218: Mid Year Report: How My Online Store Performed In The First Half Of 2018

Podcast: Download (Duration: 26:36 — 36.8MB)

The most expensive supplier quality control mistake you can make is skipping the start-of-production inspection with a vendor you trust. I skipped it with a supplier we had used for over five years without a single problem, and it cost Bumblebee Linens four months of inventory on our best selling products.

This is the mid year report for my ecommerce store covering January through July 2018. Revenue and profit were up 3%, traffic up 6%, and margins roughly flat, which under the circumstances I consider a genuine win.

The reason those numbers are barely positive comes down to three problems that landed at once: a failed color inspection that wiped out our inventory pipeline, a server auto-update that corrupted customer personalization data, and losing half our warehouse staff. Here is what happened, what it cost, and the three levers we pulled to stay level.

Key takeaways

• Always run a start-of-production inspection, even with a vendor who has been flawless for years. Catching a problem at the end of a run means restarting from zero.
• Chinese linen manufacturers collectively shifted their white dye toward a bluish white to suit European and Japanese buyers, which broke color matching across all three of our vendors at once.
• Handmade product lead times run three to four months, so a rejected production run costs a full extra cycle.
• Turn off automatic updates on production servers. A database version auto-update silently corrupted personalization data on random orders for weeks.
• Most stores under-email their lists. We send at least weekly and six or more times during a sale, with a spam rate under 0.01%.
• Facebook Messenger open rates ran 6x email and click-through rates over 10x email during this period.
• Ad spend is a two-way lever. We dialed ads down when fulfillment could not keep up and up on the products that were still in stock.
• Treating employees well means they give notice and help train replacements rather than disappearing.

Why you should inspect at the start of production, not just the end

An end-of-production inspection tells you whether to accept a finished order, and a start-of-production inspection tells you whether to let the order be made at all. That distinction cost us four months.

We got samples back from the end-of-run inspection with a slight blue tinge instead of pure white. On its own the fabric looked white, and it failed to match the other linens we sell.

That matching problem is fatal for our catalog. Customers order dinner napkins in a color and then buy cocktail napkins to match, and a mismatched white makes the cross-sell impossible.

The vendor told us they could not re-dye the fabric to our usual white without degrading it. Starting over meant another four-month production cycle, and we had been expecting the shipment in January.

We run start-of-production inspections on brand new vendors as standard practice. We dropped it for this one because they had been rock solid for five-plus years, and that is exactly the reasoning that cost us the year.

Why all three linen vendors shipped the wrong white at once

Chinese linen manufacturers shifted their standard white dye toward a bluish white because European and Japanese customers prefer it. When our number two and number three vendors started shipping the same off-white, it was clearly not coincidence.

We asked all three what was happening and got the same answer. The factories followed the market and changed their dye formulations.

That is worth knowing if you source anything color-critical overseas. Industry-wide specification drift can break your product without any single supplier doing anything wrong.

What does a four-month inventory shortage actually cost?

Being out of stock on your best sellers costs you Amazon rankings, wholesale relationships, and the customers who go find another supplier. Our Amazon listings sold out entirely, and we had nothing to send event planners who buy from us in volume.

Wholesale customers are the real exposure. An event planner who cannot get linens from you finds someone who has them, and winning that business back takes far longer than the stockout lasted.

We received a partial shipment in July with the balance due in August. The plan was to recover during the holiday season.

Why you should disable automatic updates on a production server

An automatic update on a production server can break your store in ways that take weeks to notice. Something auto-updated on my server in January despite auto-updates being off, and the site started behaving erratically.

The symptom was subtle and expensive. On random orders, the personalization text customers wrote for our custom linens vanished from the database entirely.

Someone would order ten personalized handkerchiefs and all ten messages would disappear. My wife was contacting customers to re-collect their personalization multiple times a week.

The root cause was our database software version changing on its own. I dialed down ads on personalized products while I chased it, which directly suppressed revenue.

What is involved in a major ecommerce platform upgrade?

A deferred platform upgrade compounds into a project that takes days of uninterrupted work. Since I was already under the hood, I upgraded everything at once and it took four full days of eight-hour sessions.

The scope was substantial. Migrating from PHP 5.3 to PHP 7 required a significant code rewrite because functions had been removed from the language between versions.

The rest came along with it. I upgraded the server operating system to current Linux since the old version was deprecated, upgraded the SQL database, and updated a stack of plugins.

Doing it while the store was live added the stress. Things were broken in the middle of the migration, and I was learning the process from forums as I went.

Open source or hosted shopping cart: which should you choose?

Choose a hosted platform like Shopify or BigCommerce if server migrations sound unpleasant, and choose open source if you want full control of your source code and have the technical skills to maintain it. That upgrade was genuinely hellish, and I still do not regret the choice.

The economics favor open source at scale. Across roughly a decade, I have saved close to six figures against a hosted platform once you count transaction fees, paid plugins, and app subscriptions.

Frequency matters more than difficulty. I have gone almost ten years without a major cart or server upgrade, so amortized across that period the pain was manageable.

Why an unsupported shopping cart is the real risk

The variable that determines your maintenance burden is third-party developer support, not whether the platform is open source. My upgrade was painful specifically because my cart stopped being actively supported years ago.

WordPress is the counterexample I run on my blog. Nine years of upgrades that mostly amount to pushing a button, because it is well maintained with a deep developer ecosystem behind it.

Hosted platforms carry the same abandonment risk. Yahoo Merchant Solutions was the leading hosted cart in 2007 and is a non-factor today, and I know people still stuck on it because migrating is so painful.

Arguably a dying hosted cart is worse than a dying open source one. With source code access you can at least maintain it yourself, and on a hosted platform you have no such option.

The practical rule is to pick the cart with the most popularity and the deepest third-party developer support at the time you choose. You cannot predict which platform wins in ten years, and you can guarantee you have someone to hire for help today.

Why warehouse hiring got harder for ecommerce sellers

On-demand gig work has thinned the talent pool for nine-to-five warehouse jobs, particularly in the Bay Area. Uber, Lyft, DoorDash, and similar services let people work whenever they want, which makes a fixed warehouse schedule a harder sell.

We lost two of four employees this year, a 50% turnover rate for a company that normally has none. One had been with us five years and the other two, and both left for better opportunities or changed circumstances.

Losing them meant I personally went back to packing linens while recruiting. That forced another deliberate ad reduction so we would not take orders we could not fulfill.

How do you screen job applicants who cannot follow directions?

Build a simple instruction-following test into the application itself. Our process requires applicants to find our website and send an email with a specific subject line, and the number of people who cannot do that is remarkable.

The filter costs nothing and eliminates a large share of applicants before you spend any time on them. It also tests exactly the attribute that matters for careful fulfillment work.

How often should you email your ecommerce list?

Email at least once a week, and six or more times during a promotion. Most store owners email far less than that out of fear of burning out their list, and they are leaving money on the table.

Every campaign we send makes money. Our spam complaint rate sits under 0.01%, which tells you subscribers are considerably more resilient than sellers assume.

Frequency also works as a sales dial in both directions. When we needed revenue to offset the inventory gap we sent more, and when fulfillment was strained we sent less.

How Facebook Messenger marketing compared to email

Messenger open rates ran 6x higher than email and click-through rates over 10x higher during this period. That gap was the single biggest bright spot in the first half of the year.

The advantage came largely from low adoption. Very few sellers were using Messenger marketing at that point, so inboxes were uncluttered in a way email inboxes have not been for years.

That channel has changed substantially since, as Facebook tightened its messaging policies and moved toward paid sends. The transferable lesson is that a newly opened channel produces outsized engagement before it saturates, which is an argument for testing new channels early.

How do you keep revenue flat during a stockout?

Shift ad spend and email promotion onto the products you still have, rather than trying to sell what you cannot ship. We increased ad spend behind the linens that were in stock and ran more email campaigns promoting them.

Those products are typically ones we do not push hard on Facebook, so there was unused headroom. Redirecting the budget offset a meaningful share of what the stockout cost us.

Deliberate growth control is what made the levers available. My wife and I intentionally hold our growth below its maximum so the business does not run our lives, which leaves us room to pull levers when we need them.

Why treating employees well pays off when they leave

Employees who are treated well leave properly. Both departing staff gave ample notice and volunteered to stay longer to help train their replacements.

We have given away two cars to employees as thank-yous over the years. The return on that showed up precisely at the moment when a bad exit would have been most damaging.

Frequently asked questions

Should you inspect an order from a long-term supplier?

Yes, and inspect at the start of production rather than only at the end. A five-year track record does not prevent a materials or process change, and catching a problem after the run is finished costs you an entire production cycle.

What is the lead time for handmade imported products?

Three to four months is typical for handmade goods. Our linens are made by hand in rural Chinese communities and collected by our vendor, which extends the timeline well beyond factory production.

Why did white fabric from Chinese suppliers change color?

Many Chinese linen manufacturers shifted their white dye toward a bluish white because European and Japanese buyers prefer that shade. The change hit multiple vendors at the same time.

Should you enable automatic updates on an ecommerce server?

No. An unplanned database or language version change can silently corrupt order data, which is what happened to our customer personalization records for weeks before we traced it.

Is an open source shopping cart cheaper than Shopify?

Over a long time horizon it can be. Across roughly a decade I estimate close to six figures in savings versus a hosted platform once transaction fees, plugins, and add-on costs are included.

What happens if your ecommerce platform stops being supported?

You take on all maintenance yourself, which is what made my upgrade so difficult. This risk applies to hosted platforms too, and there you cannot maintain the code yourself at all.

How many emails should you send during a sale?

Six or more. Every send during our promotions generates revenue, and our spam complaint rate stays under 0.01%.

How do you handle orders when you are short-staffed?

Reduce ad spend deliberately so incoming order volume matches what you can actually fulfill. Taking orders you cannot ship on time damages you more than the lost revenue does.