Chapter 9 of 11
Homepage, Navigation and Collections
How shoppers move through a Shopify store, what a homepage is actually for, and how to merchandise collection pages like the landing pages they already are.
Contents
- How shoppers actually move through a Shopify store
- What a homepage is for, and what it is not for
- Navigation that matches how people shop
- Collection pages are your real landing pages
- Filtering, sorting and on-site search
- Mobile navigation
- Merchandising decisions that change revenue
- New and returning visitors want different things
- What loses
- The order to work in
Most conversion advice stops at the product page. The pages that decide which product page someone reaches usually get filed under branding, handed to whoever owns the theme, and never measured.
That is backwards on any store with a real catalogue. The product page can only answer questions about the product a shopper is already looking at. Getting them to the right one is a different job, and on stores with more than a few dozen SKUs it is frequently the larger constraint.
This chapter covers that job: how people actually move through a Shopify store, what the homepage is for, how to build navigation around how people shop, and how to treat collection pages as the landing pages they already are.
How shoppers actually move through a Shopify store
Before any opinions, get the map. Open your analytics, pull the landing page report for the last 90 days, and group entries by template: homepage, collection, product, blog, everything else.
Most teams are surprised twice. The homepage is a smaller share of entries than they assumed, and collection pages are a larger one.
Then split each template group by conversion rate and revenue per session. You now know which surface is carrying the store and which one you have been redesigning out of habit.
Three journeys account for nearly everything, and they need different things from you:
“I know exactly what I want.” They search, or they arrive from a branded query or an email straight onto a product. Your job is to not get in the way: working on-site search, no interstitial, no popup before they have seen anything.
“I know the category, not the item.” They enter a collection and compare. This is where most of the real decision work happens, and where most stores invest the least.
“I do not know you yet.” They land on the homepage or an article and need orientation before they need products.
Four numbers tell you how well the middle journey is working, and almost nobody tracks them:
- Collection to product click rate. Of sessions that view a collection, how many open a product?
- On-site search usage. What share of sessions use search at all?
- Menu click-through by item. Which navigation items get used, and which are decoration?
- Revenue per session by entry template. The one that settles arguments.
If you cannot produce those four today, that is the first job. Everything below is easier to prioritise once you can see which step is leaking.
What a homepage is for, and what it is not for
The homepage has one job: orient a stranger and route them somewhere specific, fast.
Three questions, answered in the first screen on a phone:
- What do you sell?
- Who is it for, and why you rather than the alternative?
- Where do I start?
That is it. A homepage that answers those three and gets people into a collection or a product is doing its work. A homepage that wins design awards and routes nobody is a very expensive noticeboard.
What earns space: a first screen that names the category and the differentiator in plain words, direct entry points into your top four to six collections, one row of bestsellers with prices, and one piece of proof that is a fact rather than a feeling.
What does not: an auto-rotating hero carousel, a brand film, a founder letter, a stack of announcement bars, and the seven sections that exist because seven different people asked for them.
The homepage is where internal politics queue up. Email wants a signup block. Retention wants the loyalty programme. The founder wants the sustainability story. Each request is reasonable and the sum of them is a page with no route through it.
The rule that holds the line: every homepage section has one owner and one metric. If a section cannot be judged by click-through into a collection, product, or a measurable action, it does not get a slot. Sections that fail their own metric for a quarter come out.
One useful reframe: the homepage is not a sales page, it is a router. Judge it on how many sessions leave it heading somewhere specific, and on the revenue per session of the cohort that entered through it.
Navigation that matches how people shop
Most menus are an org chart. They mirror how the company thinks about its catalogue, how the Shopify admin happens to be organised, or how the product team names things internally.
Shoppers do not have your internal vocabulary. They have their own, and it is already written down.
Take the menu labels from your own data. Your on-site search log is a list of what people call your products in their own words. Your reviews and support inbox are the same thing at more length. This is the review and support mining method from the research chapter, pointed at labels instead of objections. If the search log is full of “waterproof” and your menu says “All-Weather Series”, the menu is wrong, not the customer.
Keep it to five to seven top-level items. Someone scanning a menu is looking for a word that matches their intent. Every additional item slows that scan. If you cannot get under seven, check whether About, Blog, Contact and Wholesale are eating slots that categories should own. They belong in the footer.
Mega menus are for stores with genuinely deep catalogues. Sixty links in a panel is not navigation, it is a sitemap with a hover state. If you run one, group it, cap each group, and give each group a “shop all” link that goes to a real merchandised collection rather than /collections/all.
One concept, one collection. Tag-driven collections multiply quietly until you have four near-duplicate pages for the same idea, splitting your links, your internal search results and your merchandising attention. Audit the collection list once a year and kill the duplicates.
Test the labels cheaply. A tree test with ten customers costs an afternoon: give them a task (“find something to clean a cast iron pan”), show them only the menu structure, and record where they click first. You will learn more than a month of internal debate, and at ten participants you are not measuring conversion, you are finding the label everyone misreads.
Collection pages are your real landing pages
On most stores the collection page is the highest-traffic non-product template, it takes a large share of paid and organic entries, and it is styled once at build and never touched again.
Treat it like the landing page it is.
The card carries the decision. Price, a second image on swipe or hover, rating with count, the variant range, and availability. A card with a photo and a name forces people to open products one at a time to compare, and most will not. Add whichever attribute your research says decides the purchase: weight for luggage, spice level for sauces, drop for running shoes.
Density is a real trade-off. Two columns on mobile shows twice the choice per scroll; one column shows the product properly. If your product needs to be seen to be judged, pay for the space. If people are scanning for a known item, give them more per screen. That is a testable question, not a taste question.
Write a short intro that helps someone choose. Not the SEO paragraph dumped at the bottom. Two or three lines at the top: what is in here, how the options differ, and which one suits which situation. This is the collection-page equivalent of the comparison question on a product page, and it is the single most neglected piece of copy on most Shopify stores.
Fix the back button. If someone opens a product from row nine and taps back, they must return to row nine. Broken scroll restoration is common, invisible on desktop, and quietly expensive on mobile where the whole browsing pattern is open-look-back.
Quick view is conditional. It helps when the decision is a fast visual comparison across many similar items. It hurts when the answers a shopper needs (sizing, materials, delivery date) live on the product page, because a quick view that cannot answer them adds a step to every product.
Filtering, sorting and on-site search
Filters should come from the customer’s decision criteria, not from your product database schema.
Shopify’s storefront filters are built from product options, tags and metafields, and the free first-party Search & Discovery app exposes filters, synonyms, product boosts and search analytics. The mechanics are not the hard part. Choosing which attributes matter is.
Filter on what people actually decide with. If fit is the top theme in your reviews, fit is a filter. If nobody has ever chosen a product by its collection year, that filter is furniture.
Show counts, and never allow a dead end. A filter combination that returns nothing needs a recovery path: what to relax, and a few products that nearly match. On mobile, filters belong in a drawer with an explicit apply button and a persistent “N results” line.
Default sort is a bigger lever than any individual filter, because most people never change it. Shopify gives you best selling, price, alphabetical, date and manual order. Best selling ranks on recent sales, which makes it self-reinforcing: what sold sells more, and new products struggle to ever get seen. That is fine for a stable catalogue and quietly damaging for a store that launches regularly.
On-site search deserves its own attention, because searchers are stating intent rather than browsing. Four things to check in your search analytics this week:
- Null-result queries. A list of products customers expect you to have, in their words. Some are catalogue gaps. Most are synonym gaps you can fix in an afternoon.
- Queries returning the wrong thing. Usually a naming mismatch between your titles and their vocabulary.
- Non-product queries. “Returns”, “shipping”, “size guide”. People search for policies. Surface pages in the results rather than showing them an empty grid.
- Search exit rate. People who search and then leave were ready to buy and could not find it. That is the most qualified traffic you are losing.
Mobile navigation
Your traffic is mostly phones and your team reviews the store on laptops. The gap is where the damage lives, and navigation is the worst-affected surface because the drawer hides everything.
Count the taps to a category. From the homepage, opening the drawer, expanding a parent, then choosing a child is three taps before a shopper sees a single product. Category tiles on the homepage cut that to one. On stores with a shallow catalogue, a horizontal category strip under the header does the same job with no drawer at all.
Keep search visible. A search icon in the header gets used. Search buried inside the drawer does not, and it is the fastest path for your highest-intent visitors.
Add up your sticky elements. Announcement bar, header, filter bar, chat bubble. Each was added by a different person for a good reason, and together they can eat a third of the viewport before any product appears.
Respect the thumb. Menu items need real height and real spacing. A mis-tap in a menu costs a page load and some patience.
Check in-app browsers. Instagram, TikTok and Facebook render inside their own webviews where sticky positioning and custom fonts sometimes misbehave. If paid social is a meaningful channel, that webview is a primary rendering target, not an edge case.
The test that catches most of this: hand your phone to someone who has never seen the store, name a product category, and watch. Do not help. Every hesitation is a finding.
Merchandising decisions that change revenue
Collection order is a commercial decision that most stores leave to a default. It is one of the few changes that moves revenue without touching design, code or copy.
Things worth encoding as explicit rules:
In-stock products come before out-of-stock ones. Check what your theme and sort order do with sold-out items today. It is rarely what you would have chosen.
The first two rows are prime real estate. They should hold products that are in stock, well-reviewed, well-photographed and worth selling. Margin belongs in the decision, not just volume.
New products need a placement window. Under a best-selling sort, a new product has no sales, so it ranks last, so it gets no sales. Pin new arrivals into the top rows for a fixed period, then let them earn their place.
Merchandise your money collections manually. Automated collections are fine for the long tail. Your top five collections take most of the traffic and deserve a human ordering pass each season.
Judge positions on revenue per impression, not clicks. The top row always gets the most clicks, because it is the top row. The question is which products earn their position, and that needs revenue per product view, not raw click share.
A quick illustrative example, entirely fictional: Ferndale Cookware sorts every collection by best selling. Their skillet has sold well for two years, so it sits top left in perpetuity, while a higher-margin enamel pot launched in March never rises above row six and is judged a failure by December. Nothing about the pot was tested. The sort order made the decision.
On testing merchandising: collection pages usually get a fraction of product page traffic, and merchandising changes tend to produce modest effects. Before you run one, put your real numbers into the sample size and significance calculator and check whether the test can detect anything. Frequently it cannot, and the honest answer is to make the change deliberately, document the date, and watch the trend, the same call Chapter 4 makes about underpowered tests.
New and returning visitors want different things
Split your funnel by new versus returning before you design either journey. Most stores serve one page to both and lose on each.
New visitors need orientation. What this is, who it is for, why you rather than the alternative, and where to start. They are the reason the homepage exists.
Returning visitors need retrieval. They already know what you sell. They came back for a specific thing, or to finish something. Making them scroll past the brand story to reach the catalogue is a tax on your most valuable audience.
Practical, cheap versions of that difference:
- A recently viewed row is the highest-value, lowest-risk personalisation available to most Shopify stores. It is retrieval, not prediction, so it cannot embarrass you the way a recommendation engine can.
- Reorder paths for consumables. If someone bought coffee eight weeks ago, the fastest route to that exact product should be one tap from the homepage.
- Send campaign traffic to merchandised collections, not the homepage. An email about wet weather should land on a wet weather collection, ordered for that email, not on a generic grid.
Two cautions. Do not remove navigation for returning visitors in the name of streamlining; people change their minds about what they came for. And keep personalisation proportional: at this size, a recently viewed row and good segmentation of email destinations outperform an expensive recommendation platform running on defaults.
What loses
House rule, so here is the list that does not work, including things we have stopped doing.
Auto-rotating hero carousels. The first slide is the only one most people see, and rotation moves it before slow readers finish. Every slide after the first is work performed for almost nobody. Pick the first slide and commit.
Homepage redesigns pitched as conversion projects. Large, slow, politically expensive, and often aimed at a minority of entries. Check the landing page report first. If the homepage is a small share of your sessions, the redesign is a brand project, and it should be honest about that.
Mega menus containing the whole catalogue. More links do not mean more findability. They mean a longer scan and a slower decision.
Infinite scroll with no way back. Especially when it also makes the footer unreachable, which is where your shipping and returns links live.
“Shop All” as the primary homepage call to action. It routes people to an undifferentiated grid and defers the decision you were supposed to help with.
Filters copied from a larger competitor. Their taxonomy reflects their catalogue and their customers. Importing it gives you filters nobody uses and none of the ones your buyers need.
Recommendation widgets left on defaults. “Similar products” and “bought together” are different algorithms with different jobs, and the default is usually the one that offers a competing product at the worst moment.
Entry popups before anyone has seen anything. They land on the exact screen where a new visitor is deciding whether you are worth their attention.
And one honest category that is not a loss so much as a limit: on many stores at this size, navigation and collection changes cannot be A/B tested to significance. Pretending otherwise produces confident conclusions from noise. Ship the ones that are clearly right, test only the ones large enough to detect, and say which is which.
The order to work in
If you are starting from nothing, this sequence produces results fastest:
- Measure the entry map. Sessions, conversion rate and revenue per session by template. One afternoon.
- Read the search log. Fix null results and synonyms. Usually the cheapest revenue on this list.
- Fix the collection card and the back button. Nothing else in this chapter matters if comparison is painful.
- Re-merchandise the top five collections. Deliberate order, in-stock first, new products given a window.
- Re-label the menu using customer vocabulary, and cut it to seven items.
- Only then look at the homepage, with a metric attached to every section.
That is the whole method, and it is the same one we run for clients. If you would rather have it done than do it, the CRO audit covers exactly this ground alongside the product page and checkout. Twenty minutes on a call, and if your traffic is too low for a testing programme, we will say so rather than sell you one.
What founders ask before they book.
How many items should a Shopify main menu have?
Five to seven top-level items is the working range for most stores under a few thousand SKUs. The constraint is not aesthetic. A shopper scanning a menu is recognising a word that matches what they came for, and every extra item makes that scan slower. If you cannot get under seven, the problem is usually that non-shopping links (About, Blog, Contact) are competing with categories for the same row. Move those to the footer and the menu usually solves itself.
Should collection pages use pagination or infinite scroll?
The choice matters far less than whether the back button works. If someone opens a product from row nine and taps back, they must land at row nine, not at the top of the collection. Broken scroll restoration is a common and expensive fault on mobile, and it is invisible on a desktop review because desktop shoppers open products in new tabs. Test the back button on a phone before you argue about pagination.
Is my homepage worth optimising?
Check its share of entries before you decide. On many established stores the homepage is a minority of sessions, because search, ads and email send people directly to products and collections. It still matters for new visitors and for brand search, but a homepage redesign is often a large project aimed at a small slice of traffic. Look at the landing page report first, then spend where the sessions are.
What should I do with out-of-stock products in a collection?
Decide it deliberately, because your theme's default is rarely what you would have chosen. Generally, keep the product reachable and indexed, push it below the in-stock items rather than into the top row, mark it clearly on the card, and put a back-in-stock capture on the product page. Removing it entirely loses the traffic and the link, and hiding it behind a filter default means shoppers never learn you make it.