Search-as-a-service: How we undercut the market price by a factor of 30
and beat the performance of the leading open-source solution by factor 20
It's time for turnkey, affordable, scaling, high-performance search:
We just love search!
Not that it matters for this post, but we are good at search. With 20 years in search, and three start-ups from federated search, peer-to-peer web search, real-time-search to high-performance enterprise search we gained a lot of experience in what works, what scales, and what doesn’t – sometimes the hard way.
So, we wanted to stay in search - but the search-as-a-service market wasn't pristine anymore - there was already a strong incumbent. Nevertheless, we wanted to go through the front door with a big bang, and not just sneak in unnoticed with some boring "me too". We had to offer something of big value for our potential customers, which at the same time was impossible for the incumbent to replicate.
How to differentiate?
That's probably a question many startups have that enter a market in a second wave, with competition already established: What can we come up with that they don't have and can't copy. To compete with an incumbent is hard, especially if they have a great product - and admittedly our competition has a strong one, albeit expensive.
If we come up with an extra feature, their big team could copy it. But on the other hand, an incumbent with a big team and a large existing customer base can't compete with the price. They need to maintain their high prices and a revenue stream to cover their considerable recurring costs and fulfill the expectations of their investors.
No venture-funded company can cut its revenue stream suddenly by 97%. No investor would tolerate this and say, hey, instead of waiting for 3 years for our exit we are now just waiting 100 years, instead of expanding the team we shrink it. What a fatal signal would that be if they ever wanted to go public.
So, we had to get/stay shredded, ripped, and lean - losing all fat, both in architecture and in the organization in order to be able to compete via price. But let’s see, step by step:
What problem do we solve?
We are not the first to offer search, but the solutions available so far have their own shortcomings:
- There are some great open-source solutions, but with performance and scaling limitations. And they are not really free, as soon you calculate the cost for deployment and operation.
- And there are search-as-a services that are performant, but expensive.
So, SeekStorm offering 30x more documents and queries per dollar makes the competition suddenly 3000% more expensive.
While price and performance are our market entry strategy, but with Private information retrieval (PIR) we are working on functional differentiation as well.
Why do we offer an orders of magnitude lower price, how can we afford it?
We can offer a lower price because we have a lower cost:
- We are a bootstrapped, self-funded start-up with a lean team. We have an efficient technology that allows for high performance on inexpensive hardware.
- We choose to invest in technology development, instead of buying iron (large server farms with giant recurring costs).
- We don't have to fulfill expectations of exponential growth; nobody is standing along our way with a sign "End of runway!". We have no investors who expect a high investment multiple in exchange for millions in funding and valuation. We declined taking money and becoming dependable, we invested our own time and money for years, without taking external funds. That required a deep faith and firm confidence in our abilities and the technology we built.
- Because an affordable price and generous contingents are part of our unique selling point, and when entering a market with strong incumbents you need to differentiate.
- But also because of a lower price opens up the market. We don't need to switch existing customers from our competition, but we can reach a lot more new customers. Those for which the high cost previously was prohibitive. And we think that’s the long tail. There are those who are well funded, and those who absolutely need search as the core of their business. But the vast majority of companies, startups, and researchers are on a tight budget, and while search could improve their user experience, it’s not a necessity. So, because of the prohibitive cost, they either forgo those features or use inferior, seemingly free solutions.
How low can we go?
It’s much better to paraphrase this into “how many queries and documents can we offer for a certain price”. You have to calculate how much the operation of a server of a certain class costs you, and how much customers of a specific plan and contingents it can serve. Here we can see that architecture and algorithms are paramount to keep the cost at bay and become competitive, both in price and performance. Sprinkling some marketing fairy dust on some legacy technology at the very end won’t cut it.
Add to this the cost of your company, employees, software licenses, and taxes. In the end, a company does need to pay its bills and employees. And probably some profit (there I said it). Why else would you take all the risk and uncertainty over the years to bootstrap and materialize your dreams? And it’s an eternal question which way to achieve this is the best.
Why didn't the market leader offer a lower price?
We don’t know, and we don't want to speculate. Perhaps missing competition that allows them to charge whatever they please or competition with a technology that requires expensive hardware to be performant? Or a large team that comes with huge staff costs, expensive offices at many capitals of western Europe and the US with high wages and lease? A large marketing budget, an expensive sales team, or costly affiliate programs. Or investors who expect a high investment multiple in exchange for the hundreds of millions of dollars in funding?
We believe there isn’t just one way of doing business.
A modest price allows going much broader, to approach a part of the market that was an inaccessible terra incognita due to prohibitive pricing. And it’s an organic, sustainable, and fair win-win scenario for both service provider and customer. And no, fair pricing is not ruinous, thanks for asking. There is still a stable margin, that allows us to live and invest in further development and expansion. And it’s no Predatory pricing either, because it’s sustainable in the long run.
That efficient architecture — what’s really different?
SeekStorms' efficient technology supports 200-fold concurrency compared to the benchmarked open-source solutions. This is the equivalent of 200x more users per server. Another way to look at it is, that SeekStorm requires just one server, where you would need a whole datacenter with 200 servers to achieve the same performance (concurrency & latency) with the benchmarked open-source software, either for running your own on-premises search or for operating a search-as-a-services as a provider, which could possibly explain those high market prices.
Because SeekStorm utilizes the infrastructure more efficiently, we can pass on the lower operational cost and give you more performance and more generous contingents (docs, queries) at a lower price. Additionally, the user benefits with every single query from the 20x lower latency. The 200x better efficiency on the same hardware enables SeekStorm to offer 20x better latency AND e.g. 20x more queries per dollar to our customers AND at the same time to be 10x more profitable as a company, compared to a search stack based on the benchmarked open-source software.
That's why sometimes it's worth to invest in technology research, even if it looks like reinventing the wheel at the first glance, with all those seemingly free open-source search libraries available.
Highly integrated approach
To achieve the best scalability and performance while keeping operational costs (and prices) at a minimum, we had to build everything from scratch: database, search server, crawler. Only by getting full control of every byte processed, transferred, and stored, we were able to strictly optimize for search, performance, and efficiency, getting rid of boundaries between different packages and jettisoning unnecessary ballast. Because SeekStorm utilizes the infrastructure more efficiently, we can pass on the lower operational cost and give you more performance at a lower price.
Novel Log-structured merge-tree database
To achieve low latency under high concurrent load, for billions of documents, in a multi-tenant architecture, we had to rethink search-specific database architecture and take the Log-structured merge-tree (LSM-tree) to the next level. We built a resilient, reliable, and highly parallelized architecture with uniform term-partitioning, n-fold double buffering, zero-copy, lock-free synchronization, thread-per-core architecture optimized for high throughput, short average- and tail-latencies for high-concurrency and web-scale scenarios.
Remote-first: Because there is more than just technology
Long before the whole COVID thing, we decided to stay lean this time. In my previous companies there was always a big office involved, and so were substantial costs. Today with WhatsApp and Zoom you don’t need a conference room anymore. With your potential customers around the globe, do offices in different towns brings you really closer to them? Isn’t it more about how well you understand them and how fast you react, than how close you are geographical? Remote-first isn’t just about saving office costs. It is enabling so much flexibility time-wise. Also, sourcing employees around the globe, determined by their skills only, and not where they live.
For us, what’s changed with COVID, is not the way we work, but the general acceptance of our remote-first approach.
You are invited to test
Our 14-day Free Trial allows you to test all features for free. No credit card required; you won't be charged at any time. The trial ends automatically after 14 days and all your indices and documents will be deleted permanently. No manual cancelation is required.
Wolf Garbe is CEO and co-founder of SeekStorm, Sp. z o.o., which operates an affordable, high-performance and scaling search-as-a-service.