We use “product-market-fit” to make critical decisions such as when to scale but we lack a scientific, data-driven definition of the term.
“What is product-market-fit?”
Every year, I challenge my students at Harvard Business School with this question. I find it intriguing that for a term that is so well socialized throughout the entrepreneur ecosystem and so critical to determining when to scale, it has such a varied, non-rigorous definition. Well versed students reference Marc Andreessen’s definition, “being in a good market with a product that can satisfy that market”, but worry the definition leaves too much up to subjective interpretation, especially with regard to the words “good” and “satisfy”. Other astute students reference Sean Ellis’ quantitative approach of “at least 40% percent of surveyed customers indicating they would be "very disappointed" if they no longer had the product”. However, students referencing this approach worry that data gathered in a customer survey may be corrupted with false positive risk.
So, how can we take a more data-driven, scientific approach to product-market-fit?
I find “better-in-class” companies use long term customer retention as an indicator of product-market-fit. The idea is to “let the customer’s wallet do the talking”. The argument is a customer’s decision to renew or repeat purchase is the most factual, true positive indicator of their satisfaction with the product and, in turn, product-market-fit. In aggregate, the tech sector considers an annual customer retention rate above 90% to be the world class benchmark. Therefore, we can argue that companies have product-market-fit when annual customer retention exceeds 90%.
We are getting closer. I agree customer retention is the best statistical representation of product-market-fit. However, customer retention is a lagging indicator. It often takes quarters or even a year for companies to understand the true retention rate of customers that we acquire today. We do not have years or even quarters. Time and money, especially in an early stage setting, are not on our side. We need to test, learn, and iterate in much faster cycles.
For this reason, “best-in-class” startups use a leading indicator of customer retention to quantify product-market-fit. Some entrepreneurs in Silicon Valley refer to the leading indicator as the “ah-hah'' moment. If the leading indicator is objective, rather than subjective, and truly correlates with long term retention then we have defined a data-driven, time-sensitive approach to understanding product-market-fit.
Defining the leading indicator(s) of customer retention
Unfortunately, there is not a single leading indicator of customer retention definition universally applicable to all company contexts. However, the following definition framework is universally optimal.
[Customer Success Leading Indicator] is “True” if P% of customers achieve E event(s) within T time
Documented examples of leading indicators from modern day unicorns, organized in this format, are below.
- Slack: 70% of customers send 2,000+ team messages in the first 30 days
- Dropbox: 85% of customers upload 1 file in 1 folder on 1 device within 1 hour
- HubSpot: 80% of customers use 5 features out of the 25 features in the platform within 60 days
We have deduced the question of product-market-fit to the values of P, E, and T. Below are best practices on defining these variables for our business.
P is the percentage of customers that achieve the leading indicator. If P is surpassed, we have product-market-fit. But what is an acceptable P? Evaluating the extremes, 5% seems way too low. If we acquire customers and only 5% achieve our leading indicator of retention, that will be a terrible foundation for a business. At the same time, 95% seems way too high. The primary reason for this analysis is determining when to scale. Waiting until 95% of customers achieve the leading indicator seems too cautious, exposing us to the risk of waiting too long and missing the market opportunity or losing unnecessary ground to a competitor. A final consideration is the market’s perception of strong annual customer retention, which we previously mentioned is 90%. With all of these considerations, I often see P set at between 60% and 80%. I recommend the lower end of the spectrum if the company sells to small businesses and the higher end if the company sells to large businesses. Because we will instrument and continually monitor the metric on an on-going basis, I don’t believe that a debate on whether P should be 60% or 70% is productive. If we truly have found product-market-fit, we will find that the percentage continually improves even after we have moved to the next phase of scale.
E is the actual event or set of events that represents the leading indicator. Events around product setup, usage, and results are commonly used. E is the most important variable to think through. I recommend the following considerations when defining our leading indicator:
- Objective: The event should be factual and binary. It either happened or it didn’t. There is no subjectivity or room for interpretation. “Processed the first transaction” is objective. “Customer sees value” is not.
- Instrument-able: We need to be able to automate the measurement of the event. Later in the eBook we will demonstrate why it is important to continually measure the leading indicator as the company scales to assess whether product-market-fit is lost. Therefore, it will be important to instrument the measurement of the leading indicator prior to scale. “Logging in at least once per day” is instrumentable. “Mentions of the product in executive meetings” is not instrumentable.
- Aligned with customer success and/or value creation: Intuitively, creating customer value and success will lead to customer retention. Not doing so will lead to churn. Therefore, leading indicator events that represent customer value and success are recommended. “10% reduction in processing time” represents customer value. “Signed the contract” does not.
- Correlated to the company’s unique value proposition: The go-to-market team will be focused on driving leading indicator events in the new customer base. Marketing will be focused on driving awareness with segments where leading indicator achievement is easiest. Sales will be focused on convincing prospects that the leading indicator events are most important. The customer success team will be focusing on-boarding efforts on leading indicator event achievement. If those events are aligned with our unique value proposition, we will amass a customer base that is very sticky to our strategic positioning and very difficult for our competitors to disrupt. The leading indicator example for HubSpot provided earlier is a good example. HubSpot’s strategic positioning was “all-in-one”. Prospects could replicate the HubSpot offering by assembling a number of point solutions to create a broad marketing capability. Using only one feature within HubSpot’s platform was not optimal. There were better point solutions out there. HubSpot’s competitive advantage occurred when customers adopted many features within the HubSpot platform. Therefore, their leading indicator event of “5 or more features adopted” was aligned with their unique value proposition of “all-in-one”.
- Event combinations are OK but keep it simple: As the company expands its product, there may be multiple combinations of events that represent leading indicators of customer retention. These combinations can be “AND” or “OR” definitions. For example, remember Slack’s leading indicator of “2,000 team messages”. Well, 2,000 team messages exchanged between 100 people is likely far more adopted and valuable to the customer than 2,000 team messages between 2 people. Therefore, Slack may evolve their leading indicator to be “2,000 team messages AND 20+ users involved”. They may also find that integration with the CRM represents value and predicted retention. Therefore, they may again evolve the leading indicator to be “(2,000 team messages AND 20+ users involved) OR (2,000 team messages AND integration with CRM)”. As long as the combinations can be evaluated as a binary yes/no, it works. However, keep in mind there comes a cost with this complexity. One of the advantages of the leading indicator is it provides an easy to understand “north star” for the team during the product-market-fit stage. Complex combinations of leading indicators compromises the focus of front line GTM resources.
T is the time by which the leading indicator event is achieved. T should be as short as possible to maximize the pace of learning. However, it needs to be realistic. T often depends on how complicated it is to adopt our product and how long it takes to see value. Dropbox should have a very short T because it takes minutes to download, setup, and see value from the software. Dropbox’s T could arguably be hours. Workday should have a very long T. Workday sells broad, complex HR software into large organizations. It is not uncommon for the setup and user training process to take multiple quarters. Workday could have a T of 6 months or more. On average, T is set between 1 and 3 months for most software companies.
Identify product-market-fit early through customer acquisition cohorts
Once the customer retention leading indicator is defined, we should assemble a cohort chart illustrating the percentage of newly acquired customers that achieve the leading indicator over time. This approach maximizes the speed by which we can evaluate progress toward product-market-fit. Below is an example of a company measuring their leading indicator by monthly customer cohorts.
We can bring this chart to life using a fictitious company, TeleMed. TeleMed sells software to doctors enabling them to meet with patients over video rather than in-person. A well-designed customer retention leading indicator could be:
[Customer Success Leading Indicator] is “True” if 70% of customers conduct a video conference with a patient within 2 months.
Therefore, the chart tells us that the company acquired 24 new customers in January. After 1 month, 3% of those 24 customers had actually conducted a video conference with a patient. After 2 months, 27% of those 24 customers conducted a video conference with a patient. After 3 months, 33% of those 24 customers conducted a video conference with a patient. According to TeleMed’s definition of the customer success leading indicator, they had not achieved product-market-fit in the early part of this year. However, the company executed a number of adjustments, likely changes to the product, target customer, sales process, and on-boarding approach, and the situation has greatly improved. In October they acquired 55 new customers. After 1 month, only 6% of those 55 customers conducted a video conference with a patient. However, after 2 months, 70% conducted a video conference with a patient! The execution paid off. This company has achieved product-market-fit. We do not need to wait for long term retention to surface. This company is ready to proceed to the go-to-market stage.
Here are a few guidelines as we design our customer acquisition cohort analysis.
- In order to align all levels of the organization around product-market-fit pursuit, we recommend
this chart be the first slide in the board deck, ahead of the P&L and top line revenue
- The cohorts can be organized by daily, weekly, monthly, or quarterly time periods. Selecting the
appropriate time metric is similar to defining the “T” factor in the customer retention leading indicator discussed earlier. A company like Dropbox should probably use daily customer acquisition cohorts and evaluate the cohorts’ progress toward the leading indicator on a daily basis. Workday should probably use quarterly customer acquisition cohorts and evaluate the cohorts’ progress toward the leading indicator on a quarterly basis.
- The “Customers Acquired” column are not cumulative numbers. These figures represent new customers acquired in that month.
- It is possible that the product usage within a cohort declines over time. Customers could dedicate their energy early on to using the product, find that the product is not useful, and stop using it. Companies need to instrument the cohort analysis to capture this behavior shift if it occurs.
- The time (T) of achieving the leading indicator is less important than continued improvement within the cohort over time. In the example above, we could argue transitioning to the go-to-market-fit stage in November even though the pure definition of product-market-fit had not been achieved yet. None of the prior cohorts have achieved 70% within 2 months. However, the prior cohorts were showing continued improvement month-over-month with the expectation that they would reach 70% and continue to rise. Furthermore, looking down the columns, recent cohorts at their 2 month and 3 month anniversary were substantially healthier than past cohorts at the same tenure.
Adjusting the approach for large enterprise deals
The above cohort analysis does not work for early stage ventures selling 6-digit deals or higher to large enterprises. These ventures can surpass $1 million in revenue with less than 10 customers and may only acquire 1 or 2 new customers every quarter. Therefore, an alternative approach to evaluating the pursuit of product-market-fit is necessary.
In these situations, companies assemble a customer health card with a half dozen or so criteria. Common criteria fall into the following categories:
- Status on the technical setup and integration of the product
- Number of users that are activated and active
- Breadth of product usage
- Quantifiable value realization
- Executive sign off on reference-ability
The board literally reviews the “green”, “yellow”, “red” summary status for each company as well as the statuses of each of these criteria, especially for new customers and laggard deployments.
The “Why Now” on Prioritizing Customer Retention ahead of Revenue Growth
Decades ago, prior to broad adoption of the Internet, software was sold, deployed, and used in an “on-premise” manner. This meant companies had to purchase, provision, and maintain their own server networks to run the software purchased for their organizations. The process to set up and train the organization on the software was long and expensive, often taking 12 to 18 months and costing millions and sometimes tens of millions of dollars.
In many cases, the software wasn’t very good. It was hard to use. Adoption rates were low. The term “shelfware” was popularized because most software was never adopted and “sat on the shelf”. Why did this happen? It was because adoption didn’t really matter. Sales did. Once the customer was sold, they were stuck with the purchase for the next 5 to 10 years at least. In this context, the best sales team won.
Fast forward to 2020. The internet has changed two things. First, it no longer takes months to deploy and use software. Cloud, SaaS, and the broader subscription economy have significantly reduced the friction to adopt software. These trends have also reduced the friction to stop using software. Second, every customer has a huge megaphone called social media to tell the market about good and bad product experiences. For both of these reasons, companies are starting to realize the long term health of their business is more dependent on customer retention than customer acquisition. However, the continued premature focus on top line revenue growth is misaligned with these trends. Go-to-market design and execution is, in a way, operating in a by-gone era.
Aligning go-to-market execution with customer retention
With a more scientific definition of product-market-fit in place, the company has a precise “north star” to focus on. Most companies associate customer retention issues with deficiencies in product or customer on-boarding. However, I find a much different diagnosis. Most customer retention issues originate in sales and marketing. Customer retention is driven by the types of customers targeted by marketing and the expectations set during the sales process. Remember, the odds are against us at this early stage. Only 20% of Series A funded businesses will succeed. Best-in-class companies at this stage align all aspects of the go-to-market with the “north star” of the customer retention leading indicator. The chart below summarizes how.
The first three components, target market, GTM playbook, and Sales Hire, are the most critical decisions at the Product-Market Fit stage. The buyers we choose to sell to as well as how we sell and on-board them will be the most important drivers of the customer retention leading indicator. A unique salesperson profile is needed to execute this early playbook. Scalable demand generation, pricing, and sales compensation are not important at this stage. If we are developing scalable cold calling campaigns, launching a tiered pricing model, or designing a robust sales compensation plan at this stage, we are not focused on the right things.
Target Market: Stack the deck with early adopters from smaller companies
Who should we target as our first customers? This question often leads to a debate between large customers that yield powerful reference-ability versus small customers that enable rapid learning. On one hand, we should pursue a “big-brand” customer. If we can acquire and make the customer successful, it sends a powerful signal to the market that if this new product was good enough for the big brand, it must be good enough for everyone else. On the other hand, we should pursue the smallest customers within their target market. Small customers are easier to connect with, make decisions faster, and have simpler product adoption requirements than larger companies. Therefore, pursuing smaller organizations provides the fastest path to learning.
As entrepreneurs, we pursue the “big-brand” customer most often. However, the choice is not optimal. We under-estimate the difficulty of setting a meeting, the high bar of IT and security requirements, and the “red tape” interfering with product adoption even after purchase. We should err toward the smaller customers to foster rapid learning. We should reflect on how small we can go within our target market definition where our product still creates value and start there.
The other consideration is the optimal person within the target customer. We need “early adopters” not “laggard followers”. We are still learning and refining our product and business. We need early customers who are excited to innovate with us. Often, an “early adopter” is more about the individual buyer within the organization than the organization itself. These buyers view themselves as first movers. They enjoy playing with new products and don’t mind that there are bugs. They are excited to send us lots of feedback and ideas. They enjoy being part of the innovation process. Early adopters care less about customer references or robust ROI studies. Save case study and ROI driven customers for the scale up stage. We are not ready for them.
Go-to-Market Playbook: Win at all cost
There are two themes for the GTM playbook at this stage, “win” and “do things that don’t scale”. “Win” is the customer retention leading indicator, not a signed contract or payment.
“Do things that don’t scale” is advice from Y Combinator founder, Paul Graham, and should be kept at the forefront of our minds at this stage. I remember chatting with David Cancel, CEO of Drift, at this stage of his business. He was literally flying out to have one-on-one on-boarding meetings with customers that were paying him $50 per month, as the CEO! “Do things that don’t scale”. Throw everything and the kitchen sink at achieving our early indicator of customer retention. One-on-one, “white glove” on-boarding processes, even for low value customers, are good. Mass on-boarding sequences are not optimal at this stage.
The one component of the long-term GTM playbook worth codifying at this stage is the Customer Retention Qualifying Matrix. Qualifying matrices like BANT and MEDDIC are commonly used in sales to understand the likelihood that a customer will buy. However, they do not help us understand whether the customer will succeed with the product and ultimately remain as a customer. Common components of the Customer Retention Qualifying Matrix include whether IT is aware of implementation tasks, the end user(s) are part of the purchasing process, not just the decision maker(s), the customer’s tech stack is compatible with the product, etc. As a seller, we can get a signed contract without having these items in place. In fact, accomplishing these tasks may actually slow the deal down. However, not completing these tasks before the purchase will likely put successful product adoption at risk. We are solving for customer retention, not signed contracts.
Sales and Customer Retention Qualifying Matrices
Sales Hire: Half Product Manager, Half Account Executive
Getting this hire wrong is a top 10 reason for Series A funded business failure. The first pothole most organizations fall into is pre-maturely hiring a sales leader. I can hear the Series A investor now: “Go find someone that has scaled a business to $100 million”. The hire is a complete mis-alignment for the tasks required at this stage. The second pothole is hiring an account executive from the large incumbent in our space. Yes, this account executive is successful selling to our buyer however, when they joined the incumbent, they went through weeks of sales training, were provided a sales playbook, and were scaffolded with an experienced sales manager. This “process execution” salesperson will not succeed in our environment where these assets and infrastructure do not exist.
At this stage, look for a mix between an account executive and a product manager. The first sales hire should have the skills to handle objections and comfort discussing money like an account executive. However, they should also have the ability to pattern recognize feedback from the target market and communicate the patterns to engineers. Focus on these attributes when evaluating candidates:
- Comfortable in ambiguous, rapidly changing environments. Self starter.
- Motivated more by innovating than making money. Avoid the salesperson primarily motivated
by money at this stage.
- Ability to dive into customer needs through deep discovery skills and identify patterns
- Strong collaboration skills to work in cross-functional teams, primarily with product and engineering
Demand Generation: Rely on personal network and referrals
“Do things that don’t scale.” We do not need 1,000 customers to achieve product-market-fit. We have a lot to do and adding the development of a scalable demand generation capability, such as an SDR cold calling team or a content marketing function, is mis-aligned with our phase of development. Rely on personal networks of founders, employees, and investors as well as customer referrals. These channels do not scale. However, they yield the highest quality opportunities to learn from.
Pricing: Minimize friction, maximize customer commitment
“Do things that don’t scale.” Our goal at this stage is maximum learning, not profits. Unless the biggest uncertainty in our business model is the product price, which is rarely the case, don’t spend a lot of time optimizing the price at this stage. Keep the price as low as possible to reduce friction. However, “free” is not effective as the customer will not be committed. Charge enough so the customer is committed to achieving product adoption and success. Be explicit with customers that our business model calls for a price of say $30,000 per year but we plan to sign up the first 20 customers at a 90% discount. Don’t be afraid to “grandfather” these customers into these discounts for some time. This approach will naturally attract the early adopters we need at this stage. We don’t need to raise a bunch of venture capital to afford to do this. The cost is the long hours of our small team, not an army of staff. This approach is simply a continuation of the market research and lean startup phases we have recently executed.
Sales Compensation: Simple and aligned with the customer retention leading indicator
If we hire the right salesperson profile at this stage, the design of the sales compensation plan will have minimal impact on this phase. We may even consider no sales compensation plan, using a base salary and equity just like everyone else on the team. There is no reason why the salespeople should be the only employees who suffer financially if it takes longer than expected to navigate through the product-market fit stage. Furthermore, a traditional sales compensation plan designed around new revenue acquisition mis-aligns the salesperson from the company objective of rapid learning and customer value creation.
If we did use a sales compensation plan at this stage, avoid making it too leveraged. Consider 80% base salary and 20% variable compensation. Also, align the compensation with the leading indicator of customer retention. Pay when the leading indicator is achieved, not when the contract is signed or the payment is made. Remember, the “win” here is the achievement of the customer retention leading indicator.
Verify the leading indicator of customer retention
As the company and customer base develops, we need to verify whether the leading indicator actually correlates with customer retention. In most cases, companies have moved on to the next phases before it is possible to verify this correlation. That is fine. However, it is important to conduct the analysis and continually conduct it in order to understand that the foundation is strong.
Below is an example verification for TeleMed, our fictitious doctor video company. As a reminder, TeleMed used:
[Customer Success Leading Indicator] is “True” if 70% of customers conduct a video conference with a patient within 2 months.
Correlate leading indicator of customer retentionAnalysis of customers (churned and active) acquired between 12 and 18 months ago
In the above example, the company had acquired 68 customers between 12 and 18 months ago. Of the 68 customers, 56 are still customers for an overall retention rate of 82%. Of the 68 customers, 55 had achieved the customer retention leading indicator. In other words, 55 of the 68 customers conducted a video conference with a patient within the first 2 months. Of those 55 customers, 51 are still customers for a retention rate of 93%. Similarly, 13 of the 68 customers did not achieve the customer retention leading indicator. In other words, 13 of the 68 customers did not conduct a video conference with a patient within the first 2 months. Of those 13 customers, 5 are still customers for a retention rate of 39%. In this case, the leading indicator seems to predict long term retention well.
Correlate leading indicator of customer retentionAnalysis of customers (churned and active) acquired between 12 and 18 months ago
The above example is similar to the prior one. The company had acquired 68 customers between 12 and 18 months ago. The overall retention rate was 82%. Of the 68 customers, 55 had achieved the customer retention leading indicator and 13 did not. However, in this case, only 84% of the customers that achieved the leading indicator are still customers and 77% of the customers that did not achieve the indicator are still customers. The leading indicator does not seem to predict long term retention well.
Recommendations for the verification analysis include:
- Limit the analysis to customers acquired between 12 and 18 months ago. Customers acquired
before 12 months may have not had an opportunity to churn, especially if annual contracts are in place. Customers acquired more than 18 months ago are less representative of the current
state of the go-to-market operations.
- Conduct the analysis quarterly, as the correlation may change as the market and product evolves.
- The verification analysis is not a prerequisite to moving to the go-to-market fit stage. Use the leading indicator to determine stage graduation.
- If we have historical customer data, analyze the correlation between our leading indicator hypothesis and long term retention now.
- Don’t worry if we didn't get the indicator correct. Focusing on these events, like setup or usage, probably didn’t hurt the business. Run other theories to see what events are actually correlating better and re-align the business with these events.
What We Learned
- We use “product-market-fit” to make critical decisions such as when to scale. However, we lack a scientific, data-driven definition of the term, creating timing mistakes on our decision to scale.
- Customer retention is the best statistical representation of product-market-fit. However, customer retention is a lagging indicator. We need to define the customer retention leading indicator.
- Assuming long term customer retention is the best statistical representation of product-market-fit then:
- Organizing our customers into acquisition cohorts and measuring their progress toward the customer retention early indicator enables early identification of product-market-fit.