There was a problem. Please try again.
Thank you!

Two years ago, Emile Petrone posted a humble question to reddit.com/r/arduino: "Would you support an Arduino marketplace or am I totally off base?" After observing the market for a while, he saw the need for a central destination where people could buy handmade tech. Many others agreed, and Petrone started to build. Five months later, he quit his job and raised $500,000 in funding to turn Tindie.com into the eCommerce platform for these types of products. Today, the site draws customers from all over the world, including companies like SpaceX, Google and Intel that can't find what they're looking for anywhere else.

But Tindie had another big milestone early on: Julia Grace joined as Head of Engineering after gaining expertise at companies like IBM and VigLink. Now, as CTO, she heads up a team where engineering plays an enormous role in maximizing customer acquisition and sales. In this exclusive interview, she shares the lessons she learned developing the company from its scrappy roots into a major supplier, and the tactics that can give eCommerce startups a big, early lead.

Build in Browsing to Learn About Your Audience

When Tindie first started selling indie hardware, the site enabled customers to simply browse listings. It had no search functionality. “When you’re a small startup, you don’t need to hurry to implement search,” says Grace. "If your product catalog starts out small and people don't know what to explicitly search for, then the best thing you can do is optimize for browsing. The worst thing you can do is prematurely build features like search right out of the gate."

During the initial stages of building an eCommerce site, you want to focus on building a taxonomy to categorize your expanding offerings, says Grace. This is what will eventually help you deliver a great user experience, and you can do this without getting mired in the intricacies of building out search. “If you go to the Macy’s website, you expect to see clothes or housewares as broad categories,” says Grace. “Then they give you the ability to drill down based on what they are looking for. This lets them learn more about customers’ preferences and how they find and discover products. A startup can do the same thing."

To do so, Tindie focused on building a comprehensive taxonomy to make its inventory easier to find and learn about how its customers found what they were looking for. This took multiple iterations and re-categorizing products. By experimenting and observing this process, they realized that the best way to guide traffic was based on personas. “We tried segmenting the site based on buyer profile,” says Grace. “For example: ‘I’m an electronics hobbyist,’ or ‘I’m a professional electrical engineer,’ or ‘I’m a DJ super into my equipment.’ This made sense based on our customer base.”

Before landing on this solution, the company had tinkered with tiered categories, allowing people to identify as beginner, intermediate or advanced — this didn’t work out so well. “We realized really soon that people don’t identify themselves this way,” says Grace. “The way people behaved on the site, we knew we weren’t getting it right.” It turned out that the lines between intermediate and advanced were blurred. This was a key learning: How you identify users is often not how they think of themselves. For a while, it seemed like the answer might be to offer just an advanced category and a beginner category, but this made finding products in the latter bucket very difficult because so many fit that description.

Tindie also took customer feedback very seriously, keeping a close eye on how people chose to engage or critique the site’s navigation. “When you give your customers more options for how they can engage with you, you can measure how invested they are in your products — you can classify the level of effort they’ve put into being part of your community,” she says. “For example, if someone likes a product, that’s low effort. Writing a comment shows medium effort. But if they go the extra mile to send you a note or suggest new product categories, you know they’re loyal. When someone writes in and complains about our current segments, I now ask them about the categories they want to see. We want them to feel involved.”

Empower your users to modify your navigation — but have a high barrier to entry for the ones you listen to.

As effective as advanced browse functionality can be, eventually you will get to a point where your site is home to thousands or millions of products and you’ll need to bake in search that makes sense for how your customers already find what they want.

Deploying Search Should Be All About Efficiency

When it comes to incorporating search, you have two options. One is quick and dirty — you can add ‘search as a service,’ basically using a third-party's API to handle searching and indexing your products. Depending on your goals, this can save you time and money. “There are several companies out there that will provide search for your site,” says Grace. “This is the right choice when you don’t have the horsepower, staff or bandwidth to build the backend, analytics and dashboards from scratch.”

Your second option is to build your own search tool. This is the route Tindie chose, creating very simple, text-based search based on keywords in product titles or descriptions. “Early on, we used Google Analytics and appended UTM variables to every search result so we could track where the traffic was coming from, but this got complicated fast — especially when people shared links with their friends. We had no idea if they were coming from our search box or not,” says Grace.

Most importantly, you need to be mindful of how your search system can impact your SEO — especially when you’re trying to build brand presence, she says. At first, Tindie would generate unique URLs for search results and products based on the parameters customers selected (like price range, etc.). This led to inaccurate results, which damaged the company’s Google rankings. “When you start thinking about search, you have to consider the functionality, but you also need to think long-term about URL structure and whether you're 'polluting' your Google search results with all these additional pages that are useless to customers. These can have a very negative impact on your SEO, which directly affects your brand."

The best way to do this requires a lot of effort but it’s worth it — you need to keep meticulous track of every URL on your site. “When you change something, make sure you handle it properly. If you move a URL, issue a 301 permanent move right away. Use the right HTTP response codes to signal to Google’s crawlers what’s going on. You need to remember to redirect traffic and avoid 404 pages whenever you can. You can’t afford inaccuracies.”

Knowing what people click on after they search is key, but your top priority should always be making sure results are relevant.

“Are they seeing 100 similar items or 50? Did they mean to find something else? Are they trying to search for something broad and getting a lot of granular results?” says Grace. “These are nuanced problems you need to understand, so unless you have significant in-house resources to do this, you should outsource search.”

To improve accuracy, Grace recommends implementing faceted search — a technique for retrieving information that is organized according to a classification system. This allows users to explore a collection of results through multiple features. “You often seen faceted search on airline travel sites or retail sites that segment product based on gender, age or price range.”

That said, proceed with caution. "While faceted search can provide you with good initial insights as to what matters to customers, it is very tempting to go overboard and add too many facets. We tried this, and the more facets we added (e.g. what country the hardware seller was from, etc.), the less people used them. They were quickly overwhelmed. So choose carefully."

Once you have chosen, the most important thing you can do is pay attention to the filters people use. “If you’re running an eCommerce site like Etsy where you have a lot of similar products showing up, it’s likely that users will be more discriminatory about things like price,” says Grace. “On the other hand, if you have a site with specialized products, like Tindie, the audience may be less sensitive to price. Specialized customer bases are all about finding what they’re looking for, they don’t care about what they’ll need to pay for it.”

Julia Grace, ‎CTO at Tindie

Be Thoughtful About How You Route People

When it comes to optimizing conversions, you want to make sure you’re directing traffic in ways that keep people on your site, even if they don’t find exactly what they want. This can sound time-consuming to build — especially when you see elaborate recommendation engines on sites like Amazon. But you don’t have to go all out immediately, Grace says.

For a while, Tindie didn’t have a large enough data set to know what other products they should recommend to their buyers. This becomes a big issue if people can’t find the product they want or if it happens to be out of stock. Unless you’re smart about directing them, they’ll leave and probably never come back.

“Depending on your site, you may not have the luxury of offering customized recommendations when this happens. At Tindie, most products are one-of-a-kind, so we ran into this a lot.” To retain more shoppers, they came up with a solution that would require less specific user data:

If you can't show an alternative to an out-of-stock product, suggest something complementary.

Tindie used the personas it had crafted while building its browse functionality to determine what other products customers might want. Someone building a DJ rig, for example, might be looking for audio parts and other related equipment. A hobbyist building their own robots might be interested in motors or sensors. “We would ask ourselves: ‘What would we recommend to this type of person?’ and that ended up being very powerful.”

The company also used Google Analytics heat maps to keep close track of where people clicked following a search. If you pay attention to why customers choose one product over another, you can formulate a hypothesis about who they are (generally speaking), and you can use that data to architect a better experience. “We segment our core users who are hobbyists and separate them from our corporate buyers, for example,” says Grace, adding that one way to do this would be via cookies and tracking what the users purchased, then showing them different versions of the homepage accordingly. “You want your users to identify strongly with the products on your homepage, so segmenting based on needs and interests is very important.”

When thinking about how you direct traffic on your site, Grace recommends the following:

1) Don’t force your users to create an account before they can buy. For most eCommerce companies, it doesn’t make sense to put customers through a signup flow before they can place an order. “Big eCommerce sites like Amazon and Etsy can afford to do this. They’ve reached critical mass and can make people create accounts,” says Grace. “But as a smaller startup in the space, I would say you should never do that. Our conversion rate jumped over 50% when we removed our mandatory login option.”

2) Avoid gimmicks to extract customer information. “When you're trying to appeal to a savvy audience, you need to make sure they trust you first and foremost,” she says. “In our case, if we were to introduce a 50% discount after you create an account, our churn rate would be huge. They wouldn’t be long-term valuable users, and that’s the audience you want to build early on.”

Unless your site is about deals or bargains, be careful about coupons and discounts. If your differentiator is that you offer rare products, then not only will buyers be less price discriminatory but the value of your brand should be centered around this uniqueness (and not discounts). While some of Tindie's sellers provide coupons, they do not significantly contribute to conversion, and coupons are not a reason people keep coming back to Tindie.

3) Make your checkout experience fool-proof. When you let people buy without opening an account, you have no way to capture their information or preferences. Also, the buyer can’t go back and look at their past purchases. But there are ways to work around these obstacles. “Tindie built a system where we give buyers a long, hashed, unique URL that they can use to check the status of their order, even if they didn’t sign up for anything,” says Grace. “If they decide to eventually create an account with the same email address, we seamlessly tie everything together so they can log in and see their purchase history. This has been very effective for us when it comes to attracting and retaining verified users.”

4) Keep your company’s ‘About’ section minimal. People who land on an eCommerce site are looking to buy, not read about you. “Startups have this tendency to talk a lot about themselves in a way that doesn’t keep their end consumer in mind,” says Grace. “So we’ve chosen to focus way more on the functionality — the buying and selling experience. We know that customers don’t care who we’ve raised money from or who is on our engineering team.”

The one caveat to this is recruiting. When you're in a growth phase and actively hiring, then a page talking about your team, background, and culture is important, but it should never be the first thing someone sees when they come to the site, and it shouldn't be prominent in your navigation.

Customers want a clear-cut experience. They want to find what they want, buy it and leave. Respect that.

5) Figure out what makes your buyers ‘tick.’ Some of your customers may be price sensitive, while others are more concerned with quality. This depends on what you’re selling of course, but regardless, you need to be able to group people based on these attributes. “In our case, we offer a unique set of hardware products, so price isn’t as important to our customers,” she says. “But all kinds of things might be important to them. For example, if photos of products seem to boost sales, then you want to focus more on providing better, engaging photography.”

6) Know your audience’s privacy tolerance. While some customers may appreciate tailored recommendations based on their buying history, this might make others feel monitored and uncomfortable. To neutralize this concern, Grace suggests that sites, “be very upfront and disclose that you are anonymously tracking what they're doing to improve the site and your service, not to infringe on their privacy.” This is a bigger concern than many eCommerce companies believe. It’s something they should keep squarely in mind every time they decide to collect a new type of data.

“Understand the core mission of your buyers. Then communicate that back to them so they start to identify your brand with their needs.

Invent the Best Shopping Cart

While shopping carts have largely become old hat, that doesn’t mean you shouldn’t innovate. “Most people end up modeling their carts after Amazon and other big eCommerce sites because they think people are used to that structure,” says Grace. “But you need to realize that those companies have a competitive advantage: They can force people in any direction because they're big enough — they don’t necessarily have the best interest of customers in mind. As a startup, you don’t have the luxury of experimenting like that, so doing what the big guys do isn’t always best.”

To follow in the footsteps of larger online retailers, Tindie prematurely optimized its shopping cart. “At one point we even disabled the option to click away once you were in the checkout flow, and this made people very unhappy,” says Grace. “When we rolled back those changes, we stopped getting the angry emails.”

Number of steps in a checkout flow also has a big influence on conversion rates. Tindie started out with a 5-step shopping cart process, but the drop-off was so high that they eventually condensed it to two steps. This made a sizable impact on the number of completed purchases.

What Tindie's simplified shopping cart looks like today.

Don’t Miss Out on Accepting Payments

If your customers live around the world, PayPal integration isn’t just a good idea, it’s imperative, argues Grace. In other countries, millions of people don’t have credit cards. Tindie is a global retailer, and a significant segment of its customer base uses PayPal because they're concerned about identity theft — this is even the case in the U.S.

It’s wise to integrate three payment options and to select them based on your site’s key demographics. “Perhaps include Amazon payments, Stripe, or even bitcoin through a service like Coinbase if that makes sense for the people who are buying,” she says. “The key thing is to make sure you’re not prematurely optimizing by adding payment methods that a vocal minority asks for, or methods that seem popular in Silicon Valley but don't have penetration outside."

As a specific example, Tindie tried offering direct bank account payments, and no one used the feature. It turned out to be a significant waste of time and effort. "We should have asked our customers before implementing it, particularly the demographic we thought would be most interested."

The payment systems you choose can make or break your eCommerce startup. “If you’re building an international business where buyers are ordering hundreds of thousands of dollars in products — and these have to be shipped internationally — you need to be keenly aware of fraud. There’s a high chance you’ll fall victim to it at some point,” she says. “That’s why integrating with an established payment processor like PayPal or Stripe can be a good move. They're very good with fraud detection. It’s a core part of their offering.”

At one point, Tindie was blowing up in Asia, shipping tons of product to Singapore, Taiwan and Vietnam. “We were seeing a lot of unusual transactions where people were using U.S. credit cards, but the orders were being shipped internationally,” says Grace. “We were absorbing a lot of losses because of fraud in the region, and we had to do something.” The company ended up working closely with Stripe to prevent these types of transactions. “Working with a company like this, that does a lot to fight fraud behind the scenes, can stop 90% of the bad buys on your site.” Notably, even though there are nontrivial engineering challenges to integrating it, Tindie has never seen any fraud happen through PayPal.

When it comes to both your shopping cart and your payment options, the number one challenge is to resist over-optimization and keep things simple.There’s a huge amount of complexity cost that comes with both of these processes on your site, and if you try to either catch up to the majors or exceed them with something inventive but unnecessary, you will stray off track. And, as an early eCommerce play, you can’t afford to waste the time and resources.

Gift cards, for example, are especially ripe for complications and premature fancy engineering (that could be better spent elsewhere).

“Offering some sort of gift card feature can incentivize people to buy more during the holidays, but you need to consider everything that goes into that on the backend before you pursue it,” says Grace. When will cards that get bought expire? What happens to that money when they do? Are there some products that shouldn’t be bought with gift cards? What if someone tries to open two tabs at the same time and use the gift card or promo code twice? “You’re basically creating your own form of currency. Sounds heavy, right? It is. And you need to think through all of the use cases and scenarios that go along with it.”

Seeing that a huge chunk of their annual revenue came in between Halloween and New Year’s, Tindie baked in a gift card system pretty early. But, looking back, Grace says she wishes they would have waited so they could have observed their users more and been more proactive about designing something that covered all edge cases — rather than fielding tons of complaints when things didn’t work right.

“Whenever you make a change to how people can buy something, that integration will take time — engineering time, testing time, time for users to adjust,” she says. “If you’re a small startup, you need to think twice about every extra feature or layer you add, because it will cause people to behave differently and spend differently.”

Recommended Articles