Optimizing the Interface (Band 3)

After the lessons we learned in the Band 1 and 2 Optimizations (see Band diagram) for HSN, we were able to reallocate that budget to invest in Band 3 – Optimizing the Interface. We realized during the previous Optimizations we were already getting a tremendous amount of traffic through channel surfer / clickers, but this traffic was not “converting”. In other words, we really needed to Optimize the “Landing Page” for this existing audience – the TV show itself.

Don’t suppose the above scenario sounds at all familiar to the web analytics folks out there – you know, “more traffic, any traffic” is the answer? Oops, maybe not, what about higher conversion?

I won’t bother providing the Band 3 example for the web; you all know what Optimizing Landing pages / web sites is about, or can certainly find that info elsewhere. However, you might find the Optimization of a TV shopping channel interesting…

A little background first.

You probably don’t know how interactive HSN was, at least back then; the show just looks like somebody talking about products. What you don’t know is there is a huge multi-screen war room setup steering the ship. Every competitor monitored live in real time, along with all major TV networks.

Somebody is burning a flag on the steps of the Supreme Court, and every TV network is covering it? Start selling American flags. Folks on the Today Show say “blue is the new black”? Put on blue clothing all day. Competitor selling a product we have in inventory? Put the same product up at a lower price. The war room was a wild place – and especially so with 3 of these networks, all live, going full blast at the same time, 24 hours a day!

So first, HSN was interactive with the “environment”, followed trends rather than trying to Market against them. Customer demand (sales per minute) declared what to put on a show; if they wanted to buy it, we wanted to sell it.  If they didn’t want to buy it, we tried to get rid of it by cutting price.

Sound familiar? A bit like keyword phrase / Search analysis, perhaps?

HSN was also interactive with customers. There are the callers you hear on TV, of course. What you probably don’t know is all TV Shopping customers are absolutely chock full of comments about the products and presentations, and these are tracked and reported on the very next day. Every day we had a sales meeting to review the prior day – how could we have done better – and as part of this meeting, we reviewed customer comments from the previous day, and tried to take action on these comments.

To put some context around these daily sales meetings, the President of the company ran the show, and? the attendees were all VP or above. It was the most important meeting anybody had.

Every day.

The kind of meeting a company might start having to really leverage the power of Interactivity and the output of Social Media, perhaps? We’re talking about close to a billion dollar in sales company at this time, not a start up. We sold management on the importance of an Analytical Business Culture. You can too.

So, we have this pretty interactive system despite the one-way technology of TV, and Marketing now has budget to try and Optimize it. What would you do, how would you improve this?

Here’s where we ended up, in some cases after years of working through the testing:

1. Feed customer comments from the call center to the show host live. We invented a system to send customer comments in real time to the host of the show. They scrolled on a screen as the host was pitching the item, and could be tremendously effective in improving sales. When I heard of people Twittering about speakers at conferences, that reminded me very much of this functionality.

2. Capture successful interactions in the product database for review later. After the show, the hosts wouldenter notes on which customer comments “worked” for specific products and allowed them to sell at a higher velocity. These “best of” ideas could then be reviewed by any host who would be selling the product.

3. Make real time testing actionable in the future. There was a lot of testing in real time besides the host pitch- changing displays, lighting, formats, etc. However, there was no “system” for capturing these ideas, it was all “gut” and learned by practice. So we did the same as with the comments above – got these “tips and tricks” loaded into the product database so show prep folks and producers could get a leg up using the experience of others.

Not exactly A / B testing, but there’s no A / B in live TV, you have to look at historical performance in the same Dayparts – and we did. More on this in the next initiative…

4. Program to the audience. The very difficult thing with selling product on TV Shopping versus the web, of course, is that you are selling one product at a time. How do you Optimize that?

We studied the buying patterns – rowing machines sell best on Sunday mornings, Fashion on Tuesday nights, kitchen stuff at 3 PM weekday afternoons. In other words, Dayparting.

The audiences change throughout the day, and they have a different composition on Weekdays versus Weekends. Optimize the product presentation to the audience. Using Nielsen ratings for this Optimization would be a “Push” approach. By using buying behavior, we were using a “Pull” approach – sell them what they have told us they want to buy using their actual behavior. By the hour.

5. Implement the best performance metrics and controls. In the early days, Gross Sales per Minute was everything. Analysis proved that return rate and price had a very high correlation, which in this case was indeed causation – and we proved this with price testing.

So we came up with the idea of an “average price per show” which would minimize return rates based on Daypart. Literally, we found that the time of day a high priced product was sold influenced the return rate. This was an improvement, but the system could be gamed by smart hosts looking to drive bonuses.

So, we converted from Gross Sales per Minute targets to Net Sales per Minute targets – net of returns, actual if available or forecast based on like item history, all in “real time”. But in the case of some high volume categories with thin margins, this was not Optimal either. So ultimately we made it to Net Gross Margin per Minute as the final sales metric. Very similar to moving from ROAS to ROMI.

This series of “Business Rules” linking price, return rates, margins, and time of day basically ensured that even in a completely dynamic, real time environment, the whole system was Optimized for profit.

Do you wonder if a Dayparting approach could improve your commerce profits on the web? Based on what I’ve seen, I’m pretty sure it would, as long as the costs of implementation are not prohibitive.


Look at these 5 ideas above. Ask yourself if any of this is what you think of as “Marketing”. These ideas are in reality a series of very intensive IT projects (on MainFrames!) involving telecommunications, core systems architecture, programming / code, and more.? Why was Marketing involved?? None of?this is about?”Creating Ads”.

Marketing was involved because we insisted on being involved. In an Interactive system, you can’t Optimize Marketing until you Optimize the systems delivering the Interactivity. Those of you Optimizing web sites are quite familiar with this concept by now.

It’s about removing Friction.

Said another way, these ideas were all directly related to Voice of the Customer, Customer Experience, and long-term Customer Satisfaction. The better HSN was on these issues, the more successful Marketing programs would be in terms of increasing Customer Value.

Engineers are smart people, but if they knew everything, then why would web sites, VRU’s, or countless other machine-human interfaces have to be Optimized? For best results, you really need both brains working together on these systems from the beginning. iPod / iTunes / Mac?is the most recent example – the Marketing Strategy is built into the product from the very beginning.

Failing this ideal world (and many companies are not there yet, for sure) analysts must get these Interactive systems Optimized before addressing the next challenge – Optimizing Customers – which begins in Band 4. Otherwise, Marketing success will be?undermined by?Friction instead of driving incremental Customer Value.

For more on Measuring and Managing Friction, See Chapter 5 in the book sample PDF.

Any questions on the approaches / testing / decisions above?

(A post by post index of this Marketing Bands Series is here.)

2 thoughts on “Optimizing the Interface (Band 3)

  1. *Wow*! Just wow!

    Heh. You’ve mentioned HSN a few times over the many months of my reading but it was only a few posts back that I finally … realised(?) what it was. Shoulda googled. Didn’t. :-)

    But this also explains a much older discussion we’ve had Jim where you referenced to Real Time Web Analytics. Hmmm. Perhaps there is a case for the bigger type of shopping site (a la the Amazons) to do Real Time. I’d guess it all gets down to ROI in some way shape or form. Hmmm x 3.

    We don’t really have an HSN here. The 2am period (the joys of having a baby few years back introduced me to ’em) is when they seem to come on.

    Fascinating stuff! Very!

    – Steve

  2. I consulted to a software start-up in 1999 – 2000 that built a commerce-oriented streaming analytics “cockpit” based on the TV Shopping model. Was a very cool app but they ran out of money before anybody really cared about web analytics (other than a few of us).

    Ultimately, the question is whether you think a human can do a better job of optimizing a commerce web site than a machine can – the standard personalization, customization, people who bought that also bought this stuff. Blah..

    I think people will discover over the longer run the “display management” stuff above is just the tip of the iceberg. You can have all that, but it’s low level, almost background noise stuff compared to what a skilled “daypart manager” could do with a real time system:

    1. Screen showing a “most popular searches” stream of what people are searching for in real time across the engines. This would be the equivalent of us monitoring all the TV Networks at HSN.

    2. Aggregated traffic display – where are people “bunching” on the site? This would be the equivalent of network / call center monitoring at HSN.

    3. Content management – the ability to push certain content to specific visitor segments, based either on behavior or location. This would be the equivalent of the show manager at HSN talking to the host and the product display people about trying different words or lighting, etc.

    4. An unbelivably cool “heads up” type of cockpit to run it from

    5. A back end that tracked what the pilot did and what the machines did, so there could be “post show” analysis – did the human beat the machines?

    FYI – at HSN we found out a human, on average, can take only about 3 hours of real time merchandizing optimization before they needed to take at least 3 hours off. There were an exceptional few that could pull double shifts…

    Rather like a gigantic video game with real money on the line, no?

    Seriously, these folks would come out of the war room absolutely exhausted – and then start going over the post-stats for the show, looking for ways to improve.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.