In follow up to my recent post on measuring the un-measurable in web analytics, I believe there are three general approaches to being able to quantify the things that maybe aren’t as straight forward.
- Make them measurable
- Change your measurement framework
1. Make the measurable – Get more data. There is virtually nothing that cannot be measured if you take the right approach. Vanity URLs, unique coupon codes, voice of customer surveys, unique 1-800 numbers, and viral discount codes are just a few examples. Assuming you’re in business to cause someone to do something, you can make that something unique in such a way as to measure its impact on your business bottom line.
With the right amount of effort (a lot), you can measure ALL of your marketing efforts. Nifty web intelligence data warehouses like Omniture Insight let you link all kinds of on and offline data together. If you can manage to connect them all with a primary key (a unique identifier like an account number), you can then slice and dice your way into segmentation heaven.
Want to know if more people respond to a picture of a red widget in your email marketing or a blue widget? Easy with just web analytics.
Want to know if people living in the Arctic Circle who visited your website within the past 6 months as a result of a particular banner ad, then went into one of your stores and bought your widget, then went home and registered their widget online were more likely to use online support or call your call centre?
Ok, maybe that’s a little complex, but it boils down to: Did that banner ad placement get me some tech saavy widget buyers who were less costly to support, or did it land me a bunch of people who are calling my call centre mostly because they’re lonely.
Or maybe you’re widget isn’t very cold-proof, and all those folks are calling in with problems…..maybe you should stop advertising on that snow-show website….
Actionable intelligence comes from a combination of measuring as much as you can, across as many channels as you can, linking that data with a primary key and segmenting audiences.
2. Estimate – Not my first choice. I don’t like estimating, it’s inherently dangerous. Analytics can be interpreted in so many ways that adding another dimension like estimation can lead you down the wrong path.
However if you combine estimation with something a little more measurable, like an a/b or multivariant test or a survey, then you can find some ROI validity.
Imagine you’re a retailer with both online and offline shopping. You know a lot of people use your website as a research tool then buy in-store instead of online. How much? There are a few ways to find out.
One way is to ask people in the store as they are purchasing. Did you visit our website prior to making this purchase? With a large enough sample size, you can then estimate the net impact of your website on offline sales.
If you’re lucky enough to have a complex multi-channel measurement system, you don’t even need to ask, you would know that I looked at your gizmo online before I went in to buy it and even how long it was between the first time I read the details on the website and how long it took me to fork over my cash in the store (but that’s not estimation….)
You can also run geographic a/b tests. Set your website or email advertising up to deliver two different types of content across two different geographic areas. Feature gizmo A on your home page in one city, and gizmo B in another city. Then correlate to offline sales of said gizmos only in those geographic areas. Gizmo A in Toronto outsold Gizmo A in Vancouver by X% therefore the correlation of impact of featuring on the home page is $XXXXX.
Of course, correlation doesn’t equal causation, so you’ll need to think the tests through very carefully. If Gizmo A has some feature to it that more useful in Toronto than Vancouver (like sunscreen), then the test will be flawed. However, if you do this enough times, you can develop a fairly accurate offline multiplier you can use to predict future product sales and website ROI.
3. Change your measurement framework – If you really don’t think you can measure the impact of your site on your business, ask yourself this question: Why do you have a website? If you can answer that question, go back to 1 and 2 and figure it out (or ask me, I can help).
Sometimes, all that is required is a shift in thinking. Never assume that the website is just table stakes and you don’t need to really measure its impact on your bottom line (or worse, think that you can’t measure its impact). Long gone are the days where marketing budgets were approved without any quantifiable ROI analysis.
In the end, I don’t even like the title of this post, because I think everything IS measurable, it’s just a question of how.