July 15, 2008
How To

How to Measure Website Engagement: Calculate Scores for Every Visitor

SUMMARY: Your website can attract plenty of visitors, but how long do they stay and how many get truly engaged? There is a way to know.

We talked to a Web analytics expert who measures visitor engagement and calculates a score you can use to segment your audience and measure the impact of a campaign. Find out which factors matter, which don't, and how to get started.
A campaign can drive lots of visitors to your website. Some will leave without making another click; others will click on a few links before surfing elsewhere. Then there are visitors who will spend a lot of time – the visitors every marketer targets.

But most marketers sit in the dark -- or a dimly lit room -- about how many visitors actually engage with their website. Measurements like click depth and duration shine some light on engagement, but not much – until now.

Eric Peterson, CEO and Principal Consultant, Web Analytics Demystified, Inc., has crafted a calculation to measure online engagement. The author of “Web Analytics Demystified” offers a new way for marketers to compile much more data on their website visitors.

“It’s a combined calculation,” Peterson says. “It’s based on a number of more basic Web analytic data like click depth and average time spent. But it’s combined in such a way to give you a measure of the degree and depth of engagement.”

His calculation can be applied to just about any visitor, including those from a campaign, a referring domain, a search keyword and an actionable item.

Who cares about engagement?

Most B-to-B and B-to-C marketers can benefit from knowing how much visitors interact with their websites.

-> B-to-B sites
Many online B-to-Bs rely on offline sales that occur after a customer visits the company’s website. If you measure engagement, you can segment the customers interested enough to do more research. You might even find that campaigns with low conversion rates are worth keeping if they generate lots of engagement.

-> B-to-C sites
Most retail-site visitors browse, they don’t buy. Measuring if they are highly, moderately or poorly engaged can give you another way to focus your marketing.

Peterson worked with a large multichannel retailer and discovered that moderately engaged visitors spent about $20 more per order than poorly engaged visitors. Also, highly engaged visitors had a conversion rate of about 5% compared to .2% for poorly engaged visitors -- a 2500% difference.

Engagement Calculation (Ci + Ri + Di + Li + Bi + Fi + Ii) / 7

Describing Peterson’s engagement model as a combined metric is an understatement. It takes seven metrics, each one an index that represents an engagement factor. Most of them will be familiar to you. They all depend on the variable “n.”

The “n” is a yardstick you set for each variable to make the calculation relevant to your business. Your “n” values should be different for every index, and every marketer will have different “n” values.

“One of the ways that you could find ‘n’ is you can simply look for the average. So, what is the average click depth for all visitors? … You would set ‘n’ to be that average. [Do the] same for duration, same for recency, same for loyalty,” Peterson says.

Indices: Engagement Factors

->Ci: Click Depth Index
Definition: the number of sessions having more than “n” page views divided by the total number of sessions by the user.

This index calculates the percentage of sessions a visitor clicks deeply into your website. How deeply? That depends on where you set “n.” For every session that a visitor’s page views exceed “n,” their engagement score will increase.

->Ri: Recency Index
Definition: the number of sessions having more than “n” page views that occurred in the past “n” weeks divided by the total number of sessions by the user.

The recency index calculates the percentage of sessions that a visitor returns to your website in a set amount of time (n) and views enough pages (n) to be considered engaged. Every time a visitor completes both actions, their engagement score increases.

->Di: Duration Index
Definition: the number of sessions longer than “n” minutes divided by the total number of sessions by the user.

The duration index calculates the percentage of a visitor’s sessions that exceed a set time. Each time a person spends more than the set time on your website, their engagement score will increase.

->Li: Loyalty Index
Definition: scored as 1 if the user has come to the site more than “n” times during the time frame being considered (otherwise scored as 0).

When people consistently return to your website in a period of time, this measure boosts their engagement score. “I score ‘1’ when visitors have come to [my] site more than five times in the past 12 months,” Peterson says.

->Bi: Brand Index
Definition: the number of sessions that either begin directly with your website (have no referring URL) or are initiated by an external search for a branded term, divided by the total number of sessions by the user.

The brand index calculates the percentage of sessions a visitor arrives at your website as the result of a branded action. When a visitor types your URL directly into their browser, or if they arrive on your site after using one of your brand’s search keywords, it will add to their score.

->Fi: Feedback Index
Definition: the number of sessions where the visitor gave direct feedback divided by the total number of sessions by the user.

“[This] is the sole qualitative input to this model,” Peterson says. It calculates the percentage of sessions that a visitor provides feedback – like through a contact-us form, an email to your company or any other way. It adds to their measure of engagement.

->Ii: Interaction Index
Definition: the number of sessions where a user completed one of any specific tracked events (but not full conversion), divided by the total number of sessions by the user.

This index measures the percentage of sessions that a visitor completes a “mini-conversion” – like subscribing to an email newsletter or downloading content. You can set your mini-conversions however you like.

->Quick division
After you get the value for each index, you add them together and divide by 7 – the number of indices. A decimal between 0 and 1 results. Multiply it by 100 to convert to a percentage. This percentage is the engagement score.

Tailor to Fit Your Business

Peterson says his model “is not an absolute calculation for all sites.” You should tailor it to your business to create scores that work for you.

-> Drop what you want
You can drop out an index because one not used has a zero value. The zero becomes a place-holder only.

-> Throw weight around
You can add multipliers to give more weight to a certain index, depending on your business. A multiplier will give this index more impact on your overall calculation.

For example, spending a long time on the website – duration – might be really important to a business. You can put a 2 or 3 or higher number in front of the duration index to add more weight.

"I don’t think they actually need to be weighted that heavily. The engagement calculation seems to work just fine in my use without being weighted,” Peterson says.

-> Set your scale
After you start scoring visitors, you need to decide what a high score is and what a low score is. The easiest way, Peterson says, is to set your website’s average engagement score as “moderate” and work from there.

Notes on the Calculation

o History adds up
Data for this calculation accumulates over time for each user. If a person visits your website on Monday and returns on Friday, your data on them will build.

o Main conversion not a factor
This calculation does not factor in your website’s main conversions.

“You already have a whole bunch of metrics that you’re using to look at the online transaction,” Peterson says. “If you are a retailer, you would look at these metrics [conversion and engagement] side by side…not try to boil them up into a single thing.”

Instead, if you want, you can have clicking the “add-to-cart” button as one of your mini-conversions in the interaction index.

o Engagement not satisfaction
Engagement with your website does not mean satisfaction. A user can be highly engaged and poorly satisfied, or vice versa.

For example, Peterson had computer trouble and spent a lot of time in highly engaged behavior trying to fix the problem, he says. He was highly engaged with the company’s website but not satisfied.

Advanced Software Required

Peterson’s engagement model is more complicated than most metrics. Some problems you might encounter:

->Software limitations
“A lot of technologies on the market today don’t have a persistent notion of the visitor,” Peterson says. And his calculation is visitor-based.

“You need to be able to go back and look at those visitors’ history of visits to the website and make the calculation for each one of those visits” and continue to add to the calculation over time, he says.

->Higher costs
The best software required to do the calculation will cost you something. Free software is usually too limited. And you must consider data-storage costs.

“It’s pretty expensive, from an architectural standpoint and a data-storage standpoint, to keep the entirety of the visitors’ history,” Peterson says.

->Software that Peterson says can complete the calculation:
o Omniture Discover OnPremise
o WebTrends Visitor Intelligence
o Coremetrics Explore
o Unica Affinium NetInsights
o IndexTools Rubix
o Other “homegrown data warehouses”

Start Small

Peterson’s model is not a take-it-or-leave-it calculation. You can start small if your system cannot support it by adding a few indices to your other measurements.

-> Look at few indices
Look at conversion rate and click depth, for instance, if you want to start small with engagement. Somebody who comes from a campaign will probably be more engaged and click more deeply into your website, he says.

“You can do the same with duration. You could do the same with loyalty. You could look at a campaign and say, ‘What was my response rate? What was my conversion rate? And on average, over the course of the next month, how often did those respondents come back?’”

-> Assess campaigns
A poorly converting campaign probably won’t sell a lot of widgets but it might lead to a lot of micro-conversions.

“So when you’re thinking, ‘Should I cut this campaign or not?’ if it doesn’t convert well, yeah, you’re probably going to cut it. But what if it has a lot of micro conversions? What if it’s pushing people along in the sales funnel, just not getting them all the way there? Do you really want to cut that campaign?”

Links related to this article:

Sherpa Article: How to build your own analytics system

Omniture: Discover OnPremise

WebTrends: Visitor Intelligence

Coremetrics: Explore

Unica: Affinium NetInsights

IndexTools: Rubix

Web Analytics Demystified: Blog archive: How to measure visitor engagement, redux

Web Analytics Demystified: Web analytics books, jobs, events and consulting

Improve Your Marketing

Join our thousands of weekly case study readers.

Enter your email below to receive MarketingSherpa news, updates, and promotions:

Note: Already a subscriber? Want to add a subscription?
Click Here to Manage Subscriptions