Here is my take after reading the slides from a student perspective. If we put more resource on marketing would it mean at some point we will oversell our service? In other words, we can only handle 100 users but now 120 users because of the good marketing? Then we have to hire more people (but that takes time too) to make the system more robust (from both deops and software engineering). That's a problem, isn't it?
Would it be okay to say a startup should have a good estimate of how many users they can handle at each cycle?
Capacity planning is a good exercise, but the brutal truth of the matter is that the most common scaling problem of startups is to have no scaling problems at all.
Especially with B2B startups. I could point to a client or two which had "We're having trouble holding on to the rocket ship" at some points along their growth curve. The far more common outcome, though, is to be rate limited on customer acquisition.
If you ever are in the position where you have 120 people champing at the bit to buy your stuff but can only service 100, great, there exist easy solutions to that.
The easiest is a variant on Raise Prices. For example, back in the day when Slicehost was rate-limited on the number of boxes Dell would let them put on their personal credit cards per month, and they had a substantial waiting list, they let people bid their way to the top of the waiting list by using non-refundable pre-pay for service. If you pre-paid for 6 months, you leapt ahead on the queue in front of those only willing to pre-pay for 3. This let them convert pre-paid cash into boxes, allowing them to onboard the entire wait list within 2 months of instituting the policy.
That's the very definition of a good problem to have.
It's much easier to over-estimate your potential success (and worry about it) than to under-estimate it and find you've done a bunch of work that was a waste of time. I've worked on features where I've over-thought them and designed them to limit the negative effects of being over-used, only to find that the real challenge is getting anyone to try them at all.
Fear of overselling could kill your company - it puts a brake on everything you do. Even worse, it gives you the perfect excuse not to market your product to the max, and human nature means you'll actually market it close to the minimum.
My philosophy is to ask "will it scale?" for every part of your business. EVERY step - including sales, support and things like invoicing too. This adds surprisingly little overhead, but if Jon Stewart mentions your product on The Daily Show, you'll be popping champagne corks instead of blood vessels.
On the platform side, services like Appengine make this easy. But you need to have plans in place to scale support, sales and everything else.
Rather than asking "will it scale?", I suggest asking "could it scale?". There's no point in doing up-front work to make something scalable if you don't know if it will be needed, but it's also important not to design features that absolutely could not scale under any circumstances. If you can think to yourself "if I need to scale this, I've got a rough idea of how I would do it" you'll probably be fine. And if not... once again, not being able to scale fast enough is a great problem to have (telling an investor "I don't have time to talk to you, our server's can't keep up with all of our new signups" is a great pitch).
Would it be okay to say a startup should have a good estimate of how many users they can handle at each cycle?