Transitioning from an enterprise to self-serve model isn’t easy - but it is possible

Gokul Rajaram wrote one of the most important pieces for B2B software companies: Self-serve first: the overlooked but essential paradigm underlying great software companies.

His piece convinced me that self-serve and bottoms-up SaaS is a much more scalable model.

But what do you do when you are already running a successful ($10M - $100M ARR) SaaS business selling to enterprises?  How do you create self-serve as a second motion?

We are working through this now at SafeGraph and I thought it might be helpful to others thinking about the same transition.

Even a little self-serve can help

To date, 99% of SafeGraph’s revenue comes from a salesperson-model where we contract recurring revenue with a customer - usually after a sales process where the customer talks multiple times to a salesperson, does a data evaluation, and more.  Our average ACVs are over $120k so we tend to have a small number of very large customers, though we only have one customer that is over $1 million ACV.  

About 1% of revenue comes from self-serve, and that revenue is non-recurring.

Even though self-serve has historically been a very small fraction of revenue, it has already been a very powerful sales enablement tool.  Self-serve has been part of the customer journey in most of our recent sales and it helps with data evaluations and moves deals faster.  It also has been helpful in lead generation.  

SafeGraph’s revenue will ultimately be driven by people. You need great salespeople and customer success stars to sign big contracts. But we hope the majority of our leads will come through self-serve.

Transitioning more toward self-serve

In 2021, we are making a big investment in self-serve.  We are building out features of our data download store and we will be launching a full self-serve API next month.  

We’re going to grow our revenue mix from 1% of net-new ARR coming from self-serve to over 10%.  And in future years, self-serve will be an even greater piece of the pie.

In 2022, we could conceivably see the vast majority of new customers coming through self-serve. Then our salespeople and customer success team will work with these smaller self-serve customers to turn them into enterprise accounts.

The four self-serve drivers: selection, quality, price, UX

For SafeGraph to really build a great self-serve product, we want to offer a great selection of high-quality data at the best price - and the buying experience should be really easy and really fast.  

As a true north, just think of what Amazon or Alibaba would do when selling goods.  

I’ll dive into selection in a future blog post but want to focus this post on making the case for quality, price, and UX - especially since price and UX are different values from most enterprise-heavy SaaS companies.

Clearly defining quality is really important in data

SafeGraph’s main customers are data scientists and they care DEEPLY about quality.  We need to stand behind every data element we sell.

SafeGraph data has a certain level of quality expectation.  The biggest reason for SafeGraph’s success, to date, is because our data is extremely high quality.  Our brand is associated with quality, and we need to continually level-up.  When we see data of questionable quality, we need to immediately correct it (or at least remove it).  

As we move to be increasingly self-serve, we cannot wait for the next release, which might be a month later, to correct our mistakes.  We want to ensure that bad data gets corrected or removed.  The faster we correct the data, the more feedback we will get that the data is wrong because users will see the data being changed quickly.  

To boost quality, we will also need to invest in better systems to get feedback.  That can come from our enterprise customers’ customers, from the data scientists at our enterprise customers, from our self-serve customers, or even from SafeGraph employees. We need to build systems where users can very easily flag something being wrong to then kick off a process to evaluate the claim, and ultimately give the person who logged the bug feedback on what was done.

Everyday low prices 💵

We want to have the lowest prices that one can find for a particular type of data.  This is important because we want to have a long-term relationship with our customers.  We don’t want to price-gouge them.  We don’t want pricing to be complex.  It should be easy to figure out.  

‍

Talking to a salesperson should not give you a better deal.  Of course, you almost certainly will get a better deal talking with a salesperson in the short term.  But in the future, ideally, pricing is standardized with clearly defined ways of how customers can get discounted pricing by increasing their spending.  Opaque pricing is the enemy of self-serve.

Most self-serve companies evolve their pricing overtime to pay by credits (think of Segment, etc.).  Credits can be bought for a discount up-front in bulk and spent over a period of time.  So it might always cost 10 credits for a row of geometry data, but some people might pay one cent per credit and others might pay 0.5 cents per credit if they are getting them in bulk.  

User experience is really important

For enterprise customers, the user experience is a nice feature but not necessary.  For self-serve, UX is core.

This is a problem for SafeGraph as we currently don’t have a lot of people that think about UX.  Our engineering team is mostly built of super-strong back-end engineers.  Our product team is very back-end heavy.  Even our marketing team is mostly made up of enterprise-focused people.  And, of course, our CEO (me) has historically only built businesses that sell in a more enterprise motion (rather than the bottoms-up self-serve motion).  So we need to train ourselves to really care about UX, and also bring in new talent who have a history of thinking about it.

The first rule of UX is speed.  Speed can trump a lot of things.  When you think of some of the more successful UX-driven products like Superhuman, speed is at their core.  Luckily, speed is something that back-end heavy talent can focus on (think of API response and how fast the search is, etc.).  It is an easy metric to measure and optimize.

Speed is something that has a compounding effect.  Speed also REALLY helps SEO. It is one of the most important factors, and SEO can be a huge driver of bottoms-up marketing.

As an aside, the company that does SEO landing pages best is Zapier (I also am a huge admirer of Truework’s SEO).  Zapier is one of my favorite companies (and no, I am not an investor … unfortunately … but, full disclosure, I recently became an investor in Truework).  Zapier owns landing pages about apps and also landing pages from “app 1 to app 2” -- like “Greenhouse to Hubspot.”

Today (March 2021) SafeGraph’s self-serve experience is very much on the slow end.  Searches take many seconds.  Getting data takes many minutes.  Pages load slowly and cannot be SEO’d. That is forgivable if self-serve is mainly used as a sales tool (which has been self-serve’s main use for the last year) but will hurt us significantly as we work to build out self-serve.  Our API, which we will release in April, also needs to be screamingly fast.

Another UX feature is search.  First, search needs to be really fast (see above).  The key is getting customers to find the most relevant thing really quickly.  

As we grow our data from hundreds of attributes (today) to thousands of attributes (future), one of the hardest things will be to help our customers find the right attributes for them.  Some of that will be bottoms-up, but much of it needs to be within the product where we will be surfacing columns that are right for the customer.  

The documentation needs to be excellent - written very clearly for the average person to understand.  Documentation needs to be really well organized so that people can find things.  There needs to be an easy search within the documentation.  And there needs to be quick live help (both from SafeGraphers and from a broader community).  

Another UX thing we need to obsess about is all the very little things.  We need to be constantly watching how our customers interact with the product (tools like Hotjar can be really helpful).  All transaction emails need to be written to be as helpful as possible to the customer.

We need to solve for the littlest things quickly.  We need to obsess about all the little bugs and really test everything, like every browser and mobile device.  We cannot settle.  It is ok to get things out fast, but we need to make huge efforts to continually improve them.

But that’s not all

In addition to the four self-serve drivers (selection, quality, price, UX) there are a few other things we should highlight.

Standard terms - no legal contracts

One of the beautiful things about self-serve is that there is a standard agreement that everyone agrees to, and can be updated, on occasion.  Agreements are always on your paper.  That makes everything much easier to standardize and much easier to make changes quickly.  

If a customer wants to change the terms, then they need to talk to a sales representative or customer success team member.

One thing you forget when you are an enterprise software company is how many different agreements and terms you have. Every agreement above $100k is customized.  Standardized terms simplify things and make self-serve scalable.‍

Self-serve should be marketing-driven (all things equal)

Much of the success or failure of self-serve is going to come from marketing.  Marketing needs to be engaged in every aspect of self-serve, especially the product, pricing, and documentation. Recorded demos and webinars led by marketing can go a long way in converting self-serve customers, without having to involve a salesperson.

Marketing’s job is not just to get users to come to our self-serve landing site, it is full-stack. A successful self-serve marketing team will craft the customer journey from start to finish, and then nurture customers until they are ready to become enterprise customers. In this regard, if UX is the salesperson of self-serve, marketing is the product manager.

Lessons learned from enterprise models should be used to build a self-serve model

We’ve learned a lot from our enterprise customers - about selection, quality, pricing, UX, and everything in between - that’s helping us define our self-serve model. We’re still figuring it out, but we have big aspirations and we’re excited about this next chapter.

We’re always looking for feedback so let us know what you’d like to see in the self-serve data buying process.