Forecast Forensics + Digital Profiles: Kill The Catalog???

In the spreadsheet, I build in cells that allow me to change important advertising parameters.

For instance, I use prior mail/holdout test results to estimate what happens to annual retention rates.  Combining this information with new customer acquisition information, I can forecast what would likely happen to a business if a business unit or advertising channel was "killed off".

In this case, I ran a scenario where I killed off the catalog division.  After all, with the growth of Social Media and Mobile, who needs an old school channel like the catalog?  Recall, the catalog portion of the ecosystem wasn't necessarily connected to the rest of the business.

Well, in the case of this business, if we kill off the catalog, we kill off most of the business, don't we?  Within just one year, half of the business disappears ... by year five, 70% of the business is gone.  Take a look at annual retention rates ... around 45% with the catalog, around 25% without it.

In other words, a decent portion of this business is "organic", it will happen without the catalog.  But the overall health of the business is seriously compromised when this advertising channel is taken away from the business.

This is one way that you can combine Forecast Forensics and Digital Profiles to better understand your business.  Contact me if you'd like to have this analysis run for your business!

Dear Catalog CEOs: Next Order Matters

Dear Catalog CEOs:

I've probably mentioned this fact two hundred times across 2,000 blog posts in five years.

• "It's more important to look at what a customer does next than it is to match back or attribute past orders to the advertising that caused the order."

The traditional approach is to take all February online orders, then match them against customers mailed a catalog.  If the customer was mailed a catalog, the order is credited to the catalog.  A profit and loss statement is run against the catalog, determining page count and circulation depth for next year.

By now, we know this method is patently wrong.  When we run mail/holdout tests, we immediately learn that many orders "happen anyway", without the aid of advertising.

Just as important is the concept of the "next order".  I'll give you an example of what folks are doing.  They take customers who ordered via Search in January, and they measure what these customers do if they order again in February ... in other words, where are the orders distributed in February?

• Catalog Marketing = 35%
• E-Mail Marketing = 28%
• Search Marketing = 22%
• Organic Online Orders = 9%
• All Other Channels = 6%

It turns out that this "next order", "forward looking" metric is very helpful ... it clearly helps you understand which channels the customer interacts with.


You could use these percentages in their matchback algorithm ... they credit 35% of a search order back to a catalog, 28% back to e-mail, 22% to search, 9% to e-commerce brand affinity, and 6% to other channels.  You could average channel percentages from the past "x" months and next "x" months as well.


Most important, my projects clearly indicate that history "fades quickly" ... I weight historical dollars by recency, and continually observe that historical dollars have very little weight in determining what will happen in the future.  This is a change from fifteen years ago, when historical dollars carried considerable weight.


So look forward.  We obsess with history, with "matching back", with "attribution".  Look forward, the next order matters.



Hillstrom's Catalog PhD

• Contact me now --- average payback on a project is $1,000,000 in annual profit for a $100,000,000 net sales business.
• Buy the book --- print = click here --- Kindle = click here --- do the work yourself and "reap the rewards" as the trade journals say!

Analytics Sunday: Profit, Profit, Profit

Dear Analytics Experts:

You are really smart people with a bright future in a field growing in importance.  Seriously.  A recent college graduate, with significant mathematical aptitude, asked me which profession she should consider.  I told her "Web Analytics", of course!  If I were twenty-five years younger, this is the path I would choose.


For the past eighteen months, I actively watched or participated with you in conversations on Twitter.  Continually, I am frustrated by the content.  Not the technical content, that's good stuff, there's tons of good advice out there for the aspiring analyst.


I'm frustrated by the fact that you probably don't have a mentor in your company, a mentor who helps you focus on communicating important results to management.


A mentor is that gray-haired individual who shares war stories with you, helping you get around all of the roadblocks that stall a person with great technical skills.  A mentor is somebody in your company who pushes you to be better, from a business and inter-personal standpoint.


Analysts, for instance, are taught to focus on data, they're taught to make data-driven decisions, to create a culture of analytics.


Your Executive team isn't interested in a culture of analytics.  Your Executive team is interested in profit.

Let's focus on a very simple example.  In this example, you work for a typical $100,000,000 e-commerce brand.  Your analytics skills and your A/B tests yield a strategy that increased conversion rates by 8%.

Now, how are you going to communicate this nugget to your Executive Team?

Here's what you should avoid:

• Avoid conversion rate discussions.  Conversion rate discussions don't resonate with members of your Executive Team.  They simply don't understand what eight percent means ... is that good, is that bad, is it eight percent of a five percent conversion rate (yup) ... if it is, they'll say, "well, that's nothing, what about the ninety-four percent of people who aren't converting, what are we doing about them?"  In other words, conversion rate discussions can become mysterious and mindless strolls into a maze of circular psuedo-logic.
• Don't talk about ROI.  ROI has a negative connotation among many folks --- so many vendors and marketing experts have manipulated numbers to make "ROI" look great.  You don't need to manipulate anything, you're smart, you're savvy!
• Avoid geekspeak.  I know you want to show off your A/B test design, you want to prove that your results are statistically significant at a 99% level, you want to show how you used Omniture to make magic happen.  Not one person on your Executive Team cares.  They do, however, care about the outcome.  So focus on the outcome!  And have your ducks in a row ... write up the results as if your Executive Team cares, and pass them out if questions start flying.

Here's what you can do:

• Focus on profit.  I know, you think profit is some boring outcome of an old VisiCalc spreadsheet that is housed on an IBM AT computer with a 20mb hard drive in the finance department.  It isn't.  Profit is the currency of your Executive Team, your shareholders, or your owner.  DO NOT be seduced into all that stuff about "scale" and finding investors like Goldman Sachs or VCs or whatever ... you work in the real world, and in the real world, somebody has to make money or your business doesn't exist.  So focus on profit.  People understand profit.

We're back to the image at the top of this post.  Your test shows that you can increase conversion rates by 8%.


Convert this increase to profit.  Sit down with your finance team, and demand that they give you the tools necessary to calculate profit.  Unless your finance team is doing something illegal, they will give you enough data to make a rough profit calculation --- it's their job.

In our example, an 8% increase in conversion rate results in $2.9 million in incremental, annual profit.

Your company makes $9.0 million in annual profit.

This means that you, a humble analyst, found a solution that increases company profit by 32%.  Thirty-two percent!  You didn't just find an 8% conversion rate increase, you likely made a contribution to profit that is bigger than any other employee in your company!


That's important!

Make two of those contributions, per year, coupled with good interpersonal skills, and you'll be a member of your Executive Team in no time.


You create a data-driven, analytics culture by demonstrating that your work maps directly to company profitability.



Focus on profit.  Start making a difference with your Executive Team.

Forecast Forensics + Digital Profiles: Making Adjustments

If we can predict the Digital Profile that a customer is likely to migrate to, and if we can predict how much a customer is likely to spend, then we can make adjustments to our predictions, allowing us to see how a changing business might lead to a changed business in the future.

I like to create a tab in a worksheet that allows me to make changes to the future trajectory of the business.  My tab looks something like this:

Take a look at the image above.  In the bottom half of the worksheet, I change how customer acquisition is likely to evolve, in the future.  If a Digital Profile shows a historical increase of, say, 15%, then I might type in a factor of 1.15, to reflect a future 15% increase.

Take a look at the "Mobile Mavens" row ... I've jacked that one way up, to account for projected increases in the channel.  Conversely, I ratcheted down the "Pricey Website Preference" Digital Profile, in order to account for a likely shift in customer behavior, going forward.

If we can predict where a customer is likely to migrate to ... and if we can predict how much that customer might spend ... and if we can make adjustments, allowing us to consider different possible outcomes ... well, then, we've got something interesting, don't we?

Next week, we'll explore some of the possible outcomes offered by the combination of Forecast Forensics and Digital Profiles.

FREE ANALYSIS OPPORTUNITY: Chronic Cart!

Dear MineThatData Nation:

For the past five years, you've enjoyed a lot of free information (more than most vendors would provide), and many of you paid a reasonable sum to me, in exchange for big-time profit opportunities.

Now, I'd like to present an opportunity where you can help me ... so that I can help you ... contact me if you're interested by clicking here.

Shopping Cart Abandonment

The experts will tell you that Shopping Cart Abandonment is one of the biggest issues plaguing the e-commerce industry.  Cart Abandonment programs have been proven to recapture business that may otherwise be lost.  Heck, we all know, intuitively, that if we can eliminate customer indecision, we can grow sales and profit, right?

We use shopping cart abandonment percentages to gauge how effective we are at converting a customer to a purchase.  This is a good metric, in most cases --- and given how easy it is to calculate, it is a great metric.

But we can do better.


A New Product:  Chronic Cart Analysis


I believe I am close to finalizing an algorithm that creates a metric, called "Chronic Cart".  This metric assesses shopping cart abandonment issues at a customer level.  The metric, to be determined (a percentage, a grade, an index), assesses customer issues that cloud shopping cart abandonment.  

Here are examples of "chronic cart" problems.

• A high overall abandonment rate.
• Customer places item in cart, does not visit again.
• All customer segments abandon carts at high rates, across the board.
• Customer purchases after abandoning shopping cart and then receives trigger-based e-mail with discount opportunity.
• Out of 100,000 visitors, only 1,000 visitors abandon carts on a frequent basis, resulting in an over-inflated cart abandonment rate.

Here is an example of healthy customer behavior, an example not subject to "chronic cart".

• Customer visits five times, puts item in cart on visit three, purchases on visit five.

The algorithm and index that I'm close to finalizing accounts for these issues, describing for you if you have "chronic cart" problems.  The algorithm/index reviews customer behavior over time, in order to calculate the index.

Here's Where I Need Your Help!

I am looking for two volunteers to forward online data from your web analytics platform to me for a free analysis.  In exchange, I will run the "Chronic Cart" analysis for you.  I promise to not publicly disclose your proprietary results.  You will receive a brief report, outlining the algorithm, index, and project outcome.  If you have a "Chronic Cart" problem, you will receive a file of "cookies" or household_ids that possess a "Chronic Cart" problem, so that you may take any desired marketing action with these customers.

Interested In A Free Analysis?

Contact me right now!  I'm confident we'll get a couple of volunteers in short order, so get your request in now --- I fully anticipate Chronic Cart to become a popular product in the near future that folks will be charged for, so this is a great opportunity!

Forecast Forensics + Digital Profiles: Demand Forecast

If we can predict the Digital Profiles that a customer will belong to in the future, then we can easily predict key sales metrics, right?

For instance, if we have 1,000 customers, and 40% of the customers purchase again, and 10% of customers migrate to the Mobile Mavens segment, and each Mobile Mavens customer spends $249.55, well, then we can easily calculate what is likely to happen.

• 1,000 * 0.40 * 0.10 * $249.55 = $9,982.

If I can do this for one segment / Digital Profile, I can do it for any Digital Profile.  And, I can do it for new customers as well, projecting how many new customers I'm likely to obtain by Digital Profile.

So, we can predict the Digital Profiles a customer is likely to migrate to, and we can predict the amount a customer is likely to spend.

Seems like that should be pretty useful, if you're a CEO, CFO, or a Chief Marketing Officer, right?

Forecast Forensics + Digital Profiles: Forecasting

Today is a transition day ... we're moving away from the segmentation part of this series, moving to the forecasting part of this series.

Recall, Forecast Forensics is all about using conditional probabilities to illustrate how customers are likely to migrate between segments.

In other words, if we have a customer that belongs in the "Web Masters" segment, we can calculate how likely that customer is to purchase again next year ... and if the customer purchases again, we can calculate the probability of a customer migrating to any one of the sixteen Digital Profiles.


And if we can do that for one year, well, we can do that out into infinity, or at least for five years, right?


And if we can do that, well, then we can literally "see" what the future might look like.


So take a look at the image.  This is the forecasted count of twelve-month buyers by Digital Profile, for each of the next five years.


Tell me what you see.