This is a bit of rant against robotic thinking, best practices, and testing as the savior of all things web.  This after having so many conversations lately with people at all levels of web analytics who are infatuated with the idea that robots / software and “best practices” are the answer to everything web marketing. 

To be clear, I don’t have anything against the poor robots or testing – it’s the people using them.

All the way back in 2000, Bryan Eisenberg and I wrote the Marketer’s Guide to E-Metrics – 22 Benchmarks because nobody was measuring or testing anything, and that was silly, especially when it was so easy to do.   Now, it seems web analytics has taken that mantra and run all the way to the other side with it – testing is Strategy, and Marketing is whatever the robots say it should be after the tests are done.

Yes, web marketing seems to be going IT-centric again.  Worked out well last time, didn’t it?

Here’s the bottom line: I have no doubt you can improve a faulty execution with a lot of multivariate testing, but the real question is this: if the execution is Strategically flawed, will you ever get where it is you want to go? 

I think not.

I’m sure you are convinced your Strategy is on target, based on conventional web commerce wisdom.  The following is a bit of unconventional web commerce wisdom for you to consider when you sit down around the table with your robots.

——–

The Lab Store – my wife’s pure online commerce business where I am Chief Product Assembler and also do a lot of marketing testing on the customer base – services the exotic pet customer.

It’s a very odd experience going to the pet trade shows for this biz to review merchandise and make purchases, on many levels.  The root of this odd-ness can be summed up this way: we buy narrow and deep, and most everybody else in the pet business – which means retail stores, and many online stores – buy broad and shallow.

We work with one of the largest pet supplies distributors on the East Coast.  At their show, we get a bit of a discount if we place orders directly with the vendors, which are then managed by the distributor.

As we place our order with this one vendor, he asks, “Did you know this order is nearly 40% of the entire annual volume we do on these SKU’s with the distributor?”  We chuckle, hearing this all the time.  “Yea, well we do sell a lot of them” is basically the only thing we can say.

Another common conversation goes something like this: “Are you sure you want that many of this SKU?  No offense, but this is one of our slowest moving products, and I just wanted to be sure the quantity was correct.”  And our response is always something like, “Really?  That’s one of our best sellers, it’s a great product.”

Narrow and deep.  We only sell what the customer buys – a little trick I learned at HSN (not sure how they do it now).

Kind of makes sense though, doesn’t it?  “Customer-centric”, as they say.  And we are not afraid to completely re-build / re-brand any product we think has potential but has simply not been marketed correctly.  Or to take a “poor selling” product and change the intended use of it, turning it into a best seller. 

In fact, we routinely rip off all the packaging a product comes with and create our own packaging and new name for the product.  Any online retailer who has done a great job marketing a product only to find it appearing in a competitor’s store at a lower price should understand exactly why we do this.  We absolutely love this kind of product.

In many cases, multi-variate testing can improve the sales of any product, but can it turn a dog into a best-seller by completely rethinking it?  Nope, sorry.  Are there any “best practices” a human can follow to repackage a product successfully?

What, are you kidding?

Most pet stores stock a broad range of SKU’s and buy only a few units deep on each.  We buy only a few SKU’s and buy them as deep as makes economic sense – based on volume discounts, weight to value ratio (freight cost from distributor), storage considerations (is the product large relative to value) and so forth.

In other words, everything we do in the Lab Store is really based not on Sales, but on Productivity – how can we generate the greatest amount of profit for the least amount of time, money, and effort?  I realize this approach does not square with conventional wisdom, but the Objective of the store was to replace 1 income (my wife’s) with the least amount of effort possible.  If that is the Objective, then the Strategy is Productivity, not a focus on Sales.

For example, we turn our entire inventory 21.8 times a year.  I’m pretty sure most small (micro?) online retailers in our category ($1 – $5 million in annual sales) don’t care about that stat, but I’m also sure a few of the offliners out there are feeling their jaws hit the desk.  Most of them turn at 5 – 6 times with the really good ones at 10 – 12.  This stat is one of the most important in retail, it’s an “inventory productivity” thing.  And it also points out the economic difference between a narrow-deep and broad-shallow merchandising strategy. 

I know this is going to sound insane to a lot of small online retailers, but in our online store, you do not find a lot of variety, and this is intentional.  What you find is the single very best product for each need a customer has.  And most all except commodity products are priced that way – as the super-premium product in the category.  We carry the commodity stuff not because we want to, but because customers want access to it when they order from us.  It’s a Service decision, not a Product decision.

When customers ask, “Why don’t you have more variety?” we simply tell them we don’t see a need to offer anything but the best product for each need they have. “But don’t you have any cheaper ones?”

Notice, “variety” here is a code word for price.

“No, we don’t have cheaper ones.  You can find cheaper versions on eBay.  Or try a shopping search engine, if you are shopping only on price.  If you want products we have personally tested, are vet-certified for the particular exotic pet you are dealing with, and are absolutely guaranteed to satisfy your needs, we welcome your purchase.”

As a result, we clearly narrow the ability to attract a wide audience.  But we don’t want a wide audience.  We want an audience and a business we can easily defend against the constant price wars that are a reality of the web.  We knew that would be the evolution, and designed the business that way.  We want a Productive audience, one with high demand for the best, and a low Sales to Service ratio. 

Do you know your Sales to Service ratio (orders / service inquiries) and how to optimize it?  Do your robots?

If we did an on-site survey, I’m sure a lot of casual visitors would complain the store “lacks variety” and is “over-priced”.  That we’re not being customer-centric, don’t you know.  But we are, for the customer we want – she wants a high degree of quality, professional one-to-one advice, extremely fast and accurate execution, all with no hassles.  The rest of these high maintenance, high variable cost “customers” who are buying single items on price and suck the life out of the business if you let them can go to hell.  Really.  Those shoppers looking for value, which we deliver through aggressive product bundling and flat rate shipping, find it in our store.  And those are the customers we want.

It takes nearly as much effort to process, pick, pack, ship, and service a $40 order as it does a $140 order.  Given that, we prefer to drive higher value orders, and all our marketing is set up to do just that.  We actively discourage low value orders by using flat-rate shipping.  It’s that Productivity thing again; it’s the Strategy, and the store was built from the beginning with that idea in mind. 

Please sir, can you multivariate test that idea for me?

For example, we don’t have a search engine on the site, because we want to force (sorry, I mean “encourage”) customers to look at all our products, and not to cherry-pick the product they originally came to buy.  We specifically and intentionally designed the navigation that way.  And since we have less than 80 products by design, it’s easy for customers to review every product we have very quickly. 

The end idea is ease of use by the customer.  We do it by having fewer products and really smart navigation, not by substituting technology to fix a broken execution.

I know this also sounds insane given “best practices“, but you have to realize that a lot of these “best practices” tests related to on-site search were done on sites with terrible navigation, screwed up product assortments, and lousy merchandising.  In that case, I’m pretty sure a search engine increases conversion.  In our case, a search engine did not increase conversion, but it surely did lower Average Order Value. 

C’mon, do you think I didn’t test it?  Productivity again.

Can we get a multivariate test to confirm search improves conversion on poorly merchandised web sites?  Or can we just look at the site and know it will be true because the nav sucks?

One of things we do very aggressively is cross-merchandise, bundle, and package.  We do it precisely and intentionally within the navigation, which is why a search engine doesn’t help us.  Our approach is not an automated system, it’s a carefully considered marketing decision based on known behaviors.  People who buy this will be interested in this.  We don’t need a computer to do that for us, all we need is intimate knowledge of the customer and some merchandising savvy.  This bundling and packaging doesn’t change, it uses the same format over and over (so the customer gets used to it) and the bundles don’t change dynamically, they are the same for every customer.

Could we have a more sophisticated system?  Sure, but at what cost?  Given we already know what drives buying behavior, we understand pricing theory, we attract a specific audience, and we know what they want, why do we need a machine?  What would the incremental benefit be relative to the cost?

The store itself was built with a $70 copy of FrontPage.  Our monthly costs for hosting and the MIVA Merchant shopping cart (which is all but hidden from the customer except for checkout) is $40 a month.  When the package volume got to 15 boxes a day, we bought a back-end inventory / pack / ship label processing package for $500.  That’s it, that is all the infrastructure there is.  No employees.

Does the store look “slick”?  No.  Doesn’t need to.  Instead, it oozes personality from every pore – the product copy, the newsletter e-mails (which have no offers in them, we never discount or have a sale), the customer service communications – they all speak with one voice.  People adore the site, they think it’s the easiest to use site in the entire category.  People anticipate the newsletter and actually complain when they perceive it to be “late”.

Had any complaints recently from customers about not getting the newsletter?  How about the opposite?

The average product description on our site has over 500 words – even for the most mundane products.  We tell you absolutely everything there is to know about a product.  I noted that the big thing on e-commerce retailers “to do” list for 2008 is improve product descriptions.  Did they need a multivariate test to tell them that?

We have a no questions asked returns policy without a restocking fee.  We can do this because we anticipate product problems by extensively reviewing every product..  If a product is difficult to assemble, we assemble it before we ship.  If the assembly instructions suck, we re-write them and include them with the product.  Sounds like a lot of effort, until you find out we have a return rate of 3% on units and 1% on dollars.  Yea, it’s that Productivity thing again…

What is missing in web analytics today, with all due respect to both sides, is people who understand both the Marketing and the Technology aspects of web Behavior and Analytics.  Optimization is in the middle, not at the extremes.

Following “Best Practices” leads to commodity positioning, as everybody plays Monkey-See Monkey-Do (MSMD).  The constant benchmarking that is part of the IT culture is simply wrong-headed for Marketing; why does it matter what the other guys do, especially if they do a crappy job?  Do you take pride in the fact you benchmark better than some of the crappiest folks on the planet?  That your site / performance sucks less than theirs, but still sucks?

Do you have a Marketing Strategy, and do you execute in line with it, down through every fiber of the company?  Substituting brute force robotics or worship of MSMD best practices will never replace a great Strategy.  If you are at the point where all you can do is test things to death, perhaps you need to rethink your Strategy instead.

Please understand, I am not saying you should run your commerce operation like we do.  I’m just saying there are other, highly successful ways to do it and blindly following Best Practices and robotic testing – for any web operation, commerce or not – should be reconsidered.