The site is for folks that are ok with a probability of success. There are folks that want better guarantees, probably annuities. Wade Pfau, who has been researching retirement planning for 15+ years has a good description of the issues in his Retirement Planning Guidebook  which talks about his RISA model. Looks like he's now productized it at .
I looked at the asset allocation recommended over time, and I think that should be adjustable based on the users level of risk.
I like your guide concept, but the content feels pretty generic. I think any product in this space needs to have a big emphasis on educating the user about the issues. I would like to see this integrated into the calculator better. Maybe a report that lists the issues that the product handles, then customizes the list for this particular financial plan showing if and when in the plan the issue applies and gives further resources to learn about the issue. Order it by when, so the customer has a financial issue learning plan.
In the education theme, the product is making lots of unstated assumptions now, and it makes it hard to interpret the output. As my teachers (and probably your teachers as well) said, ALWAYS state your assumptions.
Another way you can make your product more usable is to figure out what the top questions your customers want to answer and then build guides to show them how your product answers that question. I see you've got a part time position open to figure out customer needs, this could make it easier to close a sale. It may be possible to go through the forums at various groups like Mr. Money Mustache, Bogleheads, etc to find common questions.
Speaking of education, the algorithm needs to be well documented. Otherwise it could be based on a random number generator for all I know. Is it based on historical market data? Or is it Monte-Carlo based? If it's MC, then what data is generated and how? Is the data different on each run, or is it the same MC data used for every run for all folks (essentially treated the same as generated historical data.) Jim Otar has some good comments about MC in his book "Unveiling the Retirement Myth" . He's not a fan of most Monte Carlo data generators, but gives a set of tests that any good MC generator should pass. He's a fan of using historical data. Karsten (a retired economist) at Early Retirement Now  has done quite a bit of work on Safe Withdrawal Rates using historical data above and beyond what Began has done.
Have you used Financial Engines at all? It was around a dozen years or so ago much like what you are building that William Sharpe was involved in. They tried to partner with 401k providers to offer it as an upsell to large companies. I have an old 401k that still has it. They started as a tool for the DIY, but pivoted to managing your 401k for a percentage of assets.
A recommendation for you is to make the initial onboarding a big more clear as to what type of account and whose account it is for those that are married. Getting quality data here will help for the features I am going to suggest.
Some feature suggestions for you to think about that could be competitive differentiators:
1. Help estimate life insurance needs for folks. There are a number of approaches, income replacement and expense liability matching are a couple common ones. This is a feature that your competitors don't have and fits with the demographic you are targeting. There is a lot of opportunity for education in this area. There is the opportunity for referral income by referring folks that truly understand their insurance needs and are ready to buy.
2. Different forum communities have standardized formats for background data for asking questions. Make it easy to get your data in that format for asking a question. This would mean engaging more with those communities.
3. It looks like the plans go out to about 100 years old. Open Social Security uses five different mortality tables to estimate life expectancy or an entered age, which can tune a plan to meet someone's situation much better.
4. Forced early retirement. It's happened many times in the past and likely will again.
5. Divorce. This is a tough one, but it's fairly common these days, and throws a real monkey wrench in financial plans. The rules differ per state making it lots of work. Some have fairly clear rules that could be implemented, others are kinda up to the judge. Real World Divorce  may be usable as a survey to get a feel for the scope of the different rules.
I'd like to object to your statements "social security relevant (it's better for younger folks not to count on it)" and "it's tough to forecast how much younger folks should count on it, so we've left it out for now" It's really easy to handle this. Look at how Open Social Security handles it. There is a year and a reduction percentage. He fills in defaults from the Social Security Trustee report, since they are required to estimate it based on current law. For younger folks the year probably doesn't matter, since it will happen before they withdraw, just the percentage matters. This is an opportunity to educate folks rather than perpetuate the fear that it won't be there for them. For many of them it will probably be their largest income source in retirement even after the reduction.
I don't think the complexity of the questions that early retirees have like Roth conversion or 0% cap gains is any more complex than the buy a bigger house question. Look at the complexity of the NYTimes Buy vs Rent calculator . I suspect the buy a bigger house is similarly complex when you really dig into it.
Happy to answer any other questions you have.