The Bowery Capital team is kicking off a new blog series covering vertical SaaS. We are doing deep dives on various companies, interviewing founders and investors, and learning what it takes to build success in the vertical SaaS arena. This week, Sasha Novakovich, Founder and CEO of Alchemy, answers some of our questions.
Can you describe Alchemy for the readers and share a little insight on what inspired you to start the company? And who are the typical companies Alchemy is supporting?
Alchemy is a software company that serves chemical and material science businesses by providing a product development platform which enables them to make better products, faster. Our approach has been very intentional - we started with scientific workflow automation in the lab, and then we layered on specialized formulating and testing functionality. These are features typically found in an electronic lab notebook and laboratory information management systems, but once these features were incorporated into Alchemy this enabled us to generate a really great data set that spans the arc of product development regardless of what the underlying concepts are. As you test different hypotheses - formulation components and permutations - you are creating data on which Alchemy can run statistics and algorithms to speed up the development cycle. This is where we start to see a force multiplier effect; by combining that data with an AI overlay, we are able to cut down significantly on unnecessary testing. This means our users can innovate and bring products to market faster. As a company, this is super exciting because the world needs cleaner, greener, lighter, stronger, better products and we are helping bring these products to market. In terms of our sales motion, we are typically selling into the VP of R&D; however, the day-to-day users are chemists, material scientists, and the laboratory techs that support them. Alchemy helps the core lab staff to operate more efficiently while also providing reporting and visibility for the managers overseeing these labs.
For the chemicals and materials labs you sell into, what does their day-to-day workflow look like in the absence of Alchemy? Do they have existing software in place or are you basically digitizing an offline process?
Most of the workflow processes we are replacing are fairly disjointed - prior to Alchemy, many of our users relied heavily on Monday morning meetings where people would look at big Excel spreadsheets of the lab’s projects and how they were progressing. Every chemist or material scientist has a paper lab notebook, and they each tend to have a set of spreadsheet templates that they are familiar with and rely on. Many of the labs we support were also already using single-purpose software tools including custom built applications to achieve different objectives, so there is also a variety of existing software incorporated into their day-to-day operations.
The difference is that these tools tend to be disconnected from one another and from the larger company; there is often a lot of rogue software that is being used by individuals. The challenge organizationally is that often Excel is the path of least resistance for the individual because they know it, and have a template, and feel comfortable with it, but the challenge is that this does not scale company-wide. When you operate this way, the company is not in control of the superset of the information, and you cannot validate that everyone’s Excel calculations are even accurate. And if you have a calculation error in an Excel file, and it is in a document you are using over and over as a template, you are continually proliferating that mistake throughout the organization. The reality for a lot of the companies is that they still have a combination of online and offline processes and tools, but they do not have their data in one place and they are not able to optimize their processes and manage the superset of the data. That is why we started with workflow - and then as you incorporate different applications, you are generating data, and storing and consolidating that data in the Alchemy platform.
By having one platform where you do your work and where you store your data, Alchemy can help these labs with process optimization and better quality assurance. You want to be able to relate your ideas to your hypotheses, your formulations, your test data, your analyses, and the market feedback. When all those categories of data sit in different places, it is very hard to see the full picture and to optimize across your product development life cycle - that is what we are enabling our users to do when they work in a product development platform like Alchemy.
How did you arrive at the lab as the problem space you wanted to be working in as a software entrepreneur?
I am a second time founder. I started a vertical SaaS business a long time ago back in 1999 serving the telecom industry. After that, I was an angel investor for a number of years. While angel investing, I made an investment into a project management business which was serving the chemicals industry; through that investment, I saw these highly-educated people doing extremely manual and error-prone things with antiquated tools, and it was really pretty shocking. That is when I realized what the industry needed was not just project management tooling, what these scientists and technicians needed was a product development platform. A cloud-based project management tool could never be effective if all of the underlying workflows are analog and disconnected. The reality was that we had to go to the source of the problem, and the source of the problem was that these lab-based professionals did not have a modern platform to work and collaborate in. It really is one of those situations where one plus one is three - if you can see across all of your actual lab work, then suddenly you can start automating and digitizing and streamlining in a way that you simply cannot when everything is disconnected.
Can you offer some more color on how Alchemy speeds up that product development process for applied science teams who utilize the platform?
Let’s say you make some type of performance ingredient that is sold to another company and is used as an ingredient in the formulation of one of their products. Often, the right to sell this product is determined through a reverse auction. The company might reach out to four or five vendors and say “I need a material that has these 14 technical and functional performance parameters - who can make it and at what cost?” This opportunity could come in from a CRM, it could be in an email, or it could come from a phone call from a salesperson to the lab. Once the request comes in, you are trying to figure out as quickly as possible your starting point: i) Do I have anything off the shelf I can use? ii) Do I have something I can tweak that fits these parameters? or iii) Do I really need to start from scratch? On the lab side, this requires that you have really good data on everything that you have ever made that you have commercialized, as well as this whole other category of everything you have tried to make that you did not commercialize, but that may actually work under these circumstances.
Right away, what we see is a lot of rework - people will say, “I think Joe worked on something like that, over in another one of our offices, but I do not know for sure.” And even if you can find Joe's paper lab notebook or his locally-saved Excel files, you do not really know can I trust and validate this data? Were the circumstances the same for these experiments? More often than not, what happens is the chemist basically ends up redoing the work. What we can do with Alchemy is algorithmically analyze what a chemist has historically worked on, and then based on that data and the associated outcomes, we can see what worked and what did not. Using this background, we can then predict which trial formulations are most likely to work for a given problem, which provides the scientist a big head start. If we can come up with a suggested formulation that we know will work for the materials request, that is great; but even if the suggested formulation does not work 100% of the time, it can still be a good starting point and can help save time in product development.
Another possible scenario is that I might not have enough prior work in my records that is applicable to the current problem. This means I do not have enough data to use Alchemy to provide an AI recommendation and come up with some possible formulations. Instead, I need to design a set of experiments that will help me quickly figure out what is going on and will generate enough data that we can run against AI models in order to recommend a formulation. In that case, we use statistics to help our users create really efficient test plans - then we can run this new superset of the data through an AI model and actually recommend formulations. So when a possible project comes into the lab, the big questions are have I done something similar before? And if so, do I have enough data to run an AI model? And if not, can Alchemy give me an experimental plan to help me get to the finish line as quickly as possible for that prospect? Time is a huge factor in whether a project gets off the ground and if you can sell something or not - we want to help scientists achieve their targets as quickly as possible when servicing these kinds of requests or developing entirely new products.
Alchemy serves a few different verticals including building materials, industrial chemicals, and plastics and packaging - are there any verticals where you have seen particularly strong adoption of the software? If so, are there any factors that correlate with this outsized adoption?
Across the segments we serve, chemistry and materials science are at the core of product differentiation for our customers. This is a common thread across our user base. We generally tend to do very well with process-oriented companies. Where we have seen the most initial adoption would be around coatings, adhesives, sealants, and elastomers; we have done quite well serving the industrial chemicals segment more broadly and I think the reason for our strength in this category is two-fold. First off, there is a lot of complex formulation work that goes on in this segment, but frankly it is also where we started when we launched and it is kind of the legacy of our business.
These materials that your users are developing - are they being sold to someone as an ingredient or as an end product?
Our users work on both - they could be developing an intermediate - an item that is ultimately going to be sold as part of something else - or it could be an actual finished end product. In either example, there is literally chemistry every step of the way.
When you are trying to get a company to adopt Alchemy, do you take a bottoms-up, product-led approach, or is it more of an enterprise sale where you are selling it top down into the VP of R&D or someone like that?
It is definitely an enterprise-led direct sale; the data we ingest into Alchemy is often highly proprietary and not the kind of information you would want employees going out and sharing without authorization; this is not information that is just floating around on the Internet. Which actually brings us to another key topic, which is the complexity presented by the fact we deal with mostly proprietary data. This data is very closely held and chemists and material scientists are not at liberty to just put it out into the public domain. But beyond just data security, there is also the question of leverage - when people are out there each using their own tech stack for their individual lab, there is not as much leverage at the corporate level of that knowledge. You want to have a single source of truth and a shared corporate knowledge base - one benefit of relying on a top down deployment is that you can ensure that there is data labeling, taxonomy, and ontology consistency from product group to product group or business unit to business unit. You can ensure that everyone’s processes are set up optimally and you can do things like global roll-ups and other macro analysis.
At the same time, the trick is to give the individual user the flexibility that they want so that they feel like this is as easy as using Excel and I get just as much flexibility, for instance, around formulating inputs and calculation variables. We explicitly needed to build Alchemy to have all the things that people like about Excel, and I think we did take a lot of lessons from PLG in terms of building around what users want, but the sales motion itself is top down and usually involves selling to a corporate group led by the VP of R&D and blessed by the CFO/CIO who are needed to enable to sale to go through.
At Alchemy how long does it take from the initial outreach to actually go live with a company? In other words, what is the typical sales cycle and implementation period to get people in the lab up and running?
It really depends on the size of the organization and whether or not that initial contact is cold outreach or if that company is actively in some sort of buying process. What we see is that there are really three types of selling cycles we come across. First, there are companies that are already running an RFI/RFP process, and it is very structured and formal. They are literally running a process and you participate on their terms, and you follow their process, because that is what the company is mandating. Then, there are companies where they might be in some sort of buying process, and they have a clearly identified need, but it is not a formal RFP-style process. And the third type of sales cycle is one that is initiated by us doing cold outreach. What we have found is the amount of time it takes to get through each of those categories is really different. With an RFP, it could be a six month process because that is the company’s internal timetable and they have decided no decision will be made until Q4 so they can roll it out next fiscal year. When we run into companies where they are in some sort of buying process - often because whatever they are using today is not working or they have run into some real dramatic business need - then sometimes that time table can be shortened a little bit. The longest sales cycles are cold outreach - this is when we are reaching out to a company because we think it is likely to have the same common set of problems that many of their peer companies face. When we reach out, we raise those potential problems and then let them know how Alchemy can be a solution for their business. You can show them what those solutions are, and what some of the benefits are from deploying them, but it generally takes longer to close a deal as the buyer was not in a buying process to begin with. We are very conscious as we are engaging companies to try to tease out which bucket a company might be in and then act accordingly.
Can you tell us a little bit more about how Alchemy incorporates AI into its platform and what are some of those use cases?
I think we may not be that different from other verticals that can benefit from AI - when our customers are trying to devise products that are more performant than their predecessors on whatever dimension they care about, they are using the scientific method which is really just hypothesis-driven product development. It is very non-linear, very iterative, and it can be extremely time consuming, and if you have a lot of PhD scientists doing it, it can be very expensive. For certain products that require hundreds of iterations this could mean thousands of hours of lab time. If you can speed up the time to get to the finish line, it can be extremely valuable in terms of the company’s cost to develop a new product and can boost their ROI on the product itself. With Alchemy, we are using both statistics and AI together to try to streamline as much of the hypothesis testing as possible and we leverage a cloud-based, single source of truth to do this. We also use the aggregated data we have from past experiments to ensure that you are not wasting time re-doing earlier work. We identify this past work as quickly as possible, find the right starting point, and then we use statistics to help design experiments and leverage AI against the data you have already generated. Using Alchemy, we see our users cutting out a significant amount of unnecessary testing, lowering the needed testing by 80% to 90%. With Alchemy, we also put the data infrastructure in place that is generating AI-ready data as you are working; otherwise you spend most of the time trying to identify the data and determine its relevance. The way we have built out Alchemy, you do not have to waste all that time and that is what we are so excited about.
What advice would you give to someone who is looking to start an industry-focused software business?
First, I think you need to make sure the industry is big enough to support a venture-backed business if you are planning to raise venture dollars. Look for a big space that has a lot of problems, that is antiquated, and that could benefit from a vertical SaaS solution. Then, I would emphasize the importance of paying attention to feedback and being able to pattern match; as a founder you will have a lot of ideas and hypotheses about what you think is right, and you should not give up on your high-level vision but it is important to be very flexible about the execution based on what actual live prospects and customers actually tell you. Being responsive to what you are hearing is really critical. And if it is a great idea and a big space that is underserved, make sure you team up with the right partners with complimentary skills who can help you bring your vision to life.
If you liked “Vertical Visionaries: Sasha Novakovich (Alchemy)” and want to read more content from the Bowery Capital Team, check out other relevant posts from the Bowery Capital Blog.
