3 Ways OpenAI Can Drive Subscription

Intro 

OpenAI needs no introduction. They’re in the news every day and have gone from $0 to $2 billion in ARR in 3 years. 

So, this article is very much not criticizing their design but instead looking at ways they could drive greater subscription conversions for their premium product and ramp up their LTVs. Hope you enjoy, let’s dive in! 

OpenAI’s Homepage 

If you are one of few people in the technology industry that hasn’t tried out ChatGPT this is ChatGPT’s sign up screen. 

Let’s run through the positives. It’s super simple, providing LogIn and Sign Up options as well as a handy looping animation that gives new and returning users some inspiration for prompts to help better utilize the product. Not many SaaS Companies utilize animations on the log in page but it’s certainly more effective at grabbing your attention. 

However, a few tests could be run here: 

1. Make a loginless experience

A few weeks ago we explored the rise of loginless experiences, where SaaS Companies like Rows can allow you to try out the functionality of the product directly on the homepage without needing to log in or create an account. 

This can significantly reduce the barriers to adoption and gets users understanding the value prop of a product before asking for the commitment of creating an account which can lead to a far better sign up and activation rates. 

It would also make a lot of sense for ChatGPT to adopt this given competitors like Google grew so fast off the back of being able to simply type a query into a search box without needing an account. 

OpenAI is actually planning to do this, removing the final barrier to entry allowing users to try ChatGPT’s free version without creating an account but doesn’t promote this on the web and the rollout is going to be gradual. The aim behind this is to get a new cohort of users who weren’t committed enough to hand over personal information and go through creating an account to potentially try their product and get hooked. 

2. Multi-step Forms 

The next step to improve their homepage conversion rates is exploring the introduction of multi-step forms. By asking for a singular piece of information upfront you reduce customer anxiety about all the information that may be required and just give them one simple step. 

To do this most effectively you want to ask for the lowest commitment piece of information (e.g. email) and hold off on asking for the higher commitment information until the end of the signup form. 

The reason this is such a powerful trick is that it builds up sunk costs before a user gets to these higher commitment pieces of information like a password or phone number and makes users think that if they’ve made it this far, they may as well continue. As a result of multi-step forms Hubspot has reported it increased completion rates 86% while some report improvements as much as 300%, up to 53% even on lengthy forms with salary information and phone numbers!

To optimize this to an even greater degree you can implement these practices: 

a) Use image selector buttons to reduce typing

This one feature solves the largest conversion killer and reduces the source of friction in having to type, allowing users to click large, tappable buttons, meaning it is also perfectly optimized for mobile. In contrast, every text field increases the likelihood of a user typing an error, which either leads to failed submissions or incorrect data. 

b) Show users their progress: The ‘endowment effect’ shows the bias around the fact we’re far more likely to complete an action if we see the illusion of progress. 

Now this can be a trade-off. OpenAI has a super simple login process with very few steps, for most users it takes less than 10 seconds as they simply login with Google. 

However, whilst adding additional steps to the sign up form may hurt overall conversions (not necessarily), additional information gathered through multi-step forms (e.g. why you’re using ChatGPT) can actually lead to a far greater number of people hitting the ‘a ha’ moment and helping with user activation. 

c) Ask 1-2 questions per step: This way the form appears less overwhelming and feels faster to progress through. 

What could this look like for OpenAI?

3. Reduced Commitment CTAs 

The final step to improve their homepage conversion rates is exploring lower commitment Calls to Action (CTA). 

Instead simply asking users to Sign Up/Log In OpenAI could assuage some web visitors fears by highlighting the fact that it is an always free product or that they’re just continuing to the next page to find out more information, not committing to too much. 

This trick is used a LOT in SaaS and mobile games. Airbnb publicly shared that by changing one simple CTA from ‘Reserve’ to ‘Continue’ better informed users that they weren’t yet making a financial commitment and significantly boosted clickthrough rates. Overall it led to several million in new incremental revenue for Airbnb over the next month. 

Similarly, instead of choosing the word ‘Buy’ or ‘Purchase’ mobile games like Pokemon Unite choose a slightly odd word, ‘Obtain’. The reason for this is that words like ‘Purchase, Buy’ remind you of the fact you’re spending your hard earned money. ‘Obtain’ not only dulls this sense but goes even further, instead reinforcing the thing of value you’re receiving, vs the money you’re parting ways with.

Onboarding & Getting Users to Upgrade 

As a result of all this success ChatGPT now has north of 180 million registered users with more than 2 million user subscribed to ChatGPT Plus, OpenAI’s $20/month premium tier that gets you access to features such as GPT-4, the GPT Bot Store and Code Interpreter. 

This comes to about $500 million in ARR and a 1.1% conversion rate to their premium tier. Whilst this is impressive for a company that hasn’t put much effort into growing their consumer subscription revenue other freemium tools boast significantly higher conversion rates, with good ranging from 3-5% and 6-8% being great, such as Grammarly, Canva and Trello. 

1. Leverage anchoring

A great lesson to copy from the world of mobile games is how they drive users to spend on their product for the first time by leveraging the principle of anchoring. 

99% of people go into free mobile games under the premise that they won’t spend money, and whilst the statistics are less dramatic in SaaS the general rule applies - most people will say they aren’t planning to pay for the premium version of a free tool. 

As such developers need to break that first barrier so users are financially and emotionally committed to the game, and getting them thinking of themselves as “spenders”. 

To do this mobile games will typically throw up a paywall upon immediately opening the game for the first time. You haven’t played a single minute yet and so the intention isn’t to actually get most people converting here. Instead, the idea is to offer you a deal at a high price point (e.g. $10) which artificially anchors you higher. 

Anchoring, the principle used here, is essentially the idea that as humans we have a tough time accurately quantifying the value of something, especially if it’s software, which means the first price we hear is our anchor, of which we’ll compare everything going forward against. 

This then allows game developers to present you an offer shortly after which seems like a way better deal. 

In Pokemon Unite’s example after the very first game (just 6 minutes in) they’ll show you another bundle offer, but at a 90% discount for the tiny price of $1 so users feel it’s too good to pass up and is at such a small price they may as well give it a go. 

From here the wall is broken and it becomes much easier to justify spending more money because if you’ve spent $1 you may as well spend $5, and if you spend $5 why not $10. 

After throwing up a paywall promoting GPT-4 in a user’s sign up, why not let them use it for 5 minutes before offering them a one-time offer of a $1 trial of GPT-4 to get users hooked on the experience before it renews at $20/month. 

They could further improve take-up rate by leveraging the reduce commitment CTA trick, replacing ‘Upgrade’ with ‘Access GPT-4’, ‘Try for $1’ or ‘Continue’, suggesting by pressing this button they aren’t committing to paying $1, but just proceeding to find out more information. 

2. Ikea/Customisation Effect 

The Ikea effect is the principle that we value the things we put work into more highly. Whilst Ikea certainly may not use the best materials and joinings we develop a sense of attachment to the furniture because when we see it we vividly remember the work it took to construct it. 

Mobile games heavily utilize this, making you form a deeper connection with the game by enabling you to customize your character through skins and items. 

OpenAI are beginning to apply this to their models through the form of memory. For select users you can now add pieces of information within ‘Settings’ for the model to consider with every answer it gives but is working to better use your chat dialogue to develop a greater sense of who you are. 

This customisation and memory is a publicly stated priority of Sam Altman’s. Instead of having to provide every piece of information to large context window models like Claude & Gemini, the idea is that OpenAI can start building a knowledge graph around individuals which can help provide more tailored responses and reduce the need for prompt engineering. 

This is an opportunity to make answers better with usage forming meaningful differentiation vs the rest of the foundational models which are quickly becoming commodities. 

However, memory not only poses a great opportunity for making the product stickier but also could be a key driver of ChatGPT Plus subscriptions. 

We’ve talked about the idea of embedded upsells, where you click a button to perform a certain action which turns out to be a premium feature, thus throwing up a contextual paywall, which could be effortlessly combined with OpenAI’s memory. 

Imagine if OpenAI can identify queries that should be part of your memory and ask if the user would like to have this information included. If they click yes, there’s a perfect opportunity to show a paywall explaining that memory is feature gated and only available to ChatGPT Plus members. 

It’s the worst kept secret in all of tech that revenue and conversions increase almost linearly with the number of paywalls shown. This would allow OpenAI to repeatedly show contextual, well-timed paywall views to their $20/month subscription. 

3. Reciprocity Principle

However, this wouldn’t be enough to make everyone convert. Many haven’t had the opportunity to try out GPT-4 (& their other models & premium features) and so don’t understand how much better they are vs the standard free product. 

This is where the reciprocity principle comes in. Humans are reciprocal creatures by nature and so if a person or company is able to give a meaningful gift the recipient feels the need to return that generosity. 

Mobile games and apps like Duolingo heavily utilize this where users will be gifted free digital currency and cosmetics but for the feeling of reciprocity to be triggered you need to hit 3 key criteria where the gift is: (1) unexpected, (2) personal, and (3) significant. 

This already exists in SaaS in some sense with the advent of reverse trials, where a user will be automatically bumped up into the full premium experience, then downgraded when the time runs out.

The idea behind this is to make users get a sense of what they’re missing and leverage loss aversion. 

For OpenAI a 30-day reverse trial for weekly active users who have not yet converted could be a powerful lever and tick all the requirements of the reciprocity principle. Ensure there’s a great paywall or onboarding flow to help demonstrate all the features of ChatGPT Plus and conversions should jump. 

That’s it from me today folks, hope you enjoyed these ideas and that it can spark an idea or two for your own business. 

HB