Unlocking the Art of the Possible from Data Integration

1. Why Data Is Still Hard: Nick Jordan has spent his career in data-related product roles, and he describes the same pattern every time: the job looks easy until day two. As data becomes more important, there is more of it, and as there is more of it, complexity grows. Narrative.io was founded on a simple thesis: make data easy. That starts with accepting that data arrives in an infinite number of forms, and that anyone using it, whether in a spreadsheet or a large language model, must cope with that variety.

2. Normalization in Practice: Normalization projects different representations of the same thing into a single way of thinking about it. The simplest case is male versus M. Currency is another: a price in dollars means something different once translated to euros or yen. Nick's favorite example comes from veterinary clinics, where a receptionist might type golden retriever, golden, retriever, or olden retriever. If only 80% are spelled correctly, a basic trend question becomes a list of 75 questions. Narrative's Rosetta Stone technology categorizes data so those questions collapse back into one.

3. Where Large Language Models Fit: Ten years ago, before the LLM revolution, Narrative built its own neural networks to learn the thousands of ways a dog breed can be misspelled. Today it layers in more powerful AI, but the core asset is the platform primitives that record each mapping so it holds forevermore. Nick is clear about LLM limits: they are non-deterministic, expensive when a data set has billions of rows, and not observable when two users get different answers to the same task. He expects more layers to sit on top of AI to make it capable of powerful work, often outside a chat interface.

4. Integration at the Speed of the Market: Integrated CRM and data lakes have long been the holy grail of marketing, often pursued through large outsourcing projects. Nick, with 20-plus years in martech and adtech, points to the Lumascape as proof of how complex the ecosystem has become. A brand may have an agency of record, in-house data, and 20 to 40 platforms. Multi-year federation projects fall behind because the ecosystem keeps changing; the spec is out of date before work starts. If a team wants to test advertising on OpenAI before competitors, a nine-month integration timeline defeats the purpose.

5. Block and Narrative Anywhere: Block announced a commerce media network built on financial transaction data from Afterpay, Square, and Cash App. Because many of those businesses came through acquisitions, the data was not harmonious, so the first step was normalizing it into one large data set and plugging it into ecosystem platforms. Block uses Narrative's off-the-shelf technology. For sensitive data, Narrative Anywhere moves the technology into the customer's own warehouse or cloud, so Narrative never touches or sees the data. As Nick puts it, the less you move data around, the faster things work.

6. Federated Queries and the Limits of Dashboards: Data does not land in a new clean receptacle. First-party and partner data stay where they live, in their own formats. Technical users write queries in NQL, and less technical users work through interfaces. The query is federated out, normalization happens on the fly, and the result returns as a new data set or an answer. Nick argues dashboards force you to presuppose every question. When he worked at a music store next to a movie theater, the Titanic release drove sales, something no one would have built into a dashboard ahead of time.

7. Agentic Workflows and Faster Hypotheses: Narrative is leaning into agentic workflows. A user states a question, an agent hypothesizes ways to answer it, runs three or four of them against already normalized data, and suggests follow-up questions or tactics. Nick contrasts this with his time as Chief Product Officer at Tapad, where moving from a data scientist's hypothesis to finding, negotiating, and integrating the data took quarters, so every bet had to be safe. Compressing that cycle to hours or minutes lets someone with a light bulb moment in a meeting simply go and test it.

8. Rethinking the Data Warehouse: Nick does not expect data warehouses and data lakes to disappear. Snowflake, Databricks, or AWS still serve as the system of record. What changes is the design thesis. He compares exhaustive, thousand-page data specs to how NASA gets someone to the moon: very safe, but slow, with technology adoption measured in decades. A startup asks what it is working on for the next two weeks. With normalization on top of a system of record, business users can change their minds and get new analytics without waiting on one all-knowing database administrator. Dave and Nick agree the word for this is democratization.

9. Ecosystems, Culture, and Infinite Work: Dave sees new competition built on ecosystems, such as Home Depot coordinating a bathroom renovation across manufacturers, contractors, day laborers, permitting, and adjacent players like Grainger. Such partnerships have historically died on the vine. Nick says the first barrier is cultural: asking whether you are happy with how things are done today. He also notes Narrative is not always the answer, recalling a bank whose asset management problem was better suited to an asset management vendor. Nick describes himself as 80% bullish and 20% scared of AI, and believes it will leave an infinite list of more interesting, career-rewarding work.