foundation is where it starts. Meridian is the decision layer that turns that foundation into contextual answers, dashboards, Pulses, and an information base that can grow with your company.
See how Meridian turns a data foundation into continuous decision intelligence.- 01Braccio (Venice, wool)
≈ 68.3 cm, not the Florentine braccio. Your “revenue”.
- 02Year (more veneto)
Starts on 1 March. Your fiscal year.
- 03Merchant house
One entity, many spellings. Your customer.
- 04Route
Venice → Brenner → Nuremberg. Your process.
- 1 braccio ≈ 68.3 cm
- year starts 1 March
- 1 merchant = 4 spellings
- route of 12 Feb 1595