7 - How Good are You at Buying Noise?
Aside of the various things you can do (or that can be done by others) to drive users to your site, there is paid advertising.
Web Site Advertising
EN: The author talks about banner advertising, which is largely a dead horse, and much of what he has to say is no longer applicable to the current practices in online advertising. Other authors do a better job of this, so I'm skipping much of this.
It is worth noting that advertising can be designed to be measurable: you can rig your analysis to isolate the advertising-driven traffic (to measure the effectiveness of the ad) and to remove it from the ongoing measurements (so that traffic generated by an advertisement does not interfere with the "normal" traffic and skew your interpretations).
Metrics unique to a Web site advertisement is the click-through rate for an ad (number of visitors who were shown the ad versus those who clicked it), measured over the duration of the campaign. All else is normal Web site activity, but attributable to advertising response.
One interesting point: it's important to consider that a Web site ad is a brand exposure to all who see it, not just those who click through. Considering what impression it creates (whether the content of the ad or the association with the site it appears on) creates a negative brand impression. Sometimes, pimping for clicks can harm the customer's perception of your brand.
Email Promotions
EN: Once again, the author strays from his area of expertise to talk a bit about e-mail promotion. Once again, I'm skipping this in favor of getting information from better sources.
And likewise, e-mail promotions can be geared so that you can isolate and separate traffic generated by e-mail campaigns.
However, there are additional metrics to be considered:
- E-mails sent - This metric is simple: how many messages did you send out? If the campaign takes place over time, there may be repeat mailings.
- E-mails Returned - If an e-mail message is rejected (bad address), it should be deducted from the total sent.
- E-mails Rejected - The author doesn't mention this, but some of the e-mails that were returned as undeliverable aren't bad addresses, but it was simply rejected (spam filtering)
- E-Mails Opened - This is interesting, but hard to measure with much accuracy. Most users do not send a return receipt, even if you request it, and the trick of using graphics in a message body to monitor open rate is not reliable (text-only mail is gaining popularity)
- Clickthrough - As wit ha site ad, you should be able to measure the number of users who clicked through from your message. Also account for multiple click-through from the same message, by one user over time, or by multiple users if the message is forwarded.
- Unsubscribe Requests - Many e-mail marketers like to ignore this, but it's important to measure when a message has rubbed people the wrong way, to the point that they ask you not to contact them anymore.
He also mentions timing. With a site advertisement, you have a clear idea of the beginning and ending dates of the campaign. With e-mail, you know when you sent it out, but people may respond weeks or months later.
Viral Marketing
The author stumbles around on the topic of viral marketing: when you create a message or a widget that people send around to one another. He doesn't seem to have any good ideas for measuring the impact of this. There are probably better sources.
Valuing Visitors
When you advertise, your goal is to get people to visit the site - but not all visitors are created equal:
- Visitor - A "visitor" is anyone who visits your site. It's the most common measurement, but is vague.
- Unique Visitor - The concept of "visitors" is skewed and polluted. It's necessary to weed out multiple visits from the same person, multiple people using the same IP address, and the same person using multiple IP addresses. Cookies are a partial solution to this problem, though not perfect.
- Return Visitor - Once you know which customers are unique, you can then determine when a visitor who is reached by a given campaign is coming to your site for the first time, or are someone who has visited before, but is prompted to return to the site by the ad.
- Stale Visitor - A stale visitor is a return visitor whose last visit to the site was a long time ago. I think I see the importance of this: a person who visited last week and comes back is less significant than one who visited a year ago and comes back (the latter would have been less likely to return if they had not seen the ad)
- Qualified Visitor - Some sites look at behavior upon visiting to assess whether a visitor is "qualified" - but this is black magic. You can examine what they looked at on the site to determine whether they are likely to be a qualified buyer, but it's purely inference.
- User - A "visitor" becomes a user when they do more than merely visit. As with a qualified visitor, this is subjective. All the same, I see the value in measuring the number of visitors from an ad who then begin to visit the site regularly (as opposed to those who visit once because of the ad and don't come back).
- Customer - Simply stated, a customer is a visitor who buys something. Again, they might buy something today, or they might buy something a week later.
Much of this is "new" to me. It seems like a very good idea, and enables you to be more precise in your metrics, but I don't think any of this is standard practice presently.