What should you do when your startup's main web application becomes unusable, but the code is hard to maintain?

Before you can worry about where you are now focus instead on how you got there. Odds are that if you rewrote this app from scratch with the same team then in another 4/5 months you would be in the same mess with a slightly different set of features. Why were features added to this app? When features were added why was their impact on the user experience not considered? Why were features not removed when they caused problems? Why were buggy features allowed to reach production? Why are they still in production? Why was no one watching the complexity of your interface or overall user experience? A "5 whys" exercise might be helpful here. Really reflect on how you got into the position you are in. Until you identify at least some of the root causes at work here and stop doing them you cannot take any steps to actually fix the problems you identified above. Step one: stop making it worse. If you make it past step 1 then you can start to prioritize improvements. This almost certainly means removing features. Identify the few that are important, or actually work, or actually make you money. Make a list of tasks to remove or correct the rest. Prioritize that list. Find the single most important tiny little thing you could do to improve the site. Smaller is better. Do not allow large sweeping refactors to distract you, do the absolute minimum needed to make the most important deficiency on the site slightly better. Step two: fix just one thing. Got one tiny thing fixed, hopefully you even wrote a test for it? Good. Now ship it. Even if only to an internal staging version of your site get that change deployed. Do not try to fix anything else just make sure you can release even the tiniest improvement without repeating the mistakes you identified in step 1. Step three: ship it. Repeat.

If your product is viable and you have users and you can generate revenue, then persevere. First of all, I have to disagree with above. Never set yourself a metric like "No code is written until you reach 80% coverage". You'll still be sitting there 10 months from now writing unit tests for silly things because you must reach 80% code coverage. Unit tests are important when they are important and they are not when they are not. The trick is to be pragmatic and unit test where its needed. Eric made testing the highlight of his answer, while I think the highlight should be getting your start up restarted up. Second, figure out what you want your end result to be. Start from the front end. Take your designer, and together mock up the entire user interface as you'd like it to be - as it should be. At the same time have your back end team agree on what are the requirements in regards to the data. Design an API that exposes just the end points you need, i.e., only the calls that are necessary for a front end to interact with your data. The two teams along with the product people should interact constantly to determine those requirements. This shouldn't take long as you have an app already. You just have to simplify it and describe it in terms of an API. Its best if the front end people drive this car I think as how the data is interacted with will be their responsibility. Then tie it all together. The front end should talk to the API. The API should interact with the models and data that make up your application. If tomorrow you want to build another client (a mobile app for example) it will also talk to the same API. A day after you might want to open the API for external 3rd party apps. You will be able to do that easily because you have the API already. If it took you 4/5 months to build something bad, then with the lessons learned you can build something Great (tm) in 3 months or less. Remember to dilute, dilute, dilute. Features that nobody wants - out. Features only 2 people want - out. Features that don't really work and nobody ever had a chance to want - out. Start fresh and simple and build from there. If you must fire someone, fire those who do not take responsibility and acknowledge that a restart is required. A rewrite is not a good or bad thing. Its something that needs to be taken in context. If your app is unusable and the code base is impossible, it sounds like a rewrite is your only option. Embrace it as a 2nd chance.

The fact you are in this situation to begin with says a lot. My advice: hire a new CTO.