So, you are launching a new web business and your main business model is selling advertising space, at least until your customers start paying up for your services. It's likely that you will have to deal with an ad agency at some point, and here are some thoughts from the buying side of the table.
1. Demographics. Your site's demographics (or, lovingly, the demo) is what you sell to the agency. Not features, not good design, not your management team. It's the demo. Currently, there are two major ways to present a demo. One is through its descriptive characteristics: "our site is designed to attract people of such and such gender, age, occupation, geography." The other way is through the audience's collective behavior: "our site is for people who are in the market for computer peripherals"; "our site is for the die-hard ice-cream fans."
There are other ways to describe your audience. Psychographic description ("our site is for very honest people") is an interesting approach, but it is yet to be widely adopted.
If you don't have a large audience already -- large enough to register on Comscore's radar -- I think you are better off with pitching the behavioral traits of your audience. This will narrow the number of potential advertisers, but the advantages of advertising on your site and not someone else's are more obvious, and you can command higher CPMs.
2. Build your site with advertisers in mind. Very often, site designers focus on user experience but forget that advertisers, too, are users who have their own goals and needs. (Oh, look, we have some space left here -- let's slap a placeholder for banners.) Audience planning should be among the first steps in designing an ad-supported site, not an afterthought. Generic media kits behind email forms frustrate media buyers who are juggling multiple plans under pressing deadlines. And if the ad unit placement is such that it doesn't generate interactions similar to the rest of the campaign, your site will be "optimized" off the plan
3. On the other hand, don't let the ad greed cripple your design. CPM ads tempt designers to add extra steps to even the simplest processes for the sake of page views. You'll have to weigh the benefit of every extra page view against the risk of having the user leave the site in frustration to never come back.
4. Allow standard IAB ad formats on your site. You'll never guess how much overhead goes into banner production, and budgets have their limits. Plus, resizing banners is no fun.
5. But custom high-engagement formats can sell for more. Think of Facebook that sells branded pages for $300K/3 months.
6. Allow enough technological wiggle room to accommodate for ad formats the demand for which may come up in the future. Yahoo bought eGroups email list service in 2000, it is ad-supported, but I think you still can't target the banners by group categories, not to mention insertions at the individual group level. Also, create the most precise targeting mechanism possible -- you'll make more money. If you collect users' age, gender and zip code during the sign-up process, there's no reason why you can't allow targeting by any combination of the three, at a premium.
7. Allow low-cost test drives. Agencies would love to be able to incorporate new interesting media into their plans, but they can't sell them to their clients without solid numbers upfront. The only way to get solid numbers from a new medium is to test it at a cost that is not prohibitive.
8. You can't overmeasure your audience. Offer metrics that go beyond the basics of impressions and clicks. CNet's sites put a link to a feedback form under each banner (I wonder if they share results with the advertisers).
9. Make data PowerPoint-friendly. An intelligent good-looking graph goes a long way.
10. Media buying cycles are always longer than what you'd prefer them to be. At least online, media buying is rarely reactive -- every new opportunity will be evaluated and filed in an appropriate folder to be pulled out at the beginning of the next cycle.