ThoughtSpot CEO Sudheesh Nair is in an ebullient mood as he pulls up a chair at the vendor’s London offices. Originally scheduled to be in Europe for an industry event – one abandoned by several companies, including his – Nair took the opportunity to swing by ThoughtSpot’s growing presence in the City instead.
Last time we spoke, the $4.2 billion Mountain View start-up had just bought business intelligence player Mode Analytics, in a deal valued at $200 million. The aim was to enable a code-first approach to self-service business intelligence, the results of which could be queried by both data scientists and non-experts using ThoughtSpot’s natural-language AI.
In June, Nair told me the deal was step one on the road to his company being “massive”. Nearly six months on, is ThoughtSpot any closer to that aim – with its long-rumoured IPO still on the horizon? The Mode deal seemed to come out of the blue, suggesting it may have been more opportunistic than strategic…
It was not all of a sudden. We worked on it for five or six months, starting from late last year. So, it was always part of the plan.
The thought process around that acquisition was to create what we call multimodal analytics. In the world of data, all the silos are being removed. If you go to Snowflake, Databricks, or Google BigQuery, they are breaking down all the silos, whether it's Salesforce data, or SAP, or Oracle, all of those are getting consolidated. But the silos for analytics still remain.
So, our process is to say, instead of forcing people to stop using Excel, or Tableau, or stop using predictive – which is not the right thing to do, because people have developed their own styles – is it possible for us to bust the silos, but still deliver different styles of analytics?
What does he mean by that?
Some users like to watch dashboards, but someone else likes to use SQL to create their own reports, because they are advanced explorers. But someone else again wants to continue using spreadsheets. So, can we all do that on a single platform?
Consolidating data, providing flexibility in how you want to do analytics, while still having a centralized platform that delivers multimodal experiences: that’s what we are trying to solve.
Without this platform approach, analytics is “painful” for many people, he says. But of course, AI is now able to surface insights from enterprise data, whichever container it sits in, which is ThoughtSpot’s core purpose.
Be a responsible citizen in the stack that customers want to build. Focus on delivering flexibility, but still have single governance control – which is more efficient and lower cost. That’s where Mode fits in, and this is what we are executing towards.
So, how is the journey towards being “massive” going? And how is the acquisition bedding in? Nair is surprisingly candid about his progress to date:
There are some areas that are really great, and others that are much harder than I thought.
On the upside, there is not a single Mode customer that I've talked who is not loving the product. […] But the challenge has been on two sides. First, I probably underestimated the tool complexity in the back of two companies coming together.
It has been painful – in the sense that, in the case of Salesforce integration, they do marketing automation through Pardot, and we were on Marketo. So, when you pull that out and you're trying to make a single entity, everything breaks.
And the downstream effect on compensation plans, account mapping, customer plans, customer advocacy plans, all of those things get completely broken. So, I thought it'd take two months, but it has taken nearly four. But we are almost there – by the end of November, hopefully, we’ll get there.
That's the price to pay for the entire business: it grinds momentum down.
But that has not been the only challenge, he says:
The second is culture. But this one I did not underestimate: I knew it was going to be a huge issue, because of two very different companies coming together. One was built in San Francisco, with cloud-native companies as its primary thing [sic]. Whereas ThoughtSpot was built for large enterprise customers, with complexities of scale and governance.
Large customers put an extremely high premium on governance, security, and control, and ThoughtSpot was built for that primarily. Speed, flexibility, quality, reliability, agility… all those things are coming together. But both people and mindset now need to come together too. The thesis is right, and the product is good. So, we are now working through the people integration, plus the systems and tools.
Then he adds:
Acquisitions are not just about technology; they are also about finding the right talent. So, if we find phenomenal talent, a group of people who are building great product, we will absolutely be keen to do more.
Of course, the problems that Nair describes have been the story of countless tech M&As over the years. While it might be relatively straightforward to buy customers in this way, grow revenues, or acquire IP – with the Mode deal, ThoughtSpot's ARR should increase by 50% to $150 million, while its customer base doubles – integrating systems and cultures has troubled even the most experienced players.
While often seen as a fast track to growth, M&As can be a slow road to hidden costs. Historically, some have never overcome those difficulties, in fact. So, doing so before Christmas would be a notable achievement – if Nair pulls it off as quickly as he is predicting.
From what happened, to what will happen
Prior to the Mode deal, some analysts seemed hopeful of a ThoughtSpot IPO before the end of this year. But from Nair’s description of merging the disparate teams, that was never realistic; 2024 seems a more likely bet, with the added bonus of bringing those teams together.
But IPO aside, what is next on the CEO’s roadmap? His answer is intriguing:
Analytics has always been thought of as a glorious window into the past. If I'm a supply chain specialist or a marketing specialist and I want to know why customers are leaving, I would go to ThoughtSpot BI to tell me. But those dashboards and reports are all about the past.
I want to take analytics from not only talking about what happened to why it happened. Then to what will happen. There is a huge opportunity for a company like us to connect those dots.
Inevitably, this is where AI comes into the predictive analytics mix. Nair explains:
The way we are doing that is by using AI in a responsible and ethical way. We use generative AI for intent capture in a relevant natural-language question.
If I'm a user, I'm maybe asking questions like, ‘How many customers are coming to my coffee shop?’ Or ‘Who is returning from last month to this month?’ And it will deliver an answer. But then we do two things. The first is to ask, ‘Do you mean…?’ and ‘Did you know…?’
So, ‘Do you mean people who only drink coffee?’ Because the system may have found an anomaly in coffee customers churning more. Then, ‘Did you know that tea customers have not been churning?’ So, that’s interesting. Then I might say, ‘Show me the results by gender’. Because I'm a marketing specialist, I maybe have an intuition that women are not returning for some reason, or men are not returning. Or the system might reveal that the customer ratings on the coffee are down, so did we change our supplier?
In all this, my ability to make a decision is what I'm looking for. And ThoughtSpot uses machine learning to move through pattern matching. Another example, is ‘Did you know that your article readership has spiked by 50%?’ And as you click on that data, it drives automatic drill-downs through machine learning.
So, what’s the second part of this equation?
That is to show why something happened, to explain it to you. And then explain what we expect the next three weeks to look like, for example. Or to say, ‘If you post on Wednesdays at 11 o’clock instead of Mondays at 9am, these will be the differences.’
So: What happened? Why did it happen? And what will happen next? That is what we think analytics should become.
Teaching people to fish
An impressive pitch. But isn’t there a problem with this way of thinking? It all seems reminiscent of what happened when publishers – and Web businesses in general – became obsessed with search engine optimization (SEO) to game Google’s algorithm.
Over time, many content providers employed more SEO specialists than writers, and told the latter what to write, what words to use, and when publish. The end result was the disintegration of brand voice, quality, human touch, and differentiation – plus, an industry-wide quest for the same disloyal clicks? A race to the bottom, not to the top.
Might this kind of analytics create a world full of clone businesses? Or one in which entrepreneurs look to a machine to make every decision for them, based on whatever was done in the past? Might it preclude a human being having a bright idea, or an inspired hunch? Put another way, isn’t turning everything into a data query a boring way to live?
Nair accepts the point, but explains:
Sure, but the focus for us is areas such as life sciences, healthcare, and manufacturing. And if you look at supply chain cost optimization, absolutely it's all about what time you should be shipping, and what the margin and price should be. Those are all good use cases.
Because the problem we are trying to solve is, the people who speak business don't speak data. And vice versa. We want to enable the people who speak business to interact with data without any tax on their curiosity.
Before AI-infused analytics came into the picture, data scientists and analysts were enterprises’ critical, premium hires. But now it seems that their day in the sun is over: just ask the AI. Correct.
This is an important question. If you are one of those people who are unwilling to change, then complexity equals job security for you. And ThoughtSpot is about removing complexity. However, it is also important to remember that there is a hype cycle happening in generative AI at the moment. And the data it supplies is not always accurate.
Most people are inherently afraid of numbers because there's no place to hide in them. So, governance and accuracy, those things are critical.
We are very keen to elevate the analyst into a world of data modelling. So, instead of giving people fish, which is what analysts are doing, we want analysts to go and teach people how to fish.
What about the growing power of AI companies, the rush to adopt their products – despite the lack of skills in enterprises to use them – and calls for the industry to police itself and fight off regulation? Do these constitute significant risks in the market?
[Our conversation took place before the OpenAI debacle, with Sam Altman now returning as conquering hero, with whatever he did to be fired in the first place seemingly forgotten.]
Practitioners are saying, ‘Let's have patience. This can be toxic if it is not controlled properly.’ Enterprise AI should be built with trust, governance, security, and privacy. If these four things are not there, don't do anything else. It’s almost like a Hippocratic Oath. First, do no harm here.
Large enterprise customers say they have never seen a wave that is crashing so fast. So, now more than ever, vendors need to be more responsible in pushing things in the right way. What's the hurry here? That's a responsible way of talking about it. Pick use cases where the cost of failure is contained.
There are tangible benefits that we can deliver for customers by responsibly deploying AI.
Nair’s mix of forward-thinking and candour continues to impress, as does ThoughtSpot’s progress to date. However, history suggests that CEO candour tends to become muted as companies grow and the stakes get higher. On that point, we will have to see.