Facebook Ads in 2011: targeting, CPC vs CPM, good results

Facebook ads run in the right-hand column of the site, on the home page, profiles, Pages and photo views. Each ad is small: a title of up to 25 characters, body text of up to 135 characters, and an image of 110 by 80 pixels. That is less room than a Google text ad plus a thumbnail, so the image does most of the work.

Two formats matter this year:

• Standard ads link to a URL you choose, either a page on your site or a tab on your Facebook Page.
• Sponsored Stories, launched in January, promote something a user's friend already did, such as liking your Page or checking in at your location. They show the friend's name and cannot be edited, which is exactly why they get higher click rates.

If you plan to run more than a handful of ads, install Power Editor, the Chrome-based bulk tool Facebook released this spring. It makes duplicating ads across audiences and bids manageable.

Google knows what people want right now. Facebook knows who they are. The ad tool lets you combine:

• Location down to city plus a radius, which is good enough for a restaurant, a dentist or a regional retailer.
• Age and gender, in single-year ranges.
• Interests, drawn from what people list in their profiles and the Pages they like. "Yoga," "home improvement" and "Lady Gaga" are all targets.
• Connections: people who like your Page, people who do not, and friends of people who do.
• Education, workplace, relationship status and language.

The practical approach is to build many small, specific audiences rather than one large one. An ad aimed at women 30 to 45 within 15 miles of Austin who like "running" with a running-shoe image will beat a generic ad aimed at everyone in Texas on every measure. Facebook shows an estimated reach as you set the filters; for a local business, audiences of 10,000 to 100,000 are workable. Under about 5,000 and you will struggle to get impressions; over a million and you have stopped targeting.

A useful discipline: write the ad for the audience, not for the product. The copy that works says "Austin runners:" before it says anything about shoes.

You can bid per click (CPC) or per thousand impressions (CPM), and the right choice depends on your click-through rate.

The arithmetic is simple. If an ad gets 0.1 percent CTR, 1,000 impressions produce one click. At a $0.30 CPM, that click cost $0.30. At a $0.70 CPC, the same click costs $0.70. Once an ad has proven its CTR, switching it to CPM usually cuts cost per click by a third or more. Ads with low CTR should stay on CPC until you fix the creative.

Typical US ranges this spring: CPC of $0.40 to $1.20, CPM of $0.20 to $0.60, CTR of 0.03 to 0.10 percent. Yes, those click rates are low. They are normal here, and they are why creative rotation matters more than it does in search.

1. Decide on one goal per campaign. Likes, website leads, store visits, or sales. Each needs different copy and a different destination.
2. Build the destination first. For Likes, a Page tab with a reason to click Like, such as a coupon revealed to fans. For leads or sales, a landing page on your site with its own form and a thank-you page.
3. Tag every link. Use Google Analytics campaign parameters (utm_source=facebook, utm_medium=cpc, utm_campaign=your-name) so Facebook visits show up separately from the rest of your traffic.
4. Create three to five ads per audience, varying image first, headline second. Keep text identical across the image variants so you learn one thing at a time.
5. Set a daily budget you can afford to lose for a week. $10 to $25 a day is enough to learn whether an audience responds.
6. Check every day for the first week. Pause ads under 0.02 percent CTR, raise bids on ads that are not spending their budget, and lower bids on ads that spend it by noon.

After thirty days, you should be able to fill in a table like this for each audience.

Those lead and conversion figures depend heavily on your industry and your page, so use them as a starting point, not a verdict. What matters more is the trend. If cost per result falls through the month as you prune audiences and rotate creative, the channel is working. If it rises, the audience is exhausted or the offer is wrong.

Also judge Likes by what they do later. Facebook Page Insights shows how many fans see and interact with each post. A fan base built with a coupon that never engages again was cheap for a reason.

• Targeting everyone 18 to 65. The cost per click will be low and the value of each click lower.
• Using your logo as the ad image. Faces, products in use and bright colors get clicks; logos do not.
• Sending ad traffic to your home page. The visitor came from an ad about one thing and lands on a page about everything.
• Leaving one ad running for a month. Fatigue sets in after five to seven days with a small audience.
• Judging the channel after three days. Facebook ads improve with pruning, and the first week is almost always the most expensive.
• Ignoring the ad review rules. Ads that are rejected for text in images, misleading claims or landing pages that do not match the ad delay your whole campaign.

We run paid social as part of a connected brand, growth, web and technology program, through discover, plan, build, improve. Discover is the audience work: who actually buys from you, what they list as interests, where they live, and what you can afford per lead. Plan sets one goal per campaign, the budget split between prospecting and existing fans, and the landing experiences each needs. Build produces the ad library, the Page tabs or landing pages, and the tracking so that every dollar can be traced to a result in Analytics. Improve is the weekly rhythm of pruning audiences, rotating images and shifting bids from CPC to CPM as ads prove themselves.

A representative engagement is on our performance work: a regional consumer brand with a modest budget and no social advertising history. We built a dozen tightly targeted audiences by city and interest, tested image sets weekly, moved the winners to CPM bidding, and tied every click to a tagged landing page. The result was a cost per lead well under what the brand was paying in search for the same product, and a fan base that kept responding to posts after the ads stopped.