Measuring Paid Search, E-Mail, And Social Media Influence Via Matchbacks
Ted asks "how do you measure influence rather than direct sales"?
Direct marketers use Matchback Analytics to attribute sales to the activity that theoretically caused the purchase to happen. Matchbacks were originally designed to prove that catalog mailings were responsible for web sales. Now, matchbacks are well suited to measure influence.
Let's look at a few customer orders.
Order #1: Customer received catalog on November 1. Customer purchased online on November 10, using the keycode from the back of the catalog. This one is easy, the catalog gets credit for the order.
Order #2: Customer received catalog on November 1. Customer received marketing e-mails on November 3 and November 5. Customer purchased online on November 10, and did not use a keycode. The catalog brand would probably allocate this order to the catalog, ignoring any role that e-mail marketing played in the purchase.
Order #3: Customer received catalog on November 1. Customer received marketing e-mails on November 3 and November 5. Customer clicks through to the website from a paid search term on November 10, purchases online, and does not use a catalog keycode. Catalogers would like to allocate this order to the catalog, paid search mavens might want to allocate this order to paid search, e-mail marketers probably lose out in this instance.
Order #4: Customer received catalog on November 1. Customer clicks through to the website from a blog featuring merchandise offered in the catalog. Customer purchases online on November 10. Catalogers would immediately allocate this order to the catalog.
In the last three instances, the marketer "assumes" that one form of media drove the order, and creates business rules to proceed with allocation of sales. And, of course, in the last three instances, the catalog marketer makes incorrect assumptions. The assumptions are better than the assumptions made in 1999, but the assumptions are flat-out wrong.
When a catalog brand measures influence, there are two different allocations.
There is direct allocation, as illustrated above.
Then we have influence allocation.
In Order #2, the catalog gets credit (if that is how matchback business rules are written), while each e-mail campaign gets half-credit for influence.
In Order #3, the catalog gets credit, while two e-mail campaigns and paid search receive one-third credit for influence.
In Order #4, the catalog gets credit, while social media receives 100% influence credit.
Each quarter, we produce a table that illustrates, for each channel, direct attributed sales, and influenced sales.
| Direct Sales | Influenced Sales | Index | |
| Catalog Marketing | $10,000,000 | $1,200,000 | 0.12 |
| E-Mail Marketing | $1,000,000 | $4,000,000 | 4.00 |
| Paid Search | $2,000,000 | $4,000,000 | 2.00 |
| Other Online Marketing | $1,000,000 | $1,000,000 | 1.00 |
| Social Media | $100,000 | $3,000,000 | 30.00 |
| Mobile Marketing | $100,000 | $1,000,000 | 10.00 |
What you are likely to see is that emerging channels have a high "influence index". In other words, we don't attribute orders to the emerging channels --- we simply don't have business rules to do this, so we attribute orders to the most established channels. But if we focus on influenced sales, we notice that channels like e-mail and paid search and social media play a bigger role, helping cause an order to happen.
Influenced sales make a huge difference in viewing a "mutlichannel strategy". In the illustration above, e-mail, paid search and social media are key influencers. They do not get direct ROI attribution, but are clearly used by the customer as part of the purchase process. From a strategic standpoint, these channels should receive strategic attention. Or maybe catalog marketing should not receive direct credit for all orders!
Either way, the marketer views the world differently when focusing on both ROI and influenced orders.
A final note: In a perfect world, the marketer executes catalog and e-mail test/holdout groups, then measures influence in mailed/holdout groups, subtracting the differences to measure true ROI and true influence.
posted by Kevin @ 8:15 PM
3 Comments:
Thank you for your thorough response.
What you propose makes a lot of sense when looking at direct response and support media like emails and catalogs.
As you say, the holdouts are more reliable because otherwise the evaluator must make some assumptions that may or may not be wholly accurate.
When originally asking the question, I was thinking about social media efforts that probably have influence on sales, but appear at first glance untrackable.
In fact, the social media phenomenon revolves around the idea that it has enormous influence on overall sales.
Social media is usually not testable by geography. And demographic splits are generally predetermined by the product itself. So setting up measurable splits for testing don’t seem to work.
From a sales tracking perspective, social media applications fall more within the sphere of awareness and positioning advertising. How do you attribute sales to these strategies?
So how will social media proponents tie sales to their budgets? No one to my knowledge has persuasive and reliable metrics for these types of efforts.
Back in the stone ages (January 2007) at Nordstrom, we made a list of the top 1,000 referring URLs that were social media related. Anything with a blogspot or typepad or wordpress or facebook or twitter in the URL was coded as social media.
We also went through, by hand, and visited other top referring URLs, to see if they were social media related.
Once we had our list of blogs, we actively tracked the users that came from social media sites, measuring their behavior compared with other users. We knew that we had a "x.y%" conversion rate, we knew if the user visited the site then purchased in-store (as long as the user was a prior purchaser, of course).
So we could tell the impact that social media was having --- just takes some elbow grease. And we knew that what we were measuring underestimated the true impact --- but we knew that we were measuring far more than any of our competitors.
I'd just keep it that simple, do something! We get too paralyzed by trying to do things the right way --- we're better off just doing something!
That certainly deserves an "A" for effort. But I guess I would have to actually see the data to determine if buyers would have bought anyway in the absence of social media.
But I'll take your word for it that you saw impacted sales.
Good point on not getting analysis paralysis.
But it does seem better to pursue profitable activities that are accountable rather than untrackable efforts when the budget is severely limited.
In other words, try new things without holding back on what you know works.
I have consulted with some clients who are so enamored with the next new thing that they never maximize (sorry for using that word) what works.
Post a Comment
Links to this post:
Create a Link
<< Home