No specific tactic works forever. There is no "1 weird trick." You need to set up your advertising and sales funnels to track as prospects become customers, and continually experiment on the message and the channels. This is where most startups start and a lot of them get it wrong. If you have boatloads of VC money, you can run a lot of bad experiments until you hopefully hit paydirt. Paydirt is +ROI ad campaigns on a large targeted demographic. Then you treat this audience like a goldmine. At first it's easy to mine and pick up gold nuggets in river beds. As the easy pickings disappear, you'll find yourself investing in increasingly more expensive machinery and infrastructure to continue mining operations. As you've extracted most the gold, look for byproduct like copper, silver, etc (in startup terms, introduce ancillary products and services). As you hit one gold mine, keep prospecting other areas (distribution channels like display ads, native ads, social media, etc).

If you end up a multi billion dollar unicorn with tens of millions in ad budget, targeted direct ads no longer move the needle like they did when you were a startup. Now for the Director of Marketing to justify increasingly larger compensation, he needs to increase his ad spend and starts whispering about spending money on ads that have no direct correlation to the bottom line. This is called brand advertising. Hint: If an early stage startup spends more budget on brand advertising than direct marketing, run.

Effective advertising is matching the product/service with some problem, on a specific channel of eyeballs, with a specific message. There are tons of new channels popping up every few years (think google, facebook, instagram, youtube, twitch, pinterest, wechat, stackexchange, hn, reddit, etc). There are millions of permutations but you have a finite ad budget. The magic is deciding how to show what to who.

Brand awareness ads. Pretty much it’s irrelevant what you say about brand ‘X’ as long as you bombard gullible observer with ‘X’. Then buying ‘X’ would be more familiar choice for millions than buying ‘Y’ or ‘Z’.