Amazon has multiple teams developing systems for Seller Fulfilled Prime. You’d think these teams would talk to one another. You’d be wrong.

A Story About Maps

Amazon Prime is free two-day delivery. Amazon sellers can get their products on Prime by using FBA (Fulfillment by Amazon). Your products sit in an Amazon warehouse, ready to ship. At Customized Girl, we print on demand. Youc an see how custom printing is not compatible with FBA. Our items could not be on Prime.

That was true until we were invited to join Seller Fulfilled Prime (SFP) in 2016.

To offer Prime on our products we would have to take the cost of two-day shipping and bake it into the product price. But nationwide two-day shipping is very expensive. The price markup would be much too high.

Luckily, Amazon had a solution for us. We could offer Prime to only those parts of the country that were within range of two-day FedEx Ground shipping. Well, almost.

Here’s how it worked. We looked at the two-day delivery region on our FedEx Ground map. Then we looked at our Amazon Shipping Template. We tried our best to match them up.

To do this, we dove into a sea of drop down menus. Florida was split up into three different regions. Which one matched our FedEx Ground map? Well, none of them. We had to guess. We went with Northwest Florida and hoped for the best.

As I was making a decision on each and every drop down menu in the shipping template, I kept thinking: “Why are they making me do this? They know the location of our warehouse. They know exactly which streets are two days away via FedEx Ground. Just let us enter our warehouse address and automate it!”

Three Years Later

Three years later, they did exactly that. In 2019 Amazon launched Automated Shipping Rules, and it was just as I described.

You enter your warehouse address. You enter exactly which carriers you want to use (we chose FedEx Ground but NOT FedEx 2Day). Bing bang boom. Amazon displays the Prime option to only those customers who live within a FedEx Ground two-day shipping range!

This worked just fine until a week before Christmas. Amazon was not satisfied with the performance of FedEx Ground. We would no longer see Ground as an option when we go to actually ship the order.

The method for actually shipping your Seller Fulfilled Prime orders is called “Buy Shipping”.

You can use Buy Shipping directly in the Seller Central interface, or you can access it via the API. We needed to be able to mark that package as shipped in our own backend, so we built a sophisticated API integration with barcode scanners and everything.

For a variety of reasons, we had to disable Prime on December 16th. So this particular FedEx Ground outage did not impact us. We were lucky. But we wanted to get back on Prime soon and we had to wait.

Finally, we got an email in mid-January saying that FedEx Ground would be available again. We were very happy to re-enable Prime.

Two Maps

During the peak holiday season, shipping prices can fluctuate in painful ways. In December, I was worried about how much Buy Shipping was charging us for each order, and I was checking that number regularly. After peak, I relaxed.

If I had checked more often in late January, I would have been alarmed. It usually costs us around $9 per order to ship Prime. Our new average was closer to $16 per order.

What was happening?! These orders are all supposed to ship Ground for $9. But most of our orders were being shipped FedEx 2Day Express for $20.

But this didn’t make sense. We were using Automated Shipping Rules. We had only enabled certain carrier options.

Was the FedEx 2Day checkbox accidentally checked? No.

That’s when I realized: “Oh my god. They’re using two maps.”

The entire point of Automated Shipping Rules is to accurately display Prime shipping as an option to the customer based on the location of that customer. They do this using the options that we enable inside Seller Central.

But someone at Amazon HQ was angry at FedEx. They were paying close attention to certain routes and realizing that Ground only delivered on time 95% of the time. The threshold was 97%. (I am guessing here.)

So they built a new map. In theory, Ground shipping from Columbus to Manhattan was going to be two days. Not in the new map. If you needed to get it there in two days, you had to upgrade to FedEx 2Day Express.

Did they apply this map to the Automated Shipping Rules system? No.

Did they apply this map to the Buy Shipping system? Yes.

So Amazon was using one map to determine which customers could get Prime shipping.

And they were using a very different map when it came time for us to ship. Most of our orders were now going FedEx 2Day.

We disabled Prime. We filed a support ticket. We emailed everyone we could. I yelled into the void.

Please stop using two maps! We’re fine if you want to limit the number of customers who can place Prime orders, just please, use one map!

Finally, I reached out to my contact at CNBC. They reached out to Amazon PR.

The next day, all SFP sellers got a new email: FedEx Ground performance has significantly improved! You’ll see even more access to FedEx Ground inside Buy Shipping!

Remember, we had already gotten an email saying FedEx Ground was available again. Should we trust them this time?

That’s the beauty of being an Amazon seller. You don’t have to think about it too hard because you don’t have a choice.

We reenabled Prime on a Wednesday. Things looked good. The FedEx 2Day shipments disappeared.

First Breakdown

Then Friday came. My production manager came and said “we can’t access the Buy Shipping API”. I tested it in Seller Central. The error message said:

Services are not currently available. Please try again later.

I checked Amazon’s Seller Forums. Everyone was freaking out.

You can always use another method to go around Buy Shipping, but if you do, you will be immediately suspended from Prime.

There was one Amazon employee trying to answer questions on the forum. By 2 PM, she finally had an answer: just go outside of Buy Shipping. We won’t penalize you this time.

Of course the FedEx pick-up time for most sellers (including us) is around 5 PM. That gives us three hours. We wrote new code to go around the API for these orders. We deployed that update at 2:29 PM. At 2:30 PM, Buy Shipping was working again.

Rats! Revert! We reverted the code and barely managed to get all orders shipped in 2.5 hours.

The system felt very shaky. Again, we have to ask ourselves: do we disable Prime?

Valentine’s Day was a week away. One of our most important days of the year. It was my call. I enabled Prime.

Second Breakdown

I was nervous. I checked Seller Central all weekend. Clicking on orders. Clicking Buy Shipping. Checking those options. Was FedEx Ground appearing? Yes. Good… until, no.

On Sunday, there was no Ground. Buy Shipping was back to charging us $20 for FedEx 2Day. Oh no.

Again, I start filing support tickets. I start emailing. From the void, no one could hear my cries.

Except I did get a reply on my support ticket. A quick one! They said: you have to add a phone number to your warehouse location in Seller Central to see FedEx Ground.

What? How could those two things possibly be related? That form didn’t even have a field for a phone number when I first filled it out.

Also, you can’t just edit a warehouse location. You have to add a new, fake warehouse location, change your shipping templates to use the fake location, and then your original location can be edited. I did that. I added the phone number. Went back to update the shipping templates.

It worked. Somehow, it worked. FedEx Ground appeared in Buy Shipping.

Third Breakdown

On Monday I got to my desk and checked our shipping cost data. We had shipped 10 Prime orders. Every single one cost $20. They were all going FedEx 2Day.

No! I ran back to the ship station. Stop shipping!

This could not be happening. Over the weekend, we had accumulated hundreds of Prime orders for Valentines Day.

I tried to ship one in the Seller Central interface. I saw FedEx Ground. The price is reasonable. I chose it. Everything worked.

The API was returning one result and the Seller Central interface was returning another.

Unfortunately, we weren’t set up to ship via the Seller Central interface. We could do it, but each order took 5x as long. As usual, we didn’t have a choice.

I set up a new ship station in production. All hands were on deck. We shipped until 8:00 PM, and one minute before the final emergency FedEx pickup arrived, we printed the last label. We did it.

I immediately disabled Prime and began yelling into the void again. Support tickets. Emails. Phone calls. Why was the API different from Seller Central?

I only got one reply: “it’s supposed to work like that”.

This obviously could not be true. I re-opened the tickets. I heard the same answer. This went on for two weeks. My friend Ben sent me a LinkedIn profile of someone on the Buy Shipping dev team at Amazon.

I messaged them. They replied! We exchanged emails. They called me on the phone! I spoke to a human! He was nice.

He did not know what happened, but he had a suggestion. In the API, we were returning data for the “Needs By Date” column. But that field was optional. Why not leave it blank?

That’s what we did, and it worked.

In Conclusion

Remember that Friday when Buy Shipping was broken for every seller? They were working on that code. They must have changed a setting related to the Needs By Date in the API. We had not changed anything on our end. But regardless, it is fixed.

Even better, we saw that FedEx One Rate was now enabled inside Buy Shipping (as long as you enter the exact right dimensions).

We’re using One Rate now. If we had this in place 4 months ago, we could have avoided almost all of these issues.

I wish I could say that this was my only kafkaesque experience as an Amazon seller. Unfortunately, I have four more stories just like it. Those are for another day.

As far as I know, Amazon is still using two maps. If you ship a product that is too large for One Rate packaging, and if you’re using Seller Fulfilled Prime, then I say to you: good luck.

If you get tired of yelling into the void, tweet at me.

I published my last essay on this website five years ago. What have I been doing all this time? Having kids!

I’m now the proud Dad of a four year old boy and a two year old girl. They’re very cute and very defiant. We’re working on that.

I launched this website to share my work and to point out weird patterns.

Though I haven’t published anything new in these past several years, I have been secretly updating the Hand Dies First post. I’ve even been able to add a subgenre to the list: time travel movies!

As for my work: we’ve A/B tested our Custom Shirts page many more times. I’ll be honest, the version I featured in my original blog post was hard to beat. It took several attempts.

Facebook

We’ve also been doing a lot more Facebook advertising, which renders obsolete my post titled Why Facebook Fails. In my defense, the Facebook Ads platform has evolved in leaps and bounds over the past five years. Remarketing lists. Lookalike audiences. Dark posts. A million types of targeting.

Facebook is a still a dangerous platform for beginners. If you run a decent ecommerce site and if you install the Facebook tracking pixel and if you run any ad targeted toward conversions… then you will see an unbelievable conversion rate. Literally. I mean you should not believe it.

Your first impulse might be to congratulate yourself. But can you really trust a $2 CPA? Is that even possible?

If you’ve installed the tracking pixel, Facebook knows exactly who is about to checkout on your site. If any of those people are also inside the target audience for your ad, then all Facebook has to do is “show” the ad in that user’s feed. By default, Facebook will count “view-through conversions”.

At first glance, it looks like your ad performance is beautiful. But dig deeper. Did those users even see your ad? Or did they just scroll right past it? Either way, Facebook will claim credit for that conversion when they eventually checkout. They were checking out anyway and as I said earlier, Facebook knows that.

To Facebook’s credit, they also give you the tools to combat this. Just like adding negative keywords to your Adwords campaigns, so too must you exclude certain audiences from your Facebook campaigns. For example, we have audiences of 7 day, 30 day, 90 day, and 180 day visitors. If we’re targeting cold or warm traffic, we excluded these audiences from our ads, which unskews our conversion data.

Amazon

We knew that more and more customers were starting their shopping experience on Amazon rather than Google. But we also knew that it would take a lot of development resources to sell directly on Amazon. After all, we print on demand. All designs are created within our interactive design center. That design center is a fundamental part of our organization and it doesn’t exist on Amazon.

Do you remember Amazon Product Ads? You could upload your product feed to Amazon, your products would appear in search results, and you would pay per click, just like Google. The click would actually take the customer away from Amazon and onto your site.

In early 2015, our Amazon Product Ads were buggy. Half our feed would be rejected. Traffic would drop off completely. We were getting an error message that said:

We believe that you have violated our policies by attempting to redirect Amazon customers to another website.

Uhhh, yes. That is exactly what we were doing. Because that is the entire point of your platform, Amazon Product Ads.

The writing was on the wall. Amazon wanted to kill Product Ads, but they didn’t want to say that they were killing Product Ads.

The only choice was to sell on Amazon directly. We couldn’t wait for the development resources, so we built our own specialized Amazon product feeds by hand and we placed every order by hand. Within a few months, it was clear that this was a successful channel. We got the development resources. We started automating all processes.

Sidenote: when I say “we”, I mean my team. The Amazon project succeeded due to the hard work of a lot of people, but mostly Jason Kienbaum.

Unfortunately, it wasn’t all sunshine and rainbows.

In the spring of 2016, the counterfeiters arrived. They attached themselves to our listings. They undercut our prices. When an order was placed, they would take our thumbnail image, blow it up, use that as the print file, make a terrible product, and ship it out.

All of our designs were created by our designers in-house. We printed everything in-house. When we uploaded our feeds to Amazon, the brand column was filled with our brands: Customized Girl, Bridal Party Tees, and FunnyShirts.org.

Now there were terrible reviews popping up on our products, under our brand. They said things like:

• Shirt is from some Chinese company and the writing is blurry.
• I opened it and now my entire office smells like vinegar.
• I never got anything.

Re-read that last bullet point. Some of these “counterfeiters” weren’t even counterfeiting. They were literally just taking your money and not shipping anything. If you visited their seller profile, it would just be ten pages of reviews saying “I placed my order and I got a fake tracking number. Never received the package.”

We contacted our Amazon reps. Seems like an open and shut case, no? Sadly, no.

Somehow, it was our word against the “counterfeiters”. For Amazon, it was possible that these other sellers were buying from us, and re-selling our product.

But we don’t sell wholesale. No one is buying our product at retail for $25 and selling it on Amazon for $10. These other sellers provided zero proof that they had actually placed an order from us.

Instead, the burden of proof was on us. We had to buy the counterfeit goods and, in the event that we actually received the product, we had to take photos, document all the ways in which the item was clearly a knock-off, and send the reports to Amazon’s copyright department.

We tried doing this. It didn’t work. Here is why. By the time we even got the goods, that seller was already banned. Every time. Amazon has strict rules about “on time delivery percentages” and “customer satisfaction ratings”. These counterfeiters were producing garbage and their customers hated them. After all, they’d been duped.

In theory, this would be great for us. Counterfeiters were getting busted! Yay! But it didn’t matter. The sellers were using software to automate the creation of new seller accounts. The seller names were essentially a random string of alpha-numeric digits. They had thousands of accounts. As soon as a hundred got killed, no big deal, they moved on to the next hundred.

With the burden of proof on us, it was impossible to keep pace.

There had to be another way. At this point, we were selling custom products on Amazon. You could enter your own text, and we would print it and ship it. With custom products, Amazon recognizes that you might be the only company that can produce the customized print you promise. Other sellers cannot attach to those listings. So we knew Amazon had methods of preventing these attachments.

It was clear that our product pages should be ours and ours alone. No other sellers should be able to attach to our listings.

I talked to our reps. I made it clear that I was going to email them every week with one of two potential outcomes:

A) Our account would be freed from other sellers.

B) I would die.

So I emailed them. Every week. For months and months. And eventually… it worked. We found someone to help us. And that person maybe just saved the livelihoods of the 85 people who work at eRetailing.

Storefronts

One of my early essays on this site was titled Deep Thoughts on the Print-on-Demand Industry.

In that post, I compared CafePress and Zazzle to what we were doing at eRetailing.

CafePress and Zazzle were the leaders of the POD storefront industry. On those sites, anyone could open a store, post designs for sale, and earn a royalty when those designs sold.

They excelled at building a platform. But they did not excel at customization. Those storefront designs were basically static images that users had created in photoshop and uploaded to the site. The designs themselves weren’t truly customizable.

We did excel at that. And we continued to launch new businesses aimed at specific niches, offering customization: Bridal Party Tees and FunnyShirts.org. (Example design: Buy Me a Shot, I’m Tying the Knot: [Name Here]’s Bachelorette Party.)

It seemed to me, the future was a combination of both business models.

The ultimate POD website would be a platform where you could create your own custom apparel business aimed at your own favorite niche, while offering genuine customization.

Well, we built it. We launched a storefront platform on each of our three sites in 2014.

We were trying to create a platform where anyone could join up and launch their own Bridal Party Tees.

Guess what? That’s exactly what happened.

The best example might be a storefront on Customized Girl called Mom Means Business.

Just like Bridal Party Tees, she has a very specific audience: sports moms.

And just like Bridal Party Tees, she sells truly customizable designs. Usually they have a cute and clever graphic on the front and on the back it will say: MOM OF 00. You customize the jersey number to match your son or daughter.

This has been extremely exciting for me. I had spent so much time working on (and thinking about) the storefront platform project. It seemed obvious to move in this direction, but I didn’t know for sure if it would work. It took three years from the date of my original blog post. I had hoped for this exact type of store to pop up and then, almost immediately after we launched, there it was.

Leadership

I love writing about our various marketing channels. But the best part of these past five years? Building a team.

I now head up the Marketing department at eRetailing. The people on my team are smart and efficient and talented and I’m proud to be associated with each of them. I’m really looking forward to the next five years.

Anyone with a Facebook, Twitter, or Tumblr account has probably noticed a growing trend in the behavior of many web users. If certain people have a question, any question at all, they are increasingly likely to post their question to a public place, no matter how easily the answer might be found with Google.

I’m not sure if this behavior has an official name, but I call it “Asking the Aether”.

The questions range from terrifying (“Is Ohio a state?”) to depressing (“Who is Paul McCartney?”) to genuine (“Where can I buy this?”)

From reddit:

From buzzfeed (a collection of tweets during the 2012 Grammys):

From Tumblr:

I’m still hoping the first example is a fake screenshot, but regardless, I have no doubt that many similar conversations appear all over Facebook every day. It’s particularly surprising to see how upset the person gets once they realize that Ohio, is in fact, a state. Certainly they knew this information was at their fingertips? It’s just a Google search away? But they decided to ask Facebook instead. Were they unable to predict this reaction from their peers?

For me, the Tumblr post is much more interesting.

OMG Where Can I Get That?

Due to the popularity of “reblogging”, content on Tumblr is easily divorced from the original source. Images get reblogged and reuploaded, destination URLs get changed, sometimes the source button takes you to the original, but often times it does not.

This results in a lot of Tumblr users finding awesome products in their feed, but with no way of purchasing them. Well, not without a little effort at least. Most Tumblr users are much more comfortable simply reblogging the image with this new bit of text attached: “OMG where can i get that?!”

One day recently, I guessed that these types of posts represented an opportunity. I don’t mind saying that I pride myself on my Google skills. With the right keyword choice and tools like reverse image search, I was sure I could find these items for sale. I signed up with a few affiliate programs, I bought a domain (findthatthere.com) and I registered accounts named findthatthere on the biggest social sites: Reddit, Twitter, and most importantly, Tumblr.

I whipped up a logo and I was in business:

My strategy was to search Google using phrases such as:

• site:Tumblr.com “where can i buy”
• site:Tumblr.com “where did you get”
• site:Tumblr.com “take my money”

“Where can i  buy” alone results in a total of 3 million hits (as of March 2012). If you narrow your search by “past 24 hours”, you’ll see that roughly 100 people post this comment to Tumblr every day.

My next step was to reblog each post and add a link to the place where that item could be found.

Alas, FindThatThere was not successful. Unfortunately, there is no sure way to get your reblogged post in front of the right person. Tumblr only alerts you by email if another user follows you or asks you a question. A user would only notice my reblog if they were paying close attention to their blog’s activity feed. I could send them a message using the “Ask” feature, but Tumblr does not allow links to be included in messages.

FindThatThere is no longer active, but I still think there is money to be made.

How Tumblr Could Be Making Dozens of Dollars

Tumblr, the company, should embrace this trend in behavior.

The easiest way would be to hire one person to monitor all new comments that include “where can i buy”, then message each commenter with an affiliate link.

Of course, the beauty of Tumblr lies in it’s ability to spread viral content. Some of the posts on FindThatThere had already been reblogged 20,000 times. Every time someone comments “where can i buy” on a post, that post should be flagged with a badge. That badge should be an affiliate link. Example:

That commenter was certainly not the only Tumblr user to see this post and wonder where they could buy it. A bright orange badge would alert all other users of the affiliate link as well.

Now both of these initiatives might only bring in a few measly dollars, but imagine what would happen if Tumblr users started to expect this kind of functionality. Searching on Google can be difficult and frustrating for many people. What if Asking the Aether becomes the new normal? What if every time they want to buy something, they just post a comment? If that happened, Tumblr would begin to replace Google for those users. And that would be worth more than dozens of dollars.

A year ago, when this website was still in the planning stages, I was compiling a list of potential essay ideas. One of them was set to be “Why Y Combinator and Techstars Would Make Great Reality TV Shows”. I don’t always admit it readily, but I am a fan of certain reality tv shows, particularly Top Chef. Unlike most reality programming, shows like Top Chef and Project Runway take individuals with actual talent and showcase that talent. The contests are merit-based, so it’s fun to pick your favorite players and root for them all season long.

Years ago, while learning about YCombinator, it became clear to me that all the elements that made a successful reality show were present in the YC process including:

• the drama of interview day, the anguish of rejected teams and the utter joy of accepted teams
• the history of startups from previous batches, big exits and quiet closings
• the occasional interlude of YC alumni milestones (Imagine a startup from the current batch clicking on TechCrunch and seeing “AirBNB just raised $100 million on a $1 billion valuation”)
• the weekly dinners
• the dramatic pivots
• the preparation for, and excitement of, demo day
• the pressure and the potential

All of these factors would make a show worth watching (for me anyway), and on top of that, the exposure could be hugely beneficial to the startups themselves

In 2009 I started following the 5 minute videos posted weekly on Techstars.tv.  These guys were doing it right: nice editing, great music, a high quality production all around. It kept me coming back every week, and it made me even more sure this idea would be a great tv show.

Clearly, I’m a little late. The Techstars reality show launched two weeks ago and the third episode airs tonight at 9PM ET on Bloomberg.

So why should you watch it? Well, many of the reasons are in the bulleted list above, especially the pressure and the potential. But you should also check it out for the following reasons…

Everything Seems Eerily Familiar… In a Good way

Watching Top Chef, it took me a while to get used to some of the jargon, and when a celebrity chef is introduced I never recognize the name or the face. Techstars is the exact opposite.

One of the first founders they introduce is Jason Baptiste of Onswipe. Hmmm, that name sounds familiar. Wait a minute, is that jasonlbaptiste, the prolific HN commenter? Ah, indeed it is.

And what’s his company? Onswipe? It’s goal is to “make your publication look great on tablet web browsers”. Huh, that’s the exact idea my friend, whose graduate thesis is about designing long form publishing for tablets, was just talking about.

While I never recognized a celebrity restraunteur from Top Chef, it seems like every mentor on Techstars is also a prolific tech blogger. Hey, isn’t that Fred Wilson? And isn’t that Gary Vaynerchuk? Why yes, yes it is.

Heck, even David Cohen was wearing my sweater. I’ve never seen anyone with that sweater!

Companies in the Midst of Customer Development

“So we’re trying to figure out what I call Product/Market Fit.”

– Melanie from To Vie For

As I was watching the show, I couldn’t help but constantly think back to Steve Blank’s book The Four Steps to the Epiphany. In it, Blank argues that startups and new projects within existing companies should start with Customer Development instead of Product Development. When I first read it, I felt a little giddy with excitement. Maybe it was the self-published feel of the book, what with its terrible cover art (no offense, Mr. Blank), but it was like I had some kind of secret manual. Maybe everyone in the tech startup world had heard of it, but not too many in my part of Ohio had.

The Techstars show is exciting because these companies in the middle of the Customer Development process, and we get to watch how they navigate it. Based on the fact that Techstars has to turn away thousands of applicants, I’m a bit surprised more of the companies are not further along in the Customer Discovery phase.

Listening to the startups talking about their businesses and the language they use, you can sense that each team is thinking consciously about where they are in the Customer Development process. When I heard Melanie of To Vie For utter the phrase “So we’re trying to figure out what I call Product/Market Fit”, I couldn’t help but think, yeah, you know who else likes to call it that? People like Steve Blank and Marc Andreessen who use it constantly in their writings. But give her credit because she was literally “getting out of the building” and talking to potential customers, another element of the Steve Blank mantra.

In fact, the influence of Steve Blank is so thorough, the man himself appears on screen as part of a demo for another startup, SocratED.

This Show Was Made For You

Ultimately, we as readers of HN represent the ideal target market for BloombergTV. They have made this show for people with our specific set of interests, and I’m finding it very interesting indeed.

I can’t wait to hear what advice Fred Wilson and the other mentors dole out.

I can’t wait to see what startup “wins” the next 10:10 meeting. (Once a week, Techstars holds a meeting at 10:10 pm where all the founders gather and where at least some Jack Daniels and Heineken get involved. The founders present, a winner is decided, and a special prize is awarded. On episode 2, that special prize was a visit to one of the technical advisors on Spielberg’s Minority Report.)

And I can’t wait to see what some of these really smart companies, like Wiji, a company who is trying to customize outdoor advertising based on the viewer, are able to produce. I suppose I could find out just by Googling them now, but I’d hate to ruin the suspense.

I hope you tune in too.