As someone who has spent years watching people buy, browse, interact, and do all sorts of wackadoodle things online, there’s nothing that can get me riled up faster than poor navigation.
Navigation, including your search function, is my Kryptonite. If you’re getting enough traffic (and most sites get more than enough), it’s also the #1 determinant of your online success. Full stop.
I know. I know. So-called experts will say it’s your checkout, live chat, push notifications, dynamic content system, the vomit (er, legalese) all over your footers, or whatever else it is that they’re hawking that day. Still, you can’t find any of that/them without solid navigation.
Navigation is your site’s GPS and the OG of your website’s success. It includes your action bar navigation and its dropdowns, your hamburger menus, the navigation in your footer, your internal search function, breadcrumbs, facets, sorts, filters, and refinements, and your cart/checkout temperature/progress bar.
If your action bar navigation (the nav at the top of your site, near your header) is not strong, you put more pressure on your text search.
If the user lands on a page other than your home/main entry page, you put more pressure on your text search.
If the user comes from specific channels, especially the fickle ones like Shopping and Social, you put more pressure on your text search.
If your user comes from a device, especially through Voice, you put more pressure on your text search.
I could go on… and on… but you get the drill.
Your Action Bar features the main things you want people to look at – the stuff your site is about. Things you’d bet your house on. You can squeeze as many items as you want in there, but from a mobile perspective, you’ll need to pick about 6ish to focus on, as that’s what most browsers can accommodate. (Incidentally, designing for desktop is one of the biggest mistakes I see regarding navigation. Design mobile-first, and if you’re exceptionally good at mobile, spend extra time figuring out your Voice Navigation.)
Your Action Bar dropdowns can feature lots of items, as along as they’re uber-organized. The categories in the action bar and the dropdowns must be aptly named otherwise nobody will look at them.
Your Hamburger menu, which is technically a source of friction for many, is enhanced mobile navigation. It’s meant to steer users in the right direction quickly. Too often, companies let their hamburger menus run amok. They need to be clearly and concisely curated with the user in mind. You do not need to put everything and the kitchen sink into your hamburger menu. You need to put the stuff you most want your folks to do and look at. That doesn’t mean it shouldn’t be comprehensive. It means that if you have more than 1.5-2 screens, you may need to start cutting. Remember, hamburger menus should be very action-oriented, but it’s ok to use about an inch of space at the top and bottom to highlight special offers, sections, etc.
A solid Footer will house a repeat of your top nav with additional Customer Service-y items, contact information, an email capture (at the top of the area), and any info appropriate for old-school customers. This may include a catalog sign-up, your Ordering from a Catalog/Quick Order block, etc.
Your Checkout navigation should be specific for your cart and checkout areas and must be tested thoroughly as every company tends to end up with a different “best” version. Cart/checkout navigation has two parts: (1) reduced site navigation (or not) and (2) temperature bar navigation. Yes, temperature/progress bars count as navigation in users’ eyes.
The rest of the navigational items are typically dependent on your internal text search. From a desktop perspective, that would be things like facets and refinements. From a mobile perspective, it would be your sorts and filters. This second-tier navigation (aka refinement or narrow navigation) is, by far, the area that most companies fail at search-wise. It also has the most opportunity for users and traffic. Take the time to get your data in tip-top shape. Make sure your choices are clear, are worthy of the space they’re assigned, and genuinely help the user narrow down their selection(s).
Why do so many companies struggle with their internal text search?
This is the billion-dollar question.
The most common reasons that companies struggle with their internal text search are:
Weak foundation: This includes things like poorly established/maintained taxonomies; inconsistent content structure; disconnected systems; and too much or disjointed information that’s been inadequately prioritized.
Poor content: Abysmally written content; not enough content in general; only features and no benefits; lots of products with duplicate copy; variants that are undifferentiated; etc. This can also include overly optimized (SEO) copy or too much copy that has not been prioritized or put in a hierarchy. You see this a lot when things like titles, alt tags, descriptions, and the like have been written to game the system and are dumped into a search tool without oversight.
Inadequate labeling: Tags and labels that are out of date, incomplete, incorrect, ambiguous and/or meaningless, lacking proper context, and altogether missing are a H-U-G-E problem for many companies. However, one of the bigger problems is that they’re often inconsistent. This tends to happen when there’s an overwhelming amount of data and/or several merchants/divisions involved that all do things differently. It’s one of the easier things to fix and should be done, well, yesterday.
Lack of oversight/follow-up: Search is not a set-it-and-forget-it thing, but far too many companies treat it like it is. When you’re implementing a search tool/system, you need to dedicate resources for the initial project and ongoing work. The companies with the strongest search work on it daily. Catalogers and other traditional direct marketing companies tend to have the worst-performing searches. This often happens because they don’t readjust things when they drop a new book/mail piece or put things into overstock/clearance.
Whether you use a search package that was developed in-house or out-of-house, you should identify which of the areas above apply to you and how you can fix them. Voice Search, which is chomping at your heels and is not for the faint of heart, depends on your foundational structure and the data that supports it.
How is Artificial Intelligence impacting internal text search?
Artificial Intelligence is changing the game when it comes to search. No hyperbole. Using machine learning, natural language processing, and computer vision, AI-enabled search allows you to deliver highly targeted (relevant) information to your users based on their intent. It anticipates what the users are looking for and delivers it to them in one result or a stringed conversational format, whichever you/they prefer. Old search feeds the user to the wolves the minute they hit enter. They’re on their own to survive. With AI-assisted search, you hold your user’s hand throughout the process, as long as they need you.
Because Artificial Intelligence learns as it processes more and more data, it improves its recommendations over time. This allows you to deliver a more personalized experience and deeper, more helpful conversations with your users about what they’re looking for. In turn, this improves sentiment and increases customer satisfaction.
One of the most significant benefits of using AI in your search is the integration of voice and visual search at scale. Many marketers are having difficulty adjusting to voice and video inclusion, but both are proliferating and will need to be accommodated sooner than you might think.
If I power my internal search function with artificial intelligence, does it change which fields I feed it?
Yes. Because AI/ML can effectively process and analyze a lot more data, you can usually feed it a lot more. (Please establish a solid priority system before going hog wild.) It also tends to handle sentiment data in a much more productive fashion.
You already know to feed your system things like product copy (features and benefits), category information, metadata (SEO), reviews, and so on. If you’re using an AI-enabled tool, you’ll also want to consider things like the searcher’s past browsing behavior and/or purchases, the location of the user, the channel the user came in on, any profiling information you have, etc.
So, is AI-enabled search just my regular search with more fields?
This question is legit because so many vendors are selling that their package has Artificial Intelligence when it’s no more than if/then logic powered by an Excel spreadsheet.
True AI-assisted search is sophisticated. It’s not just dumping one word (ex: “widgets”) into a box and then showing a bunch of hopefully-prioritized widgets. It allows the user to ask a question or upload a picture for you to match with your closest product/service. When you search “widgets” and it returns 350 products, it stays with you to refine the list, so you get to the ONE widget that’s best for you. It also looks at the user’s intent while searching and contextualizes the information. It factors in user behavior and any available background information. Have they been to the site before? Are they a customer? What have they engaged with? What channel did they come from? How long have they been on the site? How many pages have they viewed? What other things have they looked at?
Search tools enabled with AI/ML respond to keywords, questions, phrases, and commands. The user writes what they want/need, and the system parses it. With traditional search, the user can use keywords, but other stuff often meets with a resounding “no X found” boom.
AI learns from past and current searches. Gone are the days of building dozens of directed search pages to ensure the finds you show are on target. AI helps make things progressively better for each new search user. It’s excellent at offering recommendations based on users with similar behaviors. It’s also beneficial for prioritizing content within the user path. It knows who/what/when to show content to and where it’s best to remain hidden.
It’s important to note, though, that like all other things Artificial Intelligence/Machine Learning-related, the success of your tool is based on your architecture/foundation (including training and testing), your data; and your day-to-day maintenance and oversight. Can you start without a thesaurus and have AI build one for you? Yes, but why on earth would you? Can you dump all your metadata into the mix without oversight? You can also take an axe and cut off your right hand. It’s doable but certainly not advised. Sort out your lexicons and taxonomies. Enhance them with tools like thesauruses, spelling and grammatical checkers, and text processors. Build a house you want to live in and keep adding to, not one with a cracked foundation, rotten walls, and a detached chimney.
Is it hard to train a search tool?
It’s not so hard that you need to be a rocket surgeon. You don’t. It’s challenging because you need to be uber-disciplined about the initial set-up and solid at maintenance. Otherwise, things run amok. Unfortunately, many marketers cut corners during the testing and training and/or don’t maintain things properly. (Zero out of five stars. Not advised.)
The more training data you have, the better your search will be. This includes text and images. It includes keywords, descriptions, phrases, questions and answers, sentiment, and SEO content. It includes information about past sales and may include future projections. It can consist of merchandising hierarchies, category markers, and accessibility labeling.
Maintaining your search is a constant process. It gets easier, but you must still check and ensure things are on the uppity-up. Is the search continuing to return the results that it should? Are they appropriately prioritized? How are you accommodating the new items? Is old seasonal stuff negatively impacting the results? What about clearance? Is the customer behavior (past and real-time) information affecting things the way they should? What’s the guest search experience for a new visitor? A repeat visitor? A buyer?
One of the things that can help marketers is to develop an evergreen list of things that you look at all year long and then a list of new/seasonal type items that you check during their “hot season.” One of the best reasons for having an evergreen list is to easily spot whether your results are changing. (If they are changing, are they changing the way you want them to change?) Too often, folks check the popular stuff at the very beginning and then set it and forget it. Building an evergreen list becomes increasingly important as your personalization and contextualization grow.
How do I know if my search is improving?
There are oodles of things to look at when it comes to searching; some are more straightforward than others. That’s why it’s good to look at the performance of your search function over time.
Some of the best things to look at are:
How many searches are you getting every month? (You can do this for whatever frequency you want – just be consistent about it.)
What percentage of your traffic, by channel, are searching?
What percentage of repeat visitors are searching? Are they using item numbers?
What percentage of new visitors are searching? Are their keywords different from the ones that repeat visitors use? You can do this by channel and/or as a group.
What are the top searches? Are these items adequately represented in your navigation? People search when they can’t easily spot something. You can do this by channel and/or as a group.
What’s your search exit rate? How many people search, get a results page, and then bail immediately? You can also measure how much time they spend on the search page, although it’s often a more difficult metric to gain insights unless you split it out by channel.
How many null results out of 100 are you showing? Which queries result in a no finds page? (Please be sure to build out a proper no-finds page with another chance to search AND recommended products to keep the user from bailing.)
What percentage of search results show a PDP (Product Detail Page) vs. a category or subcategory page?
What percentage of searchers adopt to cart? Checkout? Many folks focus on just measuring overall conversion – it’s fine to measure overall conversion, but it’s not the best indicator of improvement when you’re using a dynamic search. Incidentally, this is one of the most valuable metrics you can track.
How much time does it take to show the results? (Front-loading on a mobile device on a 3G network is how you should measure it.) Presentation time (the number of seconds it takes to show anything) often influences your bounce rate.
Besides personalization, how do I get more bang for my buck out of my search tool?
Whether you use a traditional tool or a tool with all the AI, there are several things marketers can do to increase revenues and improve user engagement regarding search. Here are some of the most successful:
Establish rules to push your highest-margin products. “Highest margin” may not be your thing, but whatever your thing is, once you’ve established a solid foundation, set up some rules to accommodate it. This can be especially lucrative during your best seasons. (Holiday seasons for gift catalogers, for example.)
Review the top 100 most frequently used terms weekly and act upon the information you see. If you constantly see bazillions of searches for the same exact words and phrases over and over, it’s likely because they’re not represented well enough in your navigation or your entry page(s). Reviewing your most frequently used terms will help you spot items that are heating up or things for which you may need more supporting content.
Determine how your main channels (Direct, Email, Paid, SEO, Social, Referral, etc.) influence your search function. Businesses that do a lot of offline marketing (catalogs, TV, radio) tend to disrupt their entire search ecosystem when a new book drops or another event occurs. This can be an okay/indifferent thing, but for many companies, it’s a raging dumpster fire. This is also true for companies that do a lot of Social, especially with influencers.
Make sure you use a stacked (enhanced) thesaurus and synonym finder. I get that the base tools are part of almost every package but adding your lingo from the jump goes a long way to establishing a solid foundation. , Yes, even if you are using the most robust deep learning tools ever. (Remember, training accounts for a big part of the success of any AI project.)
Review your autocomplete suggestions. Autocompleting words and phrases can help your users find what they’re looking for faster and easier. Still, companies often find that the way they have things set up actually decreases adoption to cart because of how the recommendations are ordered. (This is one of the many things that AI-assisted search does well, and you still need to pay attention to it, especially at the beginning and with new product introductions.)
Test graphics within your search. For some, even the use of small thumbnails can make a difference. For others, it’s a complete bomb. Worth testing, though, because for the people it works for, it really works.
Prioritized indexing. This isn’t discussed nearly enough, but it can materially impact your success. Smarter indexing is known for eliminating duplicates and grouping related items next to each other. This makes it easier for people to choose what they want to view most. Ecommerce companies can take it a step further and prioritize the choices based on proprietary logic (house brand, highest margin, most available) and so on. If you’re using AI, you should also dynamically personalize how many items to show each user. Every person has a threshold number, and exceeding it forces them to use filters/facets (if they’re good) or refine their search. The latter is often a hot mess.
What are some of the biggest mistakes marketers make when it comes to search?
Thinking that more results is what the user wants.
Users want high-quality, relevant results. It’s been proven repeatedly that they act on them more when they’re in priority order. This also increases user session and page depth and decreases search bounces/exits.
The number of choices you show is essential in that the nanosecond something looks out of whack; the user will think the “search is broken.” This is difficult to see in dynamic search paths, but it’s really apparent in live usability, so it’s something you should pay attention to, at least for your most popular items. If results seem duplicated (same picture/title) to the user, you tend to see a higher bounce rate and decreased adoption. (Adding a burst to an image or a slug to a title can go a long way in making search results more productive.) In usability, you’ll see that users instantaneously judge “is it broken or not” before delving deeper into the results. Broken loosely translates to things like whether you are showing any finds, if the results look accurate at first glance, and if they are low-quality results or high-quality results.
Incidentally, reaffirming what they searched for and how many results you found is helpful in building a solid search program. Showing the number of finds lets the user immediately confirm whether they searched for the right thing. (Please paginate them according to technical SEO standards.)
Users who employ search in their journey are typically some of the very highest converters you’ll get on your site. (It’s usually about 2.3x+ the average.) There is a downside to these high-potential people, though… They bail quicker than regular browsers. That’s why it’s very important to have a “do you want to search again?” page with additional recommendations of things they should look at and an abandoned search trigger program set up. Both SMS and email triggers work.
Please be sure to also look at where exactly your failed searchers are bailing. Separate failed searches by page. You want to look at who bails immediately and who bails per page/result after that. Both groups deserve separate remarketing and retargeting programs. Very few companies do these, and they work like crazy. Very lucrative.
Have questions about internal text search? Have a tip you’d like to share? Tweet @amyafrica or write email@example.com.