I’m not quite as skeptical as he is, but I do agree that it’s a far more complicated problem than our models and spreadsheets and projections would indicate.

The “marketing ROI” entry in Mother Goose’s Book of Marketing Fairy Tales, might say this is what happens with your ad campaigns.

1. A prospect sees your ad and clicks on it
   • The ad server (e.g., Google Adwords) drops a tracking cookie that corresponds to that effort
2. The prospect goes to your custom landing page and through the purchasing process
   • Your page is coded to match that effort so your back-end system knows the source of the order
3. Your new customer gets a “thank you for your order” page
   • The conversion code records an order from the ad

   • Your back-end system records an order on the corresponding priority code
4. All the codes match up nicely. Yeah.

In Grimm’s Nursery Rhymes for Real-World Marketing, this is what happens with your ad campaigns.

1. A prospect sees your ad

2. Life happens
3. Your new customer purchases
4. Your codes are all screwed up

Digging into step #2 a little deeper, here are some examples of the kinds of things that can happen to make the conversion code from your ad campaign different from the code that gets into your back-end system.

Scenario 1 – the no-click conversion

1. Your prospect sees your ad but doesn’t click on it

2. The prospect googles some of the words in your ad, clicks on a natural (not-paid) search result and lands on some other page that has nothing to do with your ad campaign
3. Your new customer purchases the product from the wrong landing page
4. Your back-end system records an order on some irrelevant code
5. The money you spent on the ad seems to be wasted.

Scenario 2 – the click and print conversion

1. Your prospect sees your ad and clicks on it
   • The ad drops a tracking cookie that corresponds to that effort
2. Your prospect reads your customized landing page with all the right promotion codes on it, prints it out and shows it to his boss / wife / palm reader.
3. The boss approves the purchase.
4. Your prospect goes to your site and navigates to a generic order page with the wrong promotion codes
5. Your new customer purchases and goes to “thank you” page.
   • The ad system records a conversion, because the customer got the tracking cookie and made it to the “thank you” page.

   • Your back-end system disagrees because the order came in on a generic order page.

Scenario 3 – the click and delete cookies conversion

1. Your prospect sees you ad and clicks on it
   • The ad tries to drop a cookie, but your customer has them turned off, or deletes them later
2. Your prospect goes to the custom landing page and purchases
3. The thank-you page can’t record a conversion because the tracking cookie is gone.
4. Your back-end system shows an order, but the ad interface doesn’t.

Scenario 4 – The “can I find a better deal?” conversion

1. Your prospect sees your ad and clicks
   • The ad drops a tracking cookie that corresponds to that effort
2. Your prospect reads your customized landing page with all the right promotion codes on it, then wonders if he can get a better deal on another page.
3. Your prospect spends ten minutes searching the web for a better price and ends up on some other page with coding that doesn’t match your ad campaign.
4. Your new customer purchases and goes to the “thank you” page.
   • The ad records a conversion, because the customer got the tracking cookie and actually purchased.

   • Your back-end system disagrees because the order came in on a generic order page.

Scenario 5 – The 30-day cookie conversion

1. Your prospect sees your ad and clicks
   • The ad drops a tracking cookie that corresponds to that effort
2. Your prospect reads your customized landing page with all the right promotion codes on it, then gets distracted by something and gets on with life.
3. 29 days later you send an email to this same person promoting the same product. The customer thinks, “yeah, I want to buy that,” clicks on the link in your email and goes to a custom purchase page – for the email campaign.
4. Your new customer purchases and goes to “thank you” page.
   • The ad records a conversion, because the customer got the tracking cookie and actually purchased within 30 days.

   • Your back-end system disagrees because the order came in on an email promotion page.

Scenario 6 – The “I don’t buy online” conversion

1. Your prospect sees your ad and clicks
   • The ad drops a tracking cookie that corresponds to that effort
2. Your prospect reads your customized landing page with all the right promotion codes on it, but wants to purchase by phone, mail or fax.
3. Your back-end system records on offline order. The online ad gets no credit.

All these scenarios happen all the time, and as you can see, there are any number of ways that the tracking from your online ad system can get messed up — either in its own right, or in its relationship to your back-end system.

The conclusion is that you have to hold on to your marketing ROI calculations somewhat loosely.

It’s fair to compare the ROI of two campaigns within the same system — for example, two different keyword campaign in Adwords — because it’s reasonable to assume that all the chaos mentioned above would happen about equally to both of them.

But it’s not necessarily reasonable to disbelieve the conversion statistics from an online marketing campaign because your back-end system isn’t showing the same results.

It’s important to track things as well as possible, but at some point you just have to believe that it’s working something like what the reports are saying.

But there’s a downside. Your stats for store.name.com/index.html and www.name.com/index.html will get combined.

There is also a solution. You can create an advanced filter to keep them separate. See How do I track all of the subdomains for my site in one profile?

Update: Unfortunately, this messes up the site overlay feature. When it tries to create a site overlay it puts your subdomain in the wrong place in the URL.

Some articles will give you some fixed guidelines. For example, How to Fix a Leaky Web Site says under 25% is good, but over 40% is too high.

I don’t buy that because your bounce rate will depend on a whole lot of things that will vary from site to site. A blog has a completely different site architecture than a store that sells dishwasher parts, and different sites get different sorts of traffic.

Here are a few things to look at to get a sense of what’s going on with your site.

Look at your traffic sources and compare the bounce rates for each. (E.g., direct traffic, referring sites, and paid search vs. natural search.) Look at your top landing and top exit pages.

Do you see any patterns? Look at the pages with a low bounce rate and see how they differ from the pages with a high bounce rate. Do they attract different sorts of visitors? Is there a clear call to action, or some obvious next step on one page and not on the other?

Remember that a high bounce rate may simply be a sign of poorly qualified traffic. Your site may rank highly for a word that has several different meanings. (E.g., “cobra” can be a snake or a kind of mustang.)

I’ve set up some “advanced segments” in Google Analytics that let me track how many people stay for one page view, or two, or three or more. I then run a report on my top search terms and see what percentage of my traffic falls into those three groups. Good pages not only have a low bounce rate, but they have a high “time on site” and a high percentage of people in the “3 or more” category.