Searching for Signal: The how and why of financial models for startups
By Paul Bennetts via paulbennetts.co
It is widely believed that building financial models are of limited benefit for a startup. They take a long time, they’re immediately shelved and need to be changed just as quickly. This is all mostly true.
But before committing yourself to a startup, you should think hard about what you are trying to achieve. Grasping the levers of a simple model can be the best way to understand what is probable, what is possible and what is just insane. It is giving you (and your investors) a framework for understanding what has to be believed for your startup to be a success. The further from probable the risker your endeavour will be.
In fact, every startup really has only 3 to 5 levers that will determine its future value. In ecommerce, this will most likely be conversion rates, level of organic traffic, repeat rates and contribution margins. Every lever can be bundled to a level of abstraction (CAC and LTV in this example) or unbundled to uncover underlying mechanics. Conversion rates, for instance, are a function of many factors including user design, usability, brand equity, convenience, advertising effectiveness, net promoter score, customer satisfaction and more.
It is through unbundling or bundling to the right level of abstraction and flexing each lever that you can understand what’s probable, what’s possible and what’s insane. For example, you may decide to target a 100,000 users for your web app 12 months post launch. Unbundling this, you will see that if you have 1,000 beta users you will need to grow your user base by 52% every month for the next 12 months, starting with 520 new users in the first month. Then unbundling those 520 new signups, you will need 52,000 unique visitors to your web app in the first month (assuming a conversion rate of 1%). Only then can you start to plan what is possible. Where will that traffic come from — organic (direct, referrals, search) or paid traffic? What activities are required for each of those channels? How much will each of those activities cost in manpower or capital? You might map out that 4,000 site visitors will come from guest contributions to blogs, which may mean writing four pieces of content that will generate on average 1,000 site visitors each. The trick to building a good operating model is to choose the right level of abstraction.
The unbundling of the levers in your startup plot what is probable, what is possible and what is insane. But it also assists as a set of goal posts to measure over time and change as you progress. As Peter Drucker says, “what gets measured gets managed”.
The best way to map this out is to get inside a spreadsheet. This exercise should allow you to build a more full hypothesis around your thinking. Unfortunately, most startups approach spreadsheets the wrong way — top down and overly complex. A model built top down rather than bottom up will start with a revenue line and grow it by some arbitrary monthly growth rate. It’s really hard to turn a bundled growth rate into an action plan.
Or, a model will be built bottom up but in an incredibly complex way. It will be multi-sheet, multi-year, multi-currency and map out to a level of detail that inhibits insight. They create more noise than signal. Startups need to find the signal and hide from the noise.
To find the signal and hide from the noise, a financial model should fit on one page and try be less than 30 rows long. Only then can you see that an investment hypothesis can be reduced down to what one has to believe from those 3–5 business levers.
Operating models should map out a journey to the next milestone. That usually means 12–18 months in length. Not five years. The world will change in 24 months let alone 60 months. The use of a milestone is a deliberate attempt to tie startup models to their funding requirements. A startup will raise enough capital to fund a cash burn for the next 12–18 months. This ensures they don’t over capitalise the business or over dilute themselves. A milestone also helps to frame the question — what do we need to achieve to raise the next portion of funding?
Let’s show this for say an ecommerce startup. From here, we will build a simple one page model. If this level of detail is not of interest — thanks for reading to this point. Please leave a comment below and join my mailing list for future articles here.
The caveat to multi-year forecasts, is to plot what the business could look like when its grown up. Getting a sense of what your startup will look as a very large business can be a valuable exercise. See how we did that looking at Etsy’s S-1 filing here.
As we build this out, take note of how to think about each business having only 3–5 levers. Versus the >20 assumptions that need to be bundled together to influence the model.
For ecommerce, the first question to ask to build a model is how a sale occurs. A visitor lands on a website from somewhere (direct, search, referral, etc.), chooses a product, purchases it and it is then sent. Let’s model this out.
Website traffic comes from many sources. It is your goal to build traffic from each of those sources. Splitting these out will force a startup to measure each going forward and think about the mechanics of building traffic from each source. The best way to split out traffic is between paid and organic. The best startups can build their business based on organic traffic.
Organic can be made up of direct, email, seo and content marketing. In the early days, this is likely to come from blog writing, press, sharing your site with everyone you know, etc. Some may break down each of these components to plan how they will generate traffic from each — but that’s beyond the scope of this simple operating model.
For this exercise, let’s assume that we can get 5,000 UVs organically in the first month. Then set a goal of growing this 10% per month. This growth will be driven by activities such as content marketing, press, search traffic, etc.
The next step is to look at paid marketing traffic. Paid marketing spend should be approached with caution. A customer you have to acquire through paid advertising is typically the worst type of customer you will acquire. You have to persuade them to turn up and make a purchase. They probably won’t return unless you persuade them again. Every customer metric (conversion rates, repeat purchase rates, referral rates) are all typically much lower through a paid channel than an organic channel. An organic customer, on the other hand, wants to buy from you. As a result, an organic customer will have a better conversion rate, repeat purchase rate and hopefully an abnormally high referral rate. The best startups shift money allocated for paid marketing to product to develop an experience that customers want to refer.
That aside, paid traffic is measured by a cost per click. The simplest way to model this is to assume a cost per click (CPC) rate and assume a monthly paid marketing spend. For example, one could assume a $5k monthly budget with a $1.00 CPC rate. This would generated 5,000 site visitors for the month.
Some suggest not to grow the budget for paid marketing. It can fool you into thinking that the key to growing is through buying clicks. Instead of the harder path of building better product-market fit.
The next lever to model is the most important — conversion rates. Conversion rates vary depending on the type of product being sold but a typical startup should hopefully be converting 3–4% of traffic. Most ecommerce startups you will see though are most likely achieving about 1% conversion rates.
You could model a low conversion rate to begin with and then asusming that through product improvements and returning customers it will improve over time. You may also choose to model out repeat rates but that is beyond the scope of this simple operating model.
The next levers are average order value (AOV) and gross margin. This will be dependent on the type of product you are selling. It’s probable that your AOV will remain flat but it’s possible that it could expand. Here we will stick with what’s probable and map out a flat AOV. Gross margin will be dependent on type of good and whether the business is vertically integrated (manufacturing your own product).
Traffic multiplied by a conversion rate gives order count. This multiplied by AOV and gross margin gives contribution margin before marketing spend. From here, we will take out marketing spend, to give a contribution margin. Contribution margin is probably the most important financial metric to focus on as you scale. After contribution margin, it’s simply your fixed cost base — number of heads, office/warehouse space, contingencies, etc.
Based on these basic assumptions, we now have a simple model for an ecommerce startup. By flexing organic traffic levels, paid traffic levels, conversion rates, AOV, gross margin and fixed cost base, we can determine what’s probable, what’s possible and what’s insane with this startup. A super useful exercise for every startup.