By Ian Lurie
That led to a lot of really poor analysis. Tools popped up everywhere claiming they could “measure social media ROI.”
- Determine the value of a single follower on a specific social network.
- Determine the value of a visitor from a social network.
- Create a metric you can compare to other channels, such as search.
First, some variables
Just for reference, here are the variables I’ll use, and what they mean: Fv = Follower lifetime value; Uv = Click lifetime value; LTv = Average customer lifetime value; TOv = Total lifetime value; F = Total followers; U = Total unique visitors; C = conversions.
Don’t worry. I’ll explain them in plain English, too.
Next, a couple of disclaimers
I’m not trying to track offline conversions. You can do that easily enough with custom phone numbers and other devices. But in these examples, a conversion is a conversion.
If you spam your Facebook page, buying followers for $5 per 5,000, these formulas don’t work. Math can’t account for silliness.
Option 1: Clicks to conversions
The first option is the most obvious and least useful:
- Measure the total number of unique visitors from a single site.
- Measure the total number of new customers acquired from those visitors.
- Multiply new customers by the average lifetime value (LTv) of a customer. That’s total LTv generated.
- Divide that by visitors. That’s your lifetime value per visitor.
- Then, divide total LTv generated by followers. That’s your lifetime follower value.
- TOv = C * LTv
- Uv = TOv / U
- Fv = TOv / F
Note that this works just as well for retail or B2B. Same variables, different LTv calculation.
This method is pretty straightforward, but it has problems:
- If the lifetime value of a social media visitor/follower is lower/higher than your average LTv, your data will always be wrong.
- It completely ignores multiple-click and indirect attribution. What if I follow you on Facebook, then refer a friend to you? That’s impossible to track. But what if I follow you on Facebook, then later see an Adwords ad and click that before hiring you?
Method 2: Contribution to conversion
You can improve on option 1 by adding attribution modeling into the mix. With an attribution model, you can track all of the channels a customer uses to find you before conversion and credit revenue to each channel.
Why bother? Because customers often click to your site from multiple sources before they convert. Consider my plumbing company as an example again: John has a leaky faucet. He’s been trying to fix it for months. It keeps him up at night and is turning his cat into a lunatic.
John searches on Google for “fix leaky faucet” and finds a blog post on my website. He reads it and tries to fix the faucet. My analytics software records that click.
The repair works. W00T! He likes the post via Facebook.
Two weeks later, his now-lunatic cat chews through the garbage disposal’s electrical cord. Bzzzzt! Now John needs a replacement (garbage disposal—the cat lived). He heads to Facebook, finds me and clicks through to the site. Then he contacts me via our form.
So, his path to conversion is:
Google Search >> Facebook like >> Facebook page >> click to site >> conversion
If I use standard last-click attribution, the Facebook page gets all of the credit.
If I use first-click attribution, Google gets all the credit.
Neither is ideal—I want full attribution of all channels. That requires an attribution model.
There are lots of different attribution models out there. I’m not going to explain them all here. You can read a good tutorial on them on Google’s support site.
I prefer to use the time decay model, regardless of the analytics software I’ve got. Historically it’s given me the most conservative attribution calculation, without totally disregarding all non-converting clicks. Time decay awards more credit to the clicks closest to the time of conversion.
If your analytics software supports multi-channel attribution, then applying an attribution model is fairly straightforward:
- From the attribution model report, get your time decay channel conversion count. Not revenue. Remember, we’re using LTv for that.
- Use this number for C, instead of last-click conversions.
- It may double count. Say you’ve got a really long conversion path, with visitors using five to six different channels to reach you before they convert. How big a problem this is depends entirely on how fussy you are.
- It can’t account for true importance of each channel. It’s always possible that the second click (from, say, Twitter) is more important than the last click (from Facebook), because of the influence of the tweeter.
- It ignores the most powerful part of social media: The social graph. If an existing fan recommended my company to you, and then you clicked through to my site, that’s a completely different transaction than if I saw a sponsored tweet (don’t laugh) or a sidebar ad.
- No one ever believes it. No matter how many times I show social media’s influence via attribution modeling, folks still shut down campaigns, then wonder why their revenue plunged. I have no fix for this. If you do, please, tell me.
- It still doesn’t account for potentially lower LTv from social media.
In a perfect world, we’d use option 2 whenever possible, option 1 when we had to and then add in an essential ingredient: brain power.
There are many ways to improve measurement, but they take some doing, differ depending on the specific social media site and require that everyone have a basic understanding of how social media works. That’s a tall order, but:
- Measure LTv for social media conversions separately. Use that LTv to adjust your formula. Not hard, but really sensitive to all those times someone removes a tracking tag, incorrectly tags an e-mail or breaks analytics for an entire site.
- Track social graph mentions and relationships. Create a separate formula to track the value of first-degree relationships: See how much a new follower who is a friend of an existing follower is worth, versus an “out of the blue” follower. This is different for every social site, unfortunately, and requires API knowledge to boot.
- Add a separate tracking score, such as the Net Promoter Score, to each channel. Then assign a value to the NPS. That will mean you have a single metric you can use across all channels.
Above all, measure
You can’t truly measure the value of any marketing: value assigned to a channel based on past transactions will never be 100 percent valid for future ones, because we’re selling to human beings, not computers. Every interaction is a little different.
You should still measure everything, especially in social media. Data may not be perfect, but it provides insight where you’d otherwise have none. More than any single formula, stick to the principle of all good analytics: Consistently collect data and analyze. Continuously check your analysis. That means improvement, which is what we’re all in this to do, right?