Stripe’s opportunity: Reinventing Customer Service

I am a big fan of Stripe, and I was recently thinking that there’s a non-obvious opportunity for them to fix one of the most annoying thing ever: customer service. I tried to imagine what an internal memo on the subject should look like. Wrote it. And decided to share it with you. Hope you find it interesting; most “corporate strategy” activity like this one is often kept secret, and there aren’t enough opportunities to discuss it openly. This is my humble contribution to this needed discussion. Read on.

Old photo of mine from 2015, just because a wall of text without any picture scares people too much.

Customer Service is currently an inefficient business function that damages the GDP of the Internet. Stripe can reinvent it for its clients as well as for Stripe itself. Let’s look at this in the context of a general framework to identify and evaluate new opportunities for a company, being it Stripe or a different one.

Customer Service is currently an inefficient business function that damages the GDP of the Internet

The right framework: reinvent industry segments with non-decreasing marginal cost curve

Nowadays human beings evolve at a much slower pace than information systems. Individuals, small businesses and large companies can improve their margins and productivity mostly by enhancing people’s ability to execute decisions; the use of APIs within and between organizations is one of the best examples of this principle at work (despite unsuccessful initiatives such as automatic tax calculation — possibly too complex and costly to maintain). When it’s correctly applied, the marginal cost curve decreases, or in other words, there’s an Increasing Return to Scale (IRS).

The right framework for Stripe should identify industry segments or services in which:

1. The marginal cost curve doesn’t decrease as it could/should (also called Constant or Decreasing Returns to Scale (CRS or DRS); further reading: Producer Theory).
2. Stripe’s clients are currently suffering from complexity and friction, but are unwilling/unable to solve the problem themselves.
3. Stripe can provide a solution leveraging a unique advantage, enabling an Increasing Return to Scale (IRS), leveraging technology correctly to increase productivity and margins.
4. Doing so increases the stickiness of Stripe’s core products (e.g. Payment).

To summarize even more succinctly:

1. Decreasing Returns to Scale;
2. Unsolved Complexity and Friction;
3. Unique Advantage;
4. Stickiness.

An example: Reinventing Customer Support

Customer service has traditionally been considered stronger in “brick and mortar” than in online sales. Better online service could move more sales to the internet (Stripe’s mission).

Customer service is primarily the consequence of product or service failure; and yet, it is mostly treated as an annoyance by the company providing it even when it already has the information necessary to provide you better service (e.g. FedEx knows 10 hours beforehand that your package will be late, but it doesn’t tell you). It seems that the most common metric measured in CS is how quickly you can send a customer away, irrespective of whether you have solved the issue and strengthen your brand.

Customer Support or Customer Service (CS) is an area where we observe all four characteristics of our framework:

1. Decreasing Returns to Scale. AI and Chatbots are mitigating the problem (up to 30% cost reduction), but this is not a high-margin business (source), and there’s no fast-growing company reinventing Customer Service in sight.
2. Unsolved Complexity and Friction. Not just Stripe’s clients, but Stripe itself has issues with Customer Service (“I’m very happy with Stripe’s API; I’m very unhappy with their support.”, source; “Stripe support is terrible”, source). Recent innovations (e.g. GPT-3, automation) suggest that Customer Service might be ready for disruption, but it is complex enough that customers are waiting for someone to come and solve it on their behalf.
3. Unique Advantage. Stripe could, as an example:

• Load up customer data (from the same merchant and others) when a user reaches out.
• Monitor high-discretion customer service reps for abuse of discretion (e.g. illicit refunds to friends), enabling companies to give reps the discretion that makes customers happy, at less cost .
• Aggregate/organize feedback from merchant’s Customer Support representatives (they are the best source of data on where pain points are, says John Goodman of CCMC, not customer surveys, which seldom work).
• Consider an on-demand, pay-as-you-go, customer service, in which you are refunded if it’s 100% the company’s fault?

4. Stickiness. We can expect a successful customer service provided to merchants to be deeply intertwined with the merchant’s workflow and automation processes, providing both a high barrier of entry and an incentive to stick with Stripe’s payment service.

More on Stripe’s unique advantage

For most businesses, Customer Support seems to generally be inefficient and ineffective. Every new customer support request starts with almost no information about the customer (with exceptions like credit ratings), and no way to improve the interaction besides a generic script (or a stupid chatbot).

A merchant might have unstructured details on past interactions which might help categorize/classify the customer (e.g. high spender; PITA; etc.), but doesn’t programmatically organize this kind of information to maximize its value.

Stripe oversees many more transactions for that same customer with a multitude of merchants: a unique opportunity to better categorize/classify him/her and provide an outstanding customer experience.

As an example, a customer who pays for an AWS account and a Sublime Text subscription is most likely a computer literate and/or a programmer; when troubleshooting a service issue he’s having with Wix (a client of Stripe), we can assume he’s more skilled than most and offer a more advanced dialogue to solve the issue (e.g. avoid “Did you try restarting your computer?” — I hate it so much!). Or, if he’s a subscriber to Because (personal hygiene products for older adults) and GoGoGrandparent, it tells us something about his age (probably 65+), and the type of customer support that would work best for this age group.

Stripe should also be able to provide tools for the merchant to use this and other insights.

More on why customer service hasn’t been reinvented elsewhere

Large merchants have to build their own customer support technology and hire a support team. This is a lot of work just to achieve a bare minimum, unappreciated by the customer. Amazon.com categorizes these things as “muck”, or “undifferentiated heavy lifting”: complex and expensive and hard to scale, but not differentiating enough for most companies.

Setting up computing infrastructure was the “muck” that many companies had to go through, and which justified the creation of Amazon Web Services, now an impressive ~$46B/year business (source) after just 14 years of operation. (I was there in the early days! Here’s how I got hired). Setting up service payment and billing was the muck Stripe was founded on.

Customer Service is the next frontier of muck. If Stripe could make it as much of a utility as servers and billing, it would create enormous value for both users and merchants.

Evaluating the Business Opportunity

It fits the framework, but should it be given the green light? We could examine how much time and money is spent by merchants on customer service; how difficult it is to scale; how much it costs them not to provide a perfect customer service; and which areas of improvement are easier to solve first. With this information in our hands, we can determine whether it’s a large enough business opportunity and an estimate of the kind of budget needed to build it.

What Stripe needs to do in 2021

Here’s how to bring this idea to life. Let’s call it Augmented Support if we want to be “corporate” and boring. Alternative name: Winston, from Pulp Fiction’s Mr. (Winston) Wolf (“I solve problems”).

1. Interview merchants (mostly their customer support reps) to deeply understand where the main pain points in customer service are, and which type/category of merchants has the highest friction and highest willingness to try something new. Should we expect low value transactions to be a bad candidate to start with?
2. Determine the best way to collect end users data from merchants, while maintaining privacy. Alternatively, just start with Stripe’s own customer support.
3. Explore the customer support landscape, and determine which companies could potentially be acquired to jumpstart our efforts (Intercom, Zapier, Embrace, Fixia, etc.). We should in any case actively track these companies and consider them a potential competitor in the future.
4. When enough data is collected, propose a specific product roadmap, timeline and budget, and understand how customers would use Stripe’s APIs to consume the service.
5. Roll out a private beta program, reassess after 3 months.

Conclusions

So, what do you think? Is this something that Stripe should build / fix? Do you like this framework when thinking about new opportunities for established companies?

Happy to hear your thoughts.