A Tale of Three ShotSpotters
A Look at the Leader in Gunshot Detection on the Eve of Its IPO
This began as a relatively straightforward project: I would work with Urban Us, where I am an active member of their network, to analyze the S-1 of ShotSpotter, the market leader for gunshot detection and location solutions for law enforcement and private security — and what we thought was a straightforward SaaS company— and compare it to other public SaaS companies of the same age cohort, across a set of key SaaS company benchmarks, a methodology developed by Tom Tunguz of Redpoint Ventures. This would provide an early point of comparison between the nascent investment area of B2G (Business-to-Government) SaaS, one of Urban Us’ key areas of investment, and the more “traditional” B2B business model of the bulk of SaaS companies. It became immediately clear, however, once I looked at ShotSpotter’s age and size, that making a meaningful set of comparisons between ShotSpotter and the larger basket of SaaS companies was going to be hard.
ShotSpotter was nowhere near the level of any of the venture-backed SaaS companies that Tunguz has analyzed, and for good reason. The company has two structural flaws: the first a challenging business model that requires expensive hardware and 24/7/365 manual review of nearly every alert the system sends to the customer, in addition to all of the ordinary costs or a SaaS company; the other a cautionary tale about the relationship between a company and its customers that emerges from a close reading of the history of the company and statements by former and current customers. The tale the S-1 tells of ShotSpotter’s current operations; how they got here; and what we can learn from its experience, is below; I will offer some thoughts as to what ShotSpotter can do to address these challenges in a future post. Meanwhile, I’m taking suggestions for a pure SaaS B2G public company to do the original analysis on; we are also conducting a short survey of B2G SaaS startups to generate a vertical-specific data set for benchmarking.
A Brief History
There are three distinct eras to ShotSpotter’s history, each representing a significant evolution from the prior, so much so that outside of the Silicon Valley jargon of the pivot they could almost be considered different companies, were it not for the reliance on the continuity of the branding. These three periods are briefly described below, followed by my analysis of ShotSpotter’s performance relative to its age cohort¹ according to Tunguz’ template.
The Science Project Era: 1996–2003
Robert Showen, a Silicon Valley engineer, started to work on ShotSpotter’s underlying technology — gunshot detection using sound waveforms picked up by at least three acoustic sensors to triangulate the location based on the difference in the timing between them — as a side project, in 1992. In 1996 he launched Trilon Technologies, LLC to commercialize the concept. However it would not become much of a commercial entity for some time. As late as 2000, its only deployment was in Redwood City, and until 2003 Showen, who at the time was President and CTO, was also working as a Staff Scientist at Raytheon, according to his LinkedIn profile.
The Product Era: 2004–2010
In 2004 Lauder Partners, a VC fund with money from the Estee Lauder cosmetics empire, invested a little under $2M in ShotSpotter, simultaneous to the company appointing James Beldock as CEO. This launched ShotSpotter as a “real company.” They hired engineers and salespeople. Showen could quit his day job and focus on his lab full time as Chief Scientist (a role he holds until today). The company merged with Centurist Systems in 2005, acquiring valuable intellectual property that enabled dramatic improvements to their technology. According Beldock’s Crunchbase profile, from the time he was hired until he left they went from four installations to 50; revenue increase 50x; they raised an additional $29M in venture capital.² Their product at this time was sold to law enforcement agencies as a traditional, legacy hardware and software package: the agency bought the sensors and a perpetual license to the ShotSpotter software for an upfront fee of around $200–300k per square mile, plus an optional 15% annually for maintenance that wasn’t really optional if they wanted to keep the system running.
While the service became more reliable and professional, there were still three primary issues that were putting drag on ShotSpotter’s growth and causing churn. 1) Initial costs were high, and because the full purchase was made up front it was tempting to just cut ties when things weren’t working as expected; 2) police departments didn’t know what to do with the data ShotSpotter was giving them. Some agencies dispatched an officer on every noise, resulting in high numbers of false positives and a perception of unreliability. 3) Deployments suffered from technical difficulties as law enforcement agencies hosted the software on their own servers.
The SaaS Era: 2011-present
Ralph Clark was appointed CEO in late 2010 and introduced a managed service model in 2011. This addressed the concerns above: it replaced the upfront payment with annual payments of between $40–60k per square mile, barely more than the previous maintenance fee; 2) it took care of the “last mile” of the data relay, screening and translating incident data into actionable messages that police could use to make tactical decisions, via the creation of an Incident Review Center (IRC) staffed with acoustics experts 24/7 responsible for manually screening alerts and supplying additional tactical information to police dispatchers; 3) like all SaaS products, it transferred the technical burden of running the system from the customer to ShotSpotter, who was better equipped to handle it.
In spite of the above, my conclusion from a review of ShotSpotter’s press and performance over the years is that ShotSpotter’s fundamental and continuing problem is not with its technology or product, but with its customer insight. From their founding as the brainchild of an inventor with no subject matter expertise to their current big data play that nobody wants, they have repeatedly taken a technology as their starting point rather than a customer need. They have gotten it right in some cases, such as the creation of the IRC, but it is stunning that even after raising outside capital it took the company seven years, four VC rounds, and a CEO change to realize that this was necessary (I am counting from the start of the Product Era. As the inventor and tinkerer who had no experience in law enforcement or business and never really seemed to get comfortable outside of the lab, Robert Showen, and with him the Science Project era, gets a pass on customer development. It is those who came after who should have known better).
Under Clark, the primary example is the treatment of data, though there are others. One of the benefits of the service/SaaS model was that it allowed them to control all rights to the data (customers have access of course). The idea is that the data is a valuable, saleable asset, particularly to federal law enforcement agencies such as the ATF or the FBI, when bundled into a national data set. Like most things that companies see as their product, ShotSpotter has refused to give away the data for free, for any purpose. Unfortunately, however, ShotSpotter has been coming under criticism for this absence of data supporting the effectiveness of what is a high-ticket line item on local government budget rolls that sits at the intersection of two politically sensitive topics — policing and government surveillance. Here this data could be invaluable towards gaining public goodwill and overcoming objections about the product’s cost, but they have refused to work with academics who could agree to maintain confidentiality regarding details of the data and only release conclusions, and have been actively trying to prevent their customers from complying with Freedom of Information laws. Although the connection has not been made explicit, I imagine that this hurts their image when it comes to the controversy around whether their acoustic sensors comprise government surveillance (in fact the company goes to great lengths to ensure that the technology is not surveillant, but their public relations effort to communicate this is lacking).
To tie this back to the lack of customer development that ShotSpotter just doesn’t seem to be able to shed, it turns out that federal agencies just aren’t interested in the data, and their incentive structure is such that this kind of a sale is unlikely to ever happen. So all of this data hoarding and data secrecy is for naught (unless of course the data demonstrates that ShotSpotter’s technology, in fact, does not work, in which case there are bigger problems).
Benchmarking Analysis
Notes about formatting: In the graphs below, I’ve used ShotSpotter’s colors as a consistent legend. ShotSpotter’s data are in red, with square end caps; median values are black; other companies’ values are in gray to highlight the comparison to the median.
The different eras of ShotSpotter raised the question of to what age cohort should the company be benchmarked.³ To avoid this debate, I compared all three possibilities for each metric, and it turns out that the difference matters less than expected. The data representing the company when benchmarked as having launched in 1996, at the start of the Science Project era and turning 21 this year, is the thinly dotted line stretching from years 19–21 (in certain cases only hitting two of those, such as when the metric required calculating a rate between data points or when making a projection was not possible with the given data). The data representing the company when benchmarked as having launched in 2004, with the start of the Product Era and turning 13 this year, is the thicker dotted line stretching from years 11–13. The data representing the company when benchmarked as having launched in 2011, at the start of the SaaS era and turning 6 this year, is the full red line stretching from years 4–6. This is benchmark that is worth paying the most attention to and that I will discuss in greatest detail, as it represents the company in its present form, and highlights the struggles that even now it has in breaking into a high growth, high value creation mode.⁴
Revenue
ShotSpotter’s revenue of $11.8M and $15.5M in the two years preceding its IPO is the lowest of the 35 companies I benchmarked them against. Year “4”, counting from the SaaS era, is close to matching five of the 35 at same age, but all of these had growth rates well over 60% for each of the next two years, with three of them having growth >100% for year 5, leaving SST in the dust. Benchmarking the company from the Product or Science Project eras is even worse, with the median SaaS company of those age cohorts doing 7 and 11x as much revenue, respectively (admittedly there are only three public SaaS companies on the chart that are old enough to have data for years 19–21 since founding, but if you look at the trajectories of the other companies it is clear that the median’s multiple of ShotSpotter’s revenue would be even worse).
Revenue Growth
Most companies grow more slowly as they age. SST’s growth of 31% in 2016 and 17% in 2017 (projected), years 5 and 6 from the launch of the SaaS Era, would be middle of the road for a company benchmarked from the beginning of the Product Era (13 years old today) or the beginning of the Science Project Era 1996 (21 years); but for benchmarking from the beginning of the SaaS Era (5 years), they are growing more slowly than 9 out of the 10 companies with data from this year. The median SaaS company at this age grows 91% in year 5 and 67% in year 6.
This makes sense, because most companies grow at a much faster clip when they are younger, and then see their growth rates drop as they “use up” their early adopter pool and begin to saturate.
Gross Margin
Most SaaS companies start off with lower gross margins due to fixed costs of server capacity, customer support, etc, which then rapidly rise in years 4–6 as they start to scale revenue and spread those fixed costs across a larger base, until they start flattening out in years 7–9 and cluster around a median of 65–70%.
Here SaaS Era ShotSpotter is a bit of a question mark.⁵ Their gross margins of 30, 38, and 41% in 2015, 2016, and 1Q 2017, respectively, are the second worst in each of those years, but they are steadily, albeit slowly, improving. The real question will be whether they can make the leap into the 50s by the end of this year (year 6), which every other company in the benchmark set did with the exception of Medidata. Medidata did not break 30% gross margins until year 9, which it’s worth noting was their IPO year. Today Medidata has revenue of $463M, 76% gross margins, and a market cap of $4.7Bn, showing that an economically significant outcome is still theoretically possible for a company of “SaaS Era” ShotSpotter’s trajectory. On the other hand, the fact that one company made this leap does not put the odds in SSTs favor.
The real question is why are ShotSpotter’s gross margins so low? There are two ways to look at this. The first is to look at all of the costs ShotSpotter has in providing their service. Like all SaaS companies, they bear the cost of hosting and maintaining their software, whether in a public cloud like AWS or on their own machines, as well as the cost of customer support. ShotSpotter provides professional services around installation and setup, which is likely a drag on gross margin, but without any information about the revenue split it is impossible to say anything specific about the impact. It may be less of a drag in the short term as it is for companies that recognizes the both the revenue and cost of installation to the year in which it occurs, as ShotSpotter recognizes these “ratably, on a straight-line basis, over the estimated customer life of five years” (what happens if the customer life is less than five years is unclear; do they go back and revise previous years’ financials if the customer unexpectedly terminates the relationship before five years are over?); on the other hand, it will take them longer to see the improvements in gross margin that other SaaS companies see with growth as professional services become a smaller portion of revenue relative to recurring contract payments.
In addition to the standard SaaS costs, one of ShotSpotter’s major cost inputs is their acoustic sensors, something that is more typical of an IoT company. These are purchased as capital goods and then enter gross margin via depreciation as well as the direct cost of providing them with power and data connectivity. SST’s depreciation schedule may in fact paint an overly optimistic picture of their gross margin; it again assumes a five year average contract length, but provides no rationale for this assumption — SST has barely even been in the SaaS business for that long.⁶ By dividing the amount of capital spent each year on deployed equipment by the growth in area covered I was able to estimate an equipment cost of approximately $25–30k per square mile, or as much as 10–15% of annual revenue assuming a five-year depreciation schedule (not to mention the working capital demanded for growth and the risk of holding sensor “inventory” that isn’t earning revenue).
It is hard to see them reaching the kind of scale that really drives manufacturing costs down any time soon. At 15–20 per square mile and expanding at its current rate of about 120–150 square miles per year, that’s 3,000 new sensors a year, best case, for a custom sensor that has to be ruggedized to survive five years outdoors. Additionally, as they state in their section on Risk Factors, they use a single manufacturer, presumably because of their limited scale, giving them limited leverage to negotiate pricing.
Finally, they have to staff an Incident Review Center with trained acoustic experts, 24/7/365 to “screen and confirm” every single incident reported by the system, before they get sent to customers. I don’t know enough about gunshot acoustics to say why they can’t develop software that is good enough to automate this process, but until they do this will be a major cost center: operations personnel comprise over half of their employees.
The other reason is scale, not only to reduce the unit costs of their acoustic sensors but for all costs, including cloud infrastructure, R&D, and general and administrative overhead. The chart below shows the relationship between scale and gross margin via a master set plotting gross margin vs revenue for all data points in the benchmark data sets (I’ve cut off the X-axis at $200M but you get the same result if you extend it). While it is possible to have good margins right from the start, once you get past $100M in ARR, every single company in every single year achieves gross margins >50%. In fact, you really only need to get past $70M; the last dot on the graph with gross margin below 50% is at $68M.
S&M Spend
This is where the story of ShotSpotter as a “reborn” SaaS company benchmarked against other 4–6 year old SaaS companies starts to unravel. By all measures, if they are trying to manage a high-growth story in these still early years of the SaaS strategy, this is the time to pour on the gas. Instead, their Sales & Marketing spend as a percentage of revenue is only 33, 29, and 24% in years 4, 5, and 6 (Q1), vs a median of 72, 49, and 48%, respectively, or around half of the median. Only LinkedIn, the most viral of platforms, spent less than them at this age.
The relatively low spend on S&M on a percentage basis is doubly strange considering the low S&M Spend per $ of new annualized contract value they claim of $0.37 in 2015 and $0.28 in 2016. If every 28 cents of S&M spend turns into a dollar of revenue within 12 months, they have a retention rate of >99%, and they a revenue retention rate of 127%, why aren’t they pouring the gas on S&M? The only reason I can think of is that they have been simply capital constrained. They are already spending so much on everything else that after this many years, the private and public financing markets are only willing to give them so much room to experiment before they demand profitability.
R&D Spend
R&D spend as a percentage of revenue is fairly unremarkable, falling right in the middle of the pack.
Net Margin
ShotSpotter’s “SaaS era” data falls in middle of the pack again on net income margin, whereas the other two eras each show Shotspotter far below the median benchmark, respectively. Almost all venture-back SaaS companies go through years 3–9 or so with high negative net margin, in their effort to achieve growth, before hitting profitability around years 10–11. Whether ShotSpotter’s trend towards break-even will continue until they cross the line to positive remains to be seen.
Conclusion
“SaaS era” ShotSpotter almost matches the benchmarks for comparable companies of similar age, but their gross margin comes up short, and despite a small improvement over the last 15 months the structural realities to delivering their product seem stacked against them. Neither hardware nor manual services businesses scale like SaaS, so to get to profitability would require a radical innovation allowing them to get rid of one of those components to service delivery. Furthermore, with such tight gross margins, they can’t spend aggressively on sales and marketing, constraining growth and the ability to scale.
Unfortunately, given the reality that ShotSpotter has been around for 20+ years and carries such a long legacy with it, people are going to find it harder to believe that the company can reinvent itself than that a new company can execute. This will constrain the company’s ability to attract further capital and top talent. In many ways, I think it would be better for ShotSpotter to have had the benefit of recapping and wiping the slate clean with a new cap table with a larger ESOP pool and perhaps even a rebranding. As is, the drag on the company from the legacy business may prevent the new SaaS model from ever attaining escape velocity.
¹ I was initially dubious about age cohort benchmarking — shouldn’t benchmarking be done to comparable companies at present, the way valuation comps are done? However after reading Tunguz’s post explaining his choice of benchmarking methods (The Best Way to Benchmark a SaaS Startup), I am convinced that this offers more nuance. Another way of thinking about it is to consider which of the following two questions yields the most information: Is Company X growing at a good rate for a 5-year old startup? Or: Is Company X growing at a good rate for a public company? There is certainly value in the latter, and if the market is currently valuing companies by its expectations of a company’s ability to weather particular market forces or turbulence, rather than the company’s fundamentals — e.g. during the credit crunch of 2008, or in the face of great political uncertainty or currency collapses — then this probably is the appropriate method for benchmarking. Under normal conditions, however, it is the company’s performance for its stage in the corporate life cycle that will tell you the most about it and its potential for growth, which is typically the most important driver of shareholder value.
² Unclear whether he means until he stepped down as CEO in 2010 or when he left the company altogether in 2014. The fundraising number refers to funds raised under his tenure as CEO.
³ While considering the company to have been launched the SaaS Era data gives the best apples-to-apples comparison between ShotSpotter and other SaaS companies, all three of them have logic — the least of which is that this last one is cheating by manipulating the rules to make them look more favorable according to Tom Tunguz’ methodology of age cohort comps rather than comps of current results from companies at all different ages.
⁴ The data on the comparison companies is all taken from data sets shared by Tunguz covering a period between 2004 and 2012. Ideally this research would be repeated with data that continued through FY2016, however there is good reason to believe that the impact would be relatively minor. By comparing companies across an age cohort, you are already saying that the calendar year in which they hit that age is less important. Whether the company is 4 years old in 2004, 2012, or 2015, the relevant metadata is that the company is four years old, and can be compared to other companies that are four years old at any of those times. Over long enough time periods benchmarks can distort as entire industries shift from growth mode to a steady-state harvest, but that has not happened yet in SaaS.
⁵ Benchmarking ShotSpotter the Science Project Era company and ShotSpotter the Product Era company puts them so far below the pack that they are not even worth discussing.
⁶ The closest they come is in Note 3 to their financial statements, the Summary of Significant Accounting Policies. Their revenue recognition policy on revenue from setup fees reads as follows (emphasis mine): “The Company recognizes revenues from setup fees ratably based on the expected customer relationship period, typically over five years, which may extend beyond the initial contract period. In determining the expected customer relationship period, the Company considers specific customer details and renewal history with similar customers.” In other words, they make a judgment call that may vary with ever case and may be longer than the signed contract. This may make sense from a business perspective, but from an accounting perspective it is ripe for manipulation, particularly when it represents such a significant portion of their overall COGS.
