Dear Management Analytics Consultant: Merchandise Control

Dear Management Analytics Consultant:

You frequently talk about the importance of merchandise. But honestly, I don't have any control over merchandise. I'm just a marketer. Shouldn't I just assume that the merchandise is great and then focus on my role in creating a frictionless customer experience?

Thanks,

Elmer

Dear Elmer:

You couldn't be more wrong. You play a major role in determining the merchandise that succeeds and the merchandise that fails. Who owns email marketing? You do! Why can't you feature new merchandise, personalized based on prior customer purchases, every Wednesday and Friday? If you give customers a chance to buy new merchandise today, those items will become winners tomorrow. This stuff isn't rocket science. Use Instagram to share the new items that have the best chance of success, and create an in-house Influencer to convince customers to buy the stuff. That's your job, and your co-workers are waiting for you to do something. So do something!! Put your best winning items in expensive marketing channels, put new items in inexpensive marketing channels.

P.S.:  Spend less time creating a frictionless customer experience and instead cause a customer to crave your merchandise.

Dear Management Analytics Consultant: Attribution

Dear Management Analytics Consultant:

Our Marketing VP is as old-school as they come. She strongly believes that the catalog is responsible for all sales, telling me that without the catalog we wouldn't have acquired the customer in the first place, meaning that all downstream orders should be credited to the catalog. At minimum, she'll accept matchback results that give credit for all online orders to the catalog recently mailed to the customer. Meanwhile we've A/B tested our catalogs (without telling her) and we learned that 42% of demand is catalog-attributable while 58% is organic and is generated without catalog mailings.

Here's my question. How do I convince my boss that we have actual data that proves the catalog has limited value? I've even hired a pair of attribution vendors to analyze a month of data, and their analyses are wildly different and are unreliable. How can I get my boss to trust me when I can't trust attribution vendors because their algorithms disagree?

Sincerely,

Sophia

Dear Sophia:

You probably won't convince your boss that you are right and her worldview is wrong. Your boss built an entire career based on a worldview not unlike the J. Peterman character on Seinfeld had. If she accepts your view of catalog marketing (which is based on data, by the way), she has to come to terms with the fact that the value she brings to the organization is diminished.

There are several things people in your position do. If Management demands that a lot of catalogs be mailed, then mail them ... but not necessarily to housefile customers. Mail them to prospects and hold out online housefile buyers. Stuff as many catalogs into the mailbox of a first-time buyer for three months. Use email marketing to communicate to pure online buyers with high organic percentages. Recommend smaller catalogs (which are more productive - chalked full of winners of course) and then run scenarios showing that you can mail MORE of the smaller catalogs, which should be pleasing to your old-school boss.

A final note. All attribution models are flawed. They are all based on assumptions, though some of the assumptions (and math behind the assumptions) are better than others. The digital age promised us clarity. Instead, the digital age delivered confusion. Use your A/B test results to inform investment decisions. Work with a trusted attribution vendor and help shape their work with your A/B test results. Vary your search and Facebook spend on a monthly basis, giving attribution vendor algorithms more opportunities to detect key relationships.

Organic Percentage as a Key Targeting Variable

I spoke at a conference a few weeks ago, and the topic of "attribution" was popular. Folks wanted to know "how" to attribute orders to marketing activities.

The most important variable to store in your database (for attribution purposes) is the "organic percentage". At a customer level, you calculate the percentage of demand that your mail/holdout tests show happens regardless of catalog mailings (in a print environment). When evaluating your matchback results, you discount the results by the organic percentage.

In other words, if your matchback analytics (a lousy form of attribution) show that you generated $3.50 for a segment of customers, you discount the matchback analytics by the 50% organic percentage (in the example above) ... and that means that a $3.50 matchback result is actually a $1.75 incremental outcome based on mail/holdout testing.

And the difference in results is staggering. When you run your simulations for optimal page counts (and yes, the ad costs above are inaccurate due to printing efficiencies, but they're outlined as they are to prove a point), you see dramatic differences between matchback analytics and the far more accurate outcome calculated via the organic percentage.

• Via matchback, your optimal page count is between 80 and 192 pages ... fatten-it-up and let-er-rip!!!
• Via the organic percentage, you are limited to a tiny 32 page catalog.

The former is inaccurate (and benefits the entire print ecosystem).

The latter is far more accurate (and benefits your brand).

Store the organic percentage in your database at a customer level ... you know how to do this, you have your mail/holdout tests to calculate the organic percentage.

Yup, this targeting stuff works!

Hillstrom's Targeting: Consequences of Targeting

You analyze hundreds of companies and you see recurring themes. One of the themes is the impact of digital messaging on a retail audience.

Here's how this works ... you have a previously retail customer who you bombard with digital messages (frequently via email or social) ... and guess what? The customer buys online.

When a former retail customer buys online, customer behavior changes. I've run countless simulations that demonstrate what happens ... but you don't need the simulations, you've got real data to prove the point.

Here's an example ... a group of equalized customers (weighted historical spend of $450 - $499, 0-12 month buyers). There are four segments. Then, I measure next-twelve-month spend based on the segment the customer belongs to. Here's the data.

The details tell a fascinating story.

If you have a retail-only buyer, that customer will spend $184 in the next year (in this dataset, your mileage will vary).

Now you convert the customer to a digital buyer instead of getting the store purchase you normally would have obtained. Look at future retail/online spend.

• Retail spend drops from $149 a year to $126 a year ... -15%.
• Online spend jumps from $36 a year to $87 a year ... +142%.
• Total spend increases from $184 a year to $213 a year ... +16%.

In this case, we find the beloved "omnichannel gain" that vendors constantly scream at us about ... the "multi-channel" buyer spends 16% more.

But ... BUT ... spend in retail is -15% to what it would have been.

If the customer eventually converts preference online (a minority currently do this, the dynamic will change over time), then look at what happens to retail spend ... it drops from $149 to $126 all the way down to $54.

In other words, if your targeting strategy constantly screams benefits of online buying, you'll get prior retail-only customers to embrace your digital messaging ... and then what?

• You close stores.

What happens when stores close?  Well, the retail history will disappear ... and then look at the future value of the online-only buyer ... it's $175 ... in our case, it's the lowest of the four segments. You close the store and the customer spends less (Macy's has gone on the record publicly that this dynamic happens) and you might be more profitable but customers are less loyal.

This is the consequence of targeting in a retail environment. Your targeting strategy shifts customers in/out (usually out) of stores. When that happens, the store dies, and when the store dies, customers who used to shop in that trade area become less valuable.

#Omnichannel!!!!!

You might have a great targeting environment ... and you might do a spectacular job of targeting. But short-term ROI measurement is feckless if you don't understand the consequences of targeting. Run simulations and understand what the long-term impact of targeting strategies are. You HAVE to do this, right? RIGHT??

Run the simulations.

Understand what the consequences of targeting are.

Then make better decisions.

Contact me (kevinh@minethatdata.com) if you need help. Pricing information is outlined here.

Hillstrom's Targeting: Calculating The Organic Percentage

I recall analyzing this test at Eddie Bauer ... in 1996.

1996. Twenty-three years ago.

Of course, we didn't have mobile in the mix, so we just removed that column and analyzed the rest.

You have your mailed segment, you have your holdout segment, and you run the test for three months or six months or preferably a year. Then, by channel, you compare what you sold in the mailed group vs. what you sold in the holdout group. Look at the results (bottom arrow).

• Mailed-In Checks = 10% Organic (90% catalog driven).
• Call Center / Phone = 30% Organic (70% catalog driven).
• Desktop / Laptop = 60% Organic (40% catalog driven).
• Mobile = 80% Organic (20% catalog driven).

The overall average was 49% Organic (51% catalog driven).

Now, you go into your database and you use weighted demand and weighted organic percentages and you calculate the historical organic percentage for each customer. Here's an example:

• 1 Purchase for $100 0-12 Months Ago, via Desktop / Laptop.
• 1 Purchase for $100 13-24 Months Ago, via Mobile.
• 1 Purchase for $100 25-36 Months Ago, via Phone.
• 1 Purchase for $100 37-48 Months Ago, via Mail.

In this project, weighting is as follows:

• 0-12 Months Ago = 100%.
• 13-24 Months Ago = 60%.
• 25-36 Months Ago = 35%.
• 37-48 Months Ago = 20%.

Therefore, I have the following amount of weighted dollars:

• 100*1.00 + 100*0.60 + 100*0.35 + 100*0.20 = $215 weighted dollars.

And each channel has an associated organic percentage, yielding weighted organic dollars.

• 100*1.00*0.60 + 100*0.60*0.80 + 100*0.35*0.30 + 100*0.20*0.10 = $120.50 weighted dollars.

The calculation for historical organic percentage is straightforward:

• $120.50 / $215.00 = 56%.

In other words, 56% of weighted historical spend is "organic", and 44% is driven by catalog marketing.

This percentage (56%) is stored in your database ... it's calculated in real-time or weekly or whatever works best for you. But it is scored for every single customer in the database, regardless.

It's a good idea to create another variable .. .an "organic percentage segment" that has Low / Medium / High designations for the organic percentage.

• 0% to 33% Organic = Low.
• 33% to 67% Organic = Medium.
• 67% to 100% Organic = High.

The designation allows you to appropriate target customers at a simple level. It also allows you to store the segment in post-campaign analytics, allowing you to measure if high-organic-percentage customers generate incremental profit when you mail catalogs.

Use the template above, and combine the template with your mail/holdout test results and you've got something interesting, don't you?!!

Hillstrom's Targeting: It Works For Catalogs Too!

You've heard me talk about the "Organic Percentage" ... over and over again. It's the percentage of sales that are not caused by catalog marketing. Smart catalogers figured this stuff out fifteen years ago (we studied this at Lands' End in the early 1990s ... we called it Cannibalization back in the day).

From a targeting standpoint, you want to do the following:

1. Mail catalogs to customers with a LOW organic percentage.
2. Greatly reduce catalogs to EVERYBODY ELSE.

It turns out that our targeting framework works very, very well when evaluating the Organic Percentage.

In our dataset, here are customers with Quality = "A" ... they're the very best customers. I segmented the customers based on prior Weighted Organic Percentage, and then measured in the next month how much customers spent ... organically and via print. Here's the table.

Again, these are the best customers ... and look at what happens in the High Weighted Organic Percentage segment ... those customers generate 79% of future demand organically. Now, because these are best customers, you'll still mail 'em catalogs.

Here's the same table for customers with Quality = "C".

Look at the High Weighted Organic Percentage segment ... they generate 75% of future demand organically. This means they'll only generate $1.19 because of catalogs ... whereas Low Weighted Organic Percentage customers generate $4.26 because of catalogs. If you mail the High segment twice a month, you're doomed!!!!

So please, get a High / Medium / Low Weighted Organic Percentage variable into your targeting framework ... and then capitalize on it!!!

And if you don't have the resources to do that, contact me (kevinh@minethatdata.com) and I'll do it for you, ok?

Hillstrom's Targeting: Combine Anniversary Events and Email Clicks

Last week we talked about the importance of recent email clicks ... and we talked about the importance of "Anniversary Events".

Combine the two and you're really got something!

Let's look at April repurchase rates by Customer Quality, 30-Day Email Click, and a Prior April Purchase. Ready? We'll simply by looking at Customer Quality = "C", ok?

• 3.7% for No Anniversary, No Recent Email Click.
• 6.2% for Yes Anniversary, No Recent Email Click.
• 7.0% for No Anniversary, Yes Recent Email Click.
• 10.2% for Yes Anniversary, Yes Recent Email Click.

Looks like the combination of targeting variables yield a highly meaningful result, don't you think??

If you have a customer approaching an Anniversary Event and showed interest by clicking through an email campaign in the past month, you better use all of the targeting tools at your disposal to encourage a purchase.

Right?

Go get busy, right now!!