Nov 13, 2019 · 20m · mad
How to Answer Data Questions Without Being Miserable // Ahmed Elsamadisi, Narrator (Data Driven NYC)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this presentation at Data Driven NYC, Ahmed Elsamadisi, Founder and CEO of Narrator.ai, addresses the frustration surrounding modern data workflows and presents a simplified architecture centered around a single Activity Stream model (Customer, Activity, Time) that streamlines queries, eliminates stack complexity, and restores business impact.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 4% of the talking time here. How this is scored →
Matt as informed peer 1.0 Guest teaching 0.6 Guest disagreement 2.4 Matt pushing back 0.8
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 3:30 Looker and BI Tool Rant
Ahmed forcefully rejects the utility of BI tools, calling them the biggest myth in data and claiming Looker intentionally makes debugging impossible.
Hardest push from Matt ▶ 13:24 Playful Sarcastic Challenge on BI ToolsMatt playfully challenges Ahmed's intense anti-BI rant with deadpan sarcasm, asking him to clarify whether he actually likes BI tools or not.
Biggest teaching moment ▶ 14:43 Reframing Transformation in the Modern Data StackAhmed educates the audience and host on how activity streams redefine transformation from a heavy, predictive ETL phase into a lightweight, on-demand query assembly model.
Matt holds his own ▶ 14:22 Framing the Evolution of ETLMatt displays clear technical fluency in data engineering by asking Ahmed to contextualize his approach within the historical shift from ETL to ELT.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| The Difficulty of Answering Business Data Questions | 0 | 0 | 4 | 0 | In this solo presentation segment, the host does not participate. Ahmed aggressively attacks conventional data stack tools, calling BI tools the 'biggest myth' and mocking Looker debugging. | |
| The Failure of Dashboard-Centric Data Architectures | 0 | 0 | 3 | 0 | Solo presentation continues without host participation. Ahmed dismisses traditional multi-table dashboard architectures and acknowledges that people call 'bullshit' on his single time-series table concept. | |
| Introducing the Activity Stream and Single Time-Series Model | 0 | 0 | 1 | 0 | A purely technical walkthrough of activity streams and SQL queries in a presentation format with zero host interaction. | |
| Reusability, Standardization, and Rediscovering Data Joy | 0 | 0 | 1 | 0 | Ahmed concludes his presentation monologue explaining how standardization restores joy to data analysis. Host remains silent during the pitch. | |
| Audience Q&A: BI Tools, Data Warehousing, and Narratives | 5 | 3 | 3 | 4 | Matt enters to moderate Q&A, teasing Ahmed about BI tools and demonstrating domain knowledge by prompting a discussion on the evolution from ETL to ELT. |
Statements from this episode (17)
Elsamadisi: Answering data questions provides a company's biggest competitive advantage
“The thing about data questions is that no matter how painful they are, it is a company's competitive advantage, and that's what makes the biggest impact on whether a company can succeed and outdo its competition.”
Elsamadisi: Cross-system data queries take weeks due to unlinked identifiers
“There's no foreign key that ties these two systems together. There's no way to deal with different user identifiers. You would have to just go to your engineering team, And ask them to copy that data in all these different places, and then you deal with other …”
Elsamadisi: Data tools tend to add more complexity than they solve
“Tools tend to add a lot more complexity than they actually end up solving.”
Elsamadisi: Buying Looker requires hiring five to six BI engineers
“All it ends up doing is you end up having to hire a team of five or six BI engineers to write LookML and maintain these systems for you.”
Elsamadisi: WeWork's 45-person data team spent over $1M with little impact
“We actually spent over a million dollars, a lot of time and resources. We had a 45 person data team, and then very little impact.”
Elsamadisi: Traditional data software builds dashboards, not answers
“Well, it turns out that all these tools and all these systems are designed to help you build dashboards, not really answer questions.”
Elsamadisi: WeWork maintained 3,000 dashboards without resolving its data issues
“We had 3000 dashboards, and it was not helping.”
Elsamadisi: WeWork's data team managed 3,000 tables and 700 transformations
“We had 3000 tables, 700 transformation tables”
Elsamadisi: Any business data structure simplifies to customer, activity, and time
“Your business can be broken into three simple things. Customer doing some activity in time.”
Elsamadisi: Companies can replace hundreds of database tables with one time-series table
“So, instead of hundreds of fact and dimension tables, you can just have one time series table centered around a customer.”
Elsamadisi: A working single time-series table eliminates need for 20-30 data engineers
“If a single time series table worked, you wouldn't have 20 or 30 data engineers building tables to answer questions.”
Elsamadisi: Time and customer identity replace foreign keys when relating multi-system data
“Instead of depending on foreign keys to relate the data, you can actually just use time and customer.”
Elsamadisi: Activity stream queries reuse consistent structures to bridge disparate systems
“The really fascinating thing here is that the query is so consistent. I can literally reuse the same query and combine layers of questions, all using the same structure, all being answered by the same table, bridging these systems invisibly.”
Elsamadisi supported 16 companies across five industries using one time-series table
“This allowed me, as a single person, to support 16 different companies using five different industries, all using the same time series table.”
Elsamadisi: Single-table activity streams guarantee all dashboard numbers will always match
“There's only one table. Everything will always match.”
Elsamadisi: Narrator reduced 40-hour Looker queries to just 10 minutes
“We've taken queries that were taking, like, literally 40 hours in Looker, and we've dropped them down to, like, 10 minutes. Queries that cross billions of rows end up happening in, like, seven to eight seconds on a warehouse.”
Elsamadisi: Redshift, Snowflake, and BigQuery perform identically on long tables
“The speed improvement on Redshift and versus Snowflake versus BigQuery becomes interchangeable. Because what they handle is for those edge cases, we have like a bunch of, like Snowflake is really good at a bunch of small tables everywhere trying to load a memo…”