Omni Analytics

Since before it became the 4th largest acquisition in Google’s history, Looker had a target on its back as every other data start-up promised to create yet another Looker. Many years later, there is still just one Looker!

Many years later, there is still just one Looker!

Looker BI (not to be confused with Looker Studio, which was renamed from Google Data Studio on the path of combining these two products into one analytics suite) is now fully integrated into Google. The exciting Santa Cruz surfer spirit that once dominated the “ethos” of both the Looker company and the Looker product is nowhere to be found.

The vast majority of its enterprise customers likely don’t event know what the product looked like pre-G. The distinctive brand colors have been swapped for the bland and dull colors of Google.

So it makes sense that some reasonable people might conclude that the product is ripe for disruption - even though dozens of other start-ups failed. Is now then the moment to take on Looker with a “better” product?

Not so fast…

Who is Omni?

Every decade, a new Billion-dollar data analytics company is born. In the early 2000s, Tableau was one. In the previous decade, it was Looker. Omni Analytics, a company founded by my former colleagues, would like you to believe that they are such a company of this decade.

Back in 2022, before they even had a product or much was known about what they were planning to do, I wrote this piece on Omni and their $27M Seed Round (I combined their Seed and Series A because both were raised pre-product during the same summer).

At the time, my suspicion was that their key innovation will be around AI-centric data model. If you read my post, I was skeptical they would be able to deliver on the “magical data model” because of just how difficult it actually is.

A lot has happened since that post - namely: Generative AI took over the world, and my own start-up is now building an AI-centric data model. But what of Omni? Have they built an AI-centric data model? The short answer is, NO. Instead, they have literally rebuilt Looker. Time for some deep analysis…

?? A note from the author

I pick on Omni NOT because they are bad. I pick on them because they are highly representative of the general problem in the industry - namely: assuming that a new BI tool can avoid the problems of an incumbent tool once it reaches maturity. Spoiler: it can’t.

A Personal Journey

Some context should be useful here.

Omni’s CEO, Colin Zima, and I started at Looker around the same time, back in 2014. It was the time when nearly everyone in the organization was using the product, and many of the decisions around what to build next came literally straight from our own usage. But by the time I left Looker, I was communicating to engineers dozens of customer-reported bugs and features on a daily basis.

It really is amazing how nuanced BI products are. Over the years, I must have reported at least 100 bugs around our handling of timezones; there was another 100 bugs around specific handling of each database; another 100 bugs around some incorrect handling of an integer value; you get the idea…

At first, I worked our engineering team to create solutions for each of the problems. Later I would focus on what was possible to achieve on the customer side - drafting a guide and the top-searched Google result for how to fix performance in AWS’s Redshift (a 3rd party database many of our customers relied on at the time). And in the process, I got to work with Data teams from such companies as Wise, Strava, Procore Technologies, and Docker, Inc. It was, without any reservations, an amazing experience!

However, much of that idealism was broken when I set out as an independent and started my data consulting practice. Over the next few years, I saw the reality of Business Intelligence.

Every company - regardless of the tool they use (be that Looker, Tableau, etc) goes roughly through the following journey:

1. A decision is made to procure a BI tool around a critical use case,
2. A new BI tool successfully delivers the critical use case,
3. 1-3 years go by,
4. A team, or the entire organization, makes a decision to procure a new BI tool

The experience left me disenchanted.

The work of data teams is complex and riddled with detail. So eventually - whether due to the pressure from inside the team or from that of outside business stakeholder - someone makes an enthusiastic bet that all these ills will be solved by just ?? swapping a BI tool. EVERY SINGLE TIME.

That is, of course, a major misperception. Swapping a BI tool is almost never the solution it is made out to be.

Looker 2.0 vs Looker

I am no Looker fanboy. When Google laid off the entire Customer Success team, I found this shocking. People running CS at Looker were the same people I started with back in 2014 - they were the very reason Looker succeeded in the first place. And then all this useless redesign… I totally get why the early adopter Looker fanbase found all this destruction tasteless and moved to criticize the product and the company.

On the other hand, the Omni fanbase would say here that Omni has fixed some of Looker’s problems and made the product better - namely the functionality to push new Explore-level table calculations into the model itself. Then there is that natural language interface to help me build Excel-like calculations. They have also avoided the mistake committed by the majority of data start-ups: “internal tools make lousy startups” (Benn Stancil). So I do think the product is not so bad.

Yet, as a Data practitioner, I don’t believe this alone is enough. BI is a complex domain - with a variety of integrations and highly nuanced enterprise edge scenarios. Google’s Looker might not be perfect (no BI product is), but from the pure product standpoint, this is still the most powerful solution out there. It is versatile for all sorts of enterprise and non-enterprise deployment scenarios: white-labeling, embed, APIs, on-prem, cloud-hosted, etc. Looker’s Data Modeling language and governance layer is second to none. And the team continues to move full steam ahead with product improvements and AI adoption.

Ultimately the decision here is about comparing the value of 10 new features (Omni) vs. foregoing of 10,000 features (Looker), and I am not sure this benefits the new “guy”.

It is like if someone set out to build a new competitor to?Tesla ???by focusing on making the car look prettier with a new set of software apps, but then does not provide the same network of charging stations. We might enjoy the moment of buying the car, but driving long distance will likely be drudgery.

And as I previously hinted in my last week’s post, there is an area where one could actually innovate in the BI: AI-centric data model.

The best way to understand an AI-centric data model is again with an analogy to Tesla. Instead of building a direct Tesla competitor - selling partially self-driving cars as luxury items - Cruise, another self-driving car manufacturer went after entirely different business model: self-driving taxis. Ultimately, both Tesla and Cruise compete for the same final market dominance - but the entry is different, and that makes all the difference.

This is exactly what I expected from Omni back in 2022, but this is not the path they have taken.

The Better Product

(Do you want to be young and stupid, or old and wise?)

Not sure about you, but personally I want to be young and wise - though I will compromise for something in the middle. But enough about me - what do people actually mean by a “better product” when they they decide to choose something like Omni over an existing incumbent product?

What I think they mean is that they make a philosophical argument about their values:

They value quick decisions and starting with clean white page, and don’t mind that it will cause them to struggle in the face of complexity.

Ultimately "new vs old" in the BI world is not a statement about the quality of the analytics product—especially when 2.0 version is so similar to 1.0—but it is a statement about one’s philosophical point of view.

Given that both Looker and Omni target a highly technical persona who is capable of writing SQL, a Software Engineer analogy is worth considering here.

The Young Engineer: - Quick to make decisions, - Seeks rapid outcomes, - Overlooks the intricacies and dependencies. The Old Engineer: - Deliberate, considering far-reaching implications, - Dependent on tried-and-tested methods, - Committed to quality from the outset. - Understands that simple beginnings != simple implementations

Choosing between the two isn't about absolutes but about assessing needs and priorities. If dealing with complexity and critical systems, or aiming for right-first-time success, the preference might lean towards the "old". If speed is of the essence, even at the risk of completing the project, "young" might be preferred.

Business Intelligence is a very old space. One could make an argument that first computer-based Analytics tooling predates Operating systems. Additionally, BI is a very competitive category.

This has implications. For example, Data Governance has been explored extensively, resulting in a comprehensive list of features mature tools should possess. Similarly, security measures have been repeatedly tested, leading to a mature tool encompassing numerous security features. New entrants like Omni may tackle the most pressing 10, 20, 100 issues but often overlook the broader spectrum of potential problems. This challenge isn't unique to Omni but is a common hurdle FOR ALL new BI tools.

So what now?

Last week I wrote that the current BI products are all struggling. The fight over AI is going to be the next fight in the Analytics space over the next decade - like one over Cloud Analytics and Cloud Databases in the previous decade.

The fight is far from predictable. Every tool under the sun understands the value, but few appear to have capability to deliver (see my previous post on this).

So if I would make one prediction, it would be this: focusing on one part of the technology stack will give that company a significant advantage. Going broad and wide (e.g. building both AI and a BI) will drive most companies nowhere.