Disclaimer: There are no good decisions in startups ! You have to pick the least bad decision from a collection of bad decisions. This is a blessing in disguise because least bad decision would always involve something that can be done fast, something that's reversible and something that's measurable. There, we have given you the gist.
What Costs Least
Any good framework needs to be objective and be able to put weights on various parameters. A good approximation, in the absence of clear numerical parameters, is the Cost Function. Simply put - it measures the cost of doing an action. The cost of anything are really limited to 2 parameters - money and time. You can think of many others, but they will ultimately boil down to these two - complexity of a task (more time), number of people needed to do a job (more money), convincing people for approvals or for making a sale (more time, also money if you need to travel to meet them) etc.
This is where we like to use Reid Hoffman's mantra on speed (link). He has talked about speed of execution and taking decisions a lot. Hoffman helped found Paypal and LinkedIn - so it's not a terrible idea to listen to him. Essentially, when he is talking about speed, he is talking about the cost function in terms of time.
Se we have our first parameter - the CHEAPNESS of doing something.
What's Least Risky
Some decisions are hard to reverse - and therefore riskier. For example, signing up a deal with a large company, or an interested investor are really irreversible decisions, for most part. It's ok to take time before deciding on these.
However, introducing a new feature is less risky - unless it changes the experience of existing features - which, if not handled well, could lead to churn. If a new feature does not massively affect existing features, it's a reversible decision, and therefore one should not worry too much before deciding. This applies specifically to small enhancements where you want to choose between 2 or 3 options.
Many a time, we have seen entrepreneurs bother too much about little details (citing "devil's in the details" maxim). But it doesn't matter - you can do it, and measure the results, and undo it at a later stage. Anything that takes less than a day to implement can be simply tried out, and iterated upon. This is where we like to follow Jeff Bezos - who has talked about(link) taking reversible decisions without much debate.
So we have our second parameter - REVERSIBILITY.
What can be measured
Well, to be honest, we can measure most things, if we try to. There is a science to statistical measurements, and while it comes with it's own caveats, it is doable. However, all measurements are not created equal. Let's say you have 100 paying customers, 5 features in your product, and 20 micro-features in each. By micro-features I mean small design elements, micro-copy, some configuration options in features and so on. It it highly likely that if we pick one micro-feature, it does not directly affect more than 2-3 of your customers. Making changes to these will likely not give you any measure-able data. Instead, changing your hottest features slightly can have massive and statistically measurable results.
We have all been guilty of introducing one or other feature because we feel it's the "right" way of doing something, or because the other option is "fugly, and visceral to my heightened sense of design". Chuck it ! If you can't measure it in a meaningful way, you might as well not do it.
So we have our third parameter - MEASURABILITY.
So here's the framework. Look at the time and money involved in doing something (the cost), reversibility, and measurability and pick the one that fits best.
Caveat: this framework assumes that the impact of the decisions you are comparing is similar. If the possible impact is dissimilar, add that as a fourth factor. That's a little harder to do though, because it depends on many factors outside your control. One easy way to measure impact is to ask your customers if they like what you just did - if the decision was related to your product or website. LightCat.io will help you there. (bet you didn't see that coming ;-))
Well, there's one more thing - the values you associate with each of these parameters depends on your own skills, network and situation. For example, if you have a decent viewership of your blogs, but not of your website, you can easily measure the impact of your blog.
Here's an example that I filled out for before deciding whether to write this very blog, or go make product page for the website.
You can see that the cost, for me, to write this blog, was substantially lower. That's because I knew my thoughts are clearer on the blog than they are on the design of the product page. A blog just needs copy, no design - a webpage is a combination of both and thus needs more mind-space.
So there we go - a small quick framework that works. It's really a simple tactic to take decisions fast and with limited data. Enjoy :-)