In the physical world, a runway is a takeoff strip for airplanes. If you were in charge of designing an airport, you'd have to figure out the ideal runway length that would allow all planes to take off and land safely. How long should your runway be? It turns out 1.25 – 2.5 miles is a typical length. You could save money by building a 1-mile version, but that's too short to be useful. On the other hand, you could build a 4-mile behemoth, but that would cost a lot more without providing additional benefits.
In the startup world, runway is how long your company can survive if your income and expenses stay constant. When companies raise money, they are literally trying to increase their runway ("We spend $100k/month and have $400k in the bank. If we raise an additional $1m, we'll have fourteen months of runway in the bank instead of four.")
The same calculus that applies to airport runways applies to startup runways. If you raise too little money, you won't have enough runway to do anything meaningful. If you raise too much, you're wasting resources (equity).
How much runway do you need? Eighteen months is a good default for most seed stage startups. That gives you 12-15 months to hit strong milestones (e.g. a product launch, a certain number of engaged users, a target for monthly revenues, etc.) and 3-6 months to raise your Series A. If you raise enough for more than ~18 months, it's likely that you either gave up too much equity or are not aggressive enough with your plans. If you don't raise enough for ~18 months, there's a large risk that you'll run out of money before your make enough progress to raise your next round.
Raising too much
It's tempting to think that having 2 or even 3 years of projected expenses in the bank is an asset, but in my experience it can actually be a liability. The more time you have, the less pressure there is to perform. You can spend months or even years being a perfectionist instead of releasing what you have and seeing if people want to use it. I've seen several companies spend years working on beautifully constructed software architectures only to find out that there were no customers for the products they were building.
Raising too little
Raising less than 18 months of runway starts to put too much pressure on you and leaves little room for iteration. Let's say your product roadmap has you launching in 8 months. Given that it usually takes at least a few months to raise a Series A round, if you 12 months of cash in the bank, you're only leaving yourself one month post-launch to turn your product into a hit. If growth is slower than expected or the initial release has bugs that you need to fix or it turns out the product doesn't adequately address its users' needs, your company is likely finished. If you had an extra 6 months of cash, you would have a time cushion that lets you iterate and experiment.
Raising way too little
I occasionally see founders raising very small amounts, like $300k at an $5m valuation. They are trying to get enough cash for 3-6 months so that they can hit a few small milestones and then raise more money at a higher valuation (e.g. an additional $1.7m at an $8m valuation). I think this is penny wise and pound foolish. The more frequently you need to raise money, the more time you waste fundraising instead of building, and the more likely you are to run out of money at some point. If you need to raise one small round every 6 months, that's 3 potential failure points over an 18-month period. If you have the opportunity to do a single larger round up front, even though you give up a little more equity, you only have one potential failure point and you have a lot more time to work on your product instead of wondering if you'll be able to make payroll next month. Tomasz Tunguz recently posted a good analysis of startups' seed rounds. Tomasz found found that companies which raised $300k were about 65% less likely to raise a Series A than companies which raised $900k or more.
Not allocating time for fundraising
Founders sometimes explain their plan as follows: "I'm raising $1.5m for 18 months of runway. In 18 months, I plan to hit milestones X, Y, and Z, which will leave me well-positioned for a Series A." The flaw with that kind of plan is that you can't start raising your next round the day before your money runs out. If you need to have more money in the bank in 18 months, then you should be hitting milestones X, Y, and Z in 12-15 months, not 18 months.
The lesson is simple: when you're building a runway, make sure it's long enough for your startup to take off. For most startups, 18 months is the magic number.