The near-decapitation (by acquisition) of the BI space in 2007-8 was perfectly timed for QlikTech, whose QlikView is rapidly becoming one of the leading independent products. This is hardly new; since its founding in Sweden in 1993, the company’s timing has been unerring. Years of slow growth – “From 1993 to ‘99 we had a grand total of 5 customers, all in Sweden,” Senior VP Anthony Deighton told me recently – ended abruptly with a funding round from a Swedish VC. That led to a revamped management team with an eye for growth, and in the next 6 years, QlikTech climbed to 1500 customers – still mostly in Sweden, though some were in Germany and a few were in the US. The next big bet was what Deighton calls “an over-investment in direct sales” based on a 2005 round of funding from some better known VCs. Since then, the takeoff has been remarkable – a happy timing of product, platform and market. With $120 million in revenue and 50% growth in 2008, Qlikview is reaping the benefits of effective timing, a conservative ramp that did not overreach, and a technology landscape that is paying off its visionary design. Now, wth QlikView release 9.0, it’s targeting enterprise scale, better performance and manageability, and mobile deployment. As the economy begins to recover and mobile platforms proliferate, it appears QlikTech’s timing is once again dead on.
QlikView’s vision has been matched by 64-bit hardware and operating systems that can leverage its simple proposition: memory is faster, and visual interfaces appeal to the long-sought business user by offering “associative analysis.” QlikView is positioned “on the right side of Moore’s law,” Deighton likes to say, and its current lead positions it well for some time to come.
Of course, there’s an element of luck. Our architectural model – associative, in-memory analysis – was ahead of the market in 1999. Memory was limited and slow, and processors were expensive. What we did correctly was see it and not jump too early.”
In its second stage, QlikView’s VC backing allowed it to prove its repeatable sales model and its vision of the market segment it would pursue for the next 5 years, ramping to local success. “We proved the sales model, and showed that we had a product that works,” Deighton says. In 2005, a $12.5M venture round funded its international growth and rampup, until today, with 10,500 customers and nearly half a million users in 92 countries, the firm is profitable and cash flow positive. QlikTech now has a formidable management team, 530 or more employees in 12 countries, and over 800 partners. Its awards fill pages, its customers and partners describe themselves as “fans.” In a market where studies have shown 80% of BI implementations take more than 6 months, QlikView’s built-in data integration and intuitive interface have opened many doors – as well as its free download, “try-then-buy” model. QlikView has institutionalized a “Seeing Is Believing” model for its proof of concept activities, and has leveraged it to an impressive string of wins. The question of scale is comfortably in hand as customer stories involving terabytes of data – tens of millions of records or rows – and thousands of users continue to pile up.
By all accounts, QlikView’s performance once in the door has anchored those accounts well. Half of its business is from existing customers. And it’s a product business: 90% is licenses; very little is services, reinforcing the “ease of implementation” story, while it drives most services to partners. The average deployment time (not project time, which is longer, as these things will be) is a couple of weeks, and QlikView offers a 30 day money back guarantee if they don’t achieve a working production environment as promised. Design is still a significant challenge for all products in this category, though, and QlikView could certainly profit from some discovery-and-design modeling help.
QlikView’s analytic interface is a revelation for first timers. The metaphor is focused on driving interaction. “Never tell the users they can’t click on something,” Deighton says. Typical displays show relevant data in context, showing grayed-out data that makes it clear what is NOT included in the display. As users move through styles and drilldowns, selections made follow them from tab to tab. For this observer, what was happening was “contextual analysis” – as new questions were asked and answered, the thread that led me there was not lost. I could backtrack and take a different tack, or even, with “trellis charts,” look across a single dimension 3 different ways on the same chart to compare.
Recently, QlikTech has delivered QlikView 9, with more than 100 new features, including a new cloud deployment option via Amazon’s Elastic Cloud Computing (EC2) platform – although not a SaaS model per se, as David Raab points out here: as more of the market shifts its pricing philosophy, this seems to be one place where QlikView’s timing is not ahead of the curve. Still, the EC2 option lowers barriers to entry still further – even if a prospect doesn’t have the hardware and storage for a new system, they can prototype in the cloud, and even opt to leave it there. In addition, QlikView Personal Edition is free, and allows QlikView Desktop to compete with the new open source competitors springing up.
Scalability and performance are also key directions, as larger deployments drive changes to the product. The previous limit of two billion rows per data set has been removed. Continuous update, load balancing features and optimization techniques help to ensure that complex queries run efficiently in clustered environments. A new control panel shows resource usage for all servers and components. QlikView has also added several new visualizations with QlikView 9, including charting features such as “spark lines”, “whiskers”, the trellis charts mentioned above, and live chart backgrounds. New global search functionality allows a user to search every QlikView field simultaneously with one click.
Finally, one of the sexiest introductions is the iPhone version. It makes effective use of familiar iPhone metaphors like coverflow, and supports GPS functionality, which should be a boon for a new class of mobile applications. (Since the iPhone version defaults to their demo server, you can see it in action immediately, by contrast to some other mobile BI products I’ve looked at.) The demo apps aren’t perfect; I played with the wine selector for a while and when I got to a detailed list I was dismayed that I couldn’t seem to zoom in using the usual iPhone gestures to read it better. But that’s execution, not product, I believe. In addition to the iPhone, other phones capable of running Java, like the BlackBerry and Symbian-based smartphones, have their own apps. The timing is right; I suspect this may become another takeoff market for Qlikview.
Hello Merv, nice read, I mentioned your post on my blog http://www.quickqlearqool.nl
Thanks. Appreciate the mention!
I think it still takes quite a while to become proficient at QV, sorry. It’s not something you can play with right out of the box. That and their in-memory architecture baffles me.
Merv touches on many of the attributes that make QlikView a valued Kalido partner. The combination of QlikView and Kalido offers end-to-end, enterprise-class BI that can be deployed and adapted very rapidly without sacrificing scalability or performance. QlikView’s EC2 option is also consistent with our desire to break down barriers to entry, as evidenced by our subscription-based offering for the midmarket. We see good things ahead for QlikView and are happy to be working with them.
For a guy like you to say that sets me back a bit. Maybe you are trying to do harder things than I was thinking about and playing with.
As for the architecture, it’s the fact that it’s in memory (and the speed that bought)that has been an advantage, and I did not go far into structural questions. As others go there too, the details will become more important.
Well I’m thinking of the richness of their ETL macros as well, for example. It takes proper training to leverage those well I think. It’s a very deep product. Sure you can get some basic stuff done fairly quickly but, to build the nice dashboards, load from multiple complex sources and full exploit the tool I think requires substantial training and experience. On the DIM side, I don’t know how they can deal with 50TB+ systems elegantly (perhaps I have not dug deep enough and things have changed since last year as you seem to point out). I do think they will encounter serious competition as the volumes increase, from SaaS, and from “simpler, gentler” approaches (like Lyzasoft.com for example).
When I said “not project times” and suggested that’s longer, sourcing and design were what I had in mind. So we agree – and getting the UI clean and simple for users takes longer.
As for what happens in memory I claim to special understanding, but it will make for interesting followup as we dig in – removing limits and working effectively with the larger volumes effectively are two different things, o be sure.