Building plans from audience metrics

Content site leaders pay more attention nowadays to measures of engagement beyond the venerable but flawed page view, per an overview from Jennifer Saba, writing for Editor & Publisher.

Be it time spent on site, page views per session, or frequency of visits per user, newspaper.com executives interviewed for the article want to grow 'em all.

"Many advertisers still buy based on CPMs (cost per thousand) per page view. 'The cold hard reality now is still the more page views we get, the more ads we serve. That translates into more dollars,' notes Anthony Moor, deputy managing editor/interactive at The Dallas Morning News. But he adds that executives also pay close attention to time spent: 'As an editor, I like seeing time spent go up — that's a good thing.' ...

"Chris Jennewein, senior vice president and publisher of Greenspun Interactive, thinks that time spent is a metric that has gotten shortchanged — and one that will eventually carry more currency. 'As this medium matures,' he says, 'sites that can show greater time spent per user are going to be able to command higher CPMs.'"

All true, all good. I begin business planning for any Web venture from a set of assumptions, some about traffic, some about ad performance, some constant through the plan period, some variable as time goes by.

I make the traffic assumptions in this order:

• Monthly Unique Visitors: This is the "reach" metric, a variable component typically presuming growth through the plan period. For a new product, the growth curve will be steeper, while a mature online product might see only limited growth from organic increases in Internet usage.
• Monthly Sessions per MUV: This is the "frequency" metric, which experience shows does not vary through an introductory business plan cycle as much as I might hope. So I start most plans with this as a constant. Better to be pleasantly surprised later, I figure, and reforecast to the positive.
• Session Depth: I could express this as "time spent," as Saba's article suggests, but today's Internet advertising economics do not allow us to translate time spent into money. So I use page views per session as my depth, or "engagement," metric. Again, experience shows this metric does not vary as much as I might dream as a site grows from introduction, so I start most plans with this as a constant.
• Ratio of Indexes to Details: Most sites have a different ad inventory model for home and index pages than for article or detail-level pages. So it's important to make assumptions about the split of a session between those two page types. Increasingly, detail-level pages dominate most user sessions as people arrive at content on our sites directly from external searches. I treat this as a constant through my business plan period; it is unlikely to vary enough to break it out as a monthly variable.

Then I make the advertising assumptions separately for each ad form (banner/display, text/context, listing) like so:

• Average Number of Units by Page Type: This assumption follows the ratio of indexes to details. If I have average four units on an index page and three units on detail pages, I can treat that as a constant in my plan.
• Average Cost Per Thousand: The rate my hypothetical enterprise will get for ads it sells. Another constant, though it's fun to be able to make a rate increase stick during a plan period, and adjust forecasts accordingly.
• Average Sell-out Percentage: High rate plus high sell-out plus high traffic equals success. Without high sell-out rate, all that unused inventory drags on a business' ability to manage rate and advertiser expectations. This metric, which I treat as a constant in business planning, remains critically important yet the hardest to maintain or improve even on a high-volume, fast-growing site. In fact, traffic growth naturally conflicts with sell-out rate, especially in Web environments where inventory forecasting is unreliable.

Now, it's one thing to say these are the metrics for which I make assumptions to build a business plan. It's quite another to devise reliable formulas for arriving at the assumptions.

Much as I would love to say the data to plot assumptions and forecast usage sits at my fingertips every time I do this, that's just not life. More often, we draw on personal experience, or call our pals who might be doing something similar, to see some shadow of the facts that might make our assumptions more valid than wet fingers held up in the wind.