Where this session sits
Session 2 showed you the five doors sessions enter through. This session follows them one step further: once inside, how do they find a product? Two roads exist - search (type what you want) and browse (walk the categories) - and they behave so differently that treating them as one "traffic" number hides your biggest levers. As always on this track: nothing to compute, everything to ask.
Same store, two very different roads 15 min live
Of Mango Lane's 144,000 product-page sessions this month, 43,200 (30%) arrived via search and 100,800 (70%) arrived via browse - the category menus, the homepage rails, the listing grids. The search road converts at 12%. The browse road converts at 4.5%. That is a 2.67x gap between two groups of shoppers looking at the same products, at the same prices, in the same month.
LiveReading the two roads like a leader6 min▶
Three things this picture tells you before anyone opens a dashboard:
- The minority road carries outsized money. Search is 30% of product-page traffic but produces 5,184 of the 9,720 orders - more than half. Small road, heavy freight.
- The majority road is where most shoppers actually live. 70% of everyone who ever sees a product got there by browsing. Whatever you think of its 4.5% conversion, this road IS the store for most visitors.
- The gap is a property of the shopper, not just the store. The same site serves both roads. The 2.67x difference is mostly about who chooses each road - which is exactly why "make browse convert like search" is usually the wrong goal. Part 2 explains.
Self-studyWhere these numbers come from2 min read▶
The analyst track computes this split in b3 using GA4's path exploration - the report that draws the actual sequences of screens users walk, forward from a search or backward from a purchase. If you want to see the machinery, that session is the one to sit in on. For this track you only need the shape: two roads, 30/70 traffic, 12% vs 4.5%, orders 5,184 + 4,536 = 9,720 exactly.
Typing is a declaration of intent 15 min live
The search box is a confession booth. To use it, a shopper must already know - and be willing to type - what they want. "Linen midi dress" is a shopper halfway to a decision; a shopper tapping through "New in" is still deciding whether to decide. The conversion gap between the roads is mostly this self-selection of intent, and only partly the quality of your search engine. Both halves matter, but confusing them leads to bad investments.
LiveThe honest numbers: what the industry actually knows6 min▶
You will hear dramatic claims about search users. Here is the range, honestly flagged - all of it is vendor and analyst data, not academic research:
| Source | Claim | Weight it as |
|---|---|---|
| Econsultancy | Search users convert ~1.8x browse users | Analyst study - the modest end |
| Forrester (commonly cited) | 2-3x browse conversion | The mid-range most teams see |
| Algolia-relayed figures | Amazon 6x · Walmart 2.4x · Etsy 3x | Vendor-relayed - directional only |
So the defensible sentence is: "vendor and analyst studies put search users at 1.8-6x browse conversion." Mango Lane's 2.67x sits comfortably mid-range - which is reassuring, not remarkable. If someone pitches you a search vendor using only the 6x number, they picked the most flattering point on a wide range.
LiveThe search road's failure modes - what leaders should ask about5 min▶
Precisely because searchers arrive with declared intent, failing them is expensive - you are dropping the shoppers most ready to buy. Three failure modes account for most of the loss:
- Zero-result searches. The shopper typed what they want and the store said "nothing found". Sometimes you truly do not stock it (useful demand signal - see Part 3); often you do stock it and the engine missed it.
- Misspellings. "Linnen dress" should find linen dresses. An engine without forgiving matching quietly zeroes out a slice of your highest-intent traffic.
- Synonym gaps. The shopper says "jumper", your catalog says "sweater". Same garment, zero results. Every market and generation brings its own vocabulary; the catalog only has one.
The zero-result report is the cheapest audit in ecommerce. Most analytics setups already record every search term (GA4 does it automatically). Sorting for terms that returned nothing takes an analyst an hour and routinely finds both fixable engine gaps and products worth stocking.
Two roads, two maintenance crews 10 min live
Because the roads serve different shoppers, they need different investments - and usually different owners. Funding "site improvements" as one bucket guarantees the louder road wins. Name the two crews explicitly:
LiveThe two maintenance lists5 min▶
| Road | What it needs | What failure looks like |
|---|---|---|
| Search (30% of PDP traffic, 12% CR) | Relevance (right products, right order) · autocomplete that guides the query · zero-result rescue (forgiving matching, synonyms, "did you mean") | High-intent shoppers hitting "no results" and leaving |
| Browse (70% of PDP traffic, 4.5% CR) | Category logic that matches how shoppers think · merchandising order (what leads each grid) · PLP scannability (can you judge 20 products in 10 seconds?) | Shoppers scrolling grids without ever tapping into a product |
Note what is NOT on either list: making browse shoppers behave like search shoppers. Browse conversion of 4.5% on 70% of traffic is not a bug to eliminate - it is the browsing population being itself. The browse crew's job is a better 4.5%, not a fake 12%.
Self-studyThe search bar as market research3 min read▶
Every search is a shopper telling you demand in their own words - unprompted, unfiltered, spelled however they spell it. GA4 records this automatically: a view_search_results event fires with the search_term attached whenever on-site search runs. That log is a free, continuously updated market-research panel: what people call your products, what they want that you do not stock, which trends arrive before your buyers notice. Teams pay agencies for demand studies while this sits unread in their own analytics. One honest caveat from the analyst track: app and single-page-app search that never touches the URL can silently vanish from this data - worth one question to confirm yours is captured.
Three judgment calls you will actually face ★ read-along - no computation
These three proposals cross every ecommerce leader's desk eventually. Walk each one with the two-roads model - the answers are judgments, not calculations, but the model keeps the judgments honest.
LiveCall 1 · "Let's make search more prominent"4 min▶
The pitch: search converts 2.67x better, so push more shoppers onto it - bigger bar, front and center.
The model says: probably yes. Search carries only 30% of product-page sessions; if even some browse-inclined shoppers with formed intent switch roads, they convert better. The 2.67x gap means modest traffic shifts move real money.
The caveat to voice: check the search road can absorb more traffic without degrading. New search users are lighter-intent than current ones (the current 12% is partly self-selection), and the failure modes from Part 2 hit harder at volume. Expect the search CR to dip as its share grows - that can still be a win overall.
LiveCall 2 · "Our search conversion fell"4 min▶
The alarm: search CR dropped from 12% toward 10% over two months. Someone wants to replace the search vendor.
The question to ask first: did the search engine get worse, or did the search population change? A relevance regression and an influx of first-time browsers trying search look identical in the headline number.
How to tell them apart: ask for the CR split by user type (returning searchers vs first-time), and for the zero-result rate trend. Regression: zero-results up, everyone's CR down. Mix shift: zero-results flat, returning searchers' CR steady, more low-intent newcomers in the pool - often the harmless side effect of Call 1 succeeding.
LiveCall 3 · "Kill the category menu - everyone searches now"4 min▶
The pitch: search converts better, modern shoppers search, the menu is clutter. Strip the navigation, celebrate the minimalism.
The model says: no. 70% of product-page reach still comes through the browse road. Removing the category menu does not convert browsers into searchers - many do not yet know what to type. It converts them into exits.
The reframe to offer: the browse road being slower does not make it optional - it is how most shoppers form the intent that search then harvests. Fund the browse list from Part 3 instead of demolishing the road.
Two asks - this week ◐ two emails, no computation
- Ask your team for your search-vs-browse split of product-page traffic, and the conversion rate of each road. Compare against Mango Lane's shape: 30/70 traffic, 12% vs 4.5%, 2.67x apart.
- Ask for the top 20 zero-result search terms this month. Mark each one: do we stock it (engine gap) or not (demand signal)?
Sources this session draws on
Leader sessions cite the same sources the analyst track verifies - with honesty flags kept visible:
Three questions before you go 🎯 ◐ 90 seconds
1 · Searchers convert at 12%, browsers at 4.5%, on the same store. What mostly explains the gap?
The gap is mostly a property of who chooses each road, not the roads' engineering. That is why "make browse convert like search" is the wrong goal - and why new search traffic converts below the current 12%.
2 · Search conversion fell over two months. What should a leader ask before approving a new search vendor?
The headline CR mixes engine quality with who is searching. Zero-result rate trend plus a returning-vs-new split separates the two - and a mix shift can be the harmless side effect of successfully promoting search.
3 · "Everyone searches now - kill the category menu." Why is this wrong at Mango Lane?
Search's higher CR does not mean browse is disposable. Most shoppers form intent BY browsing; search harvests intent that browse helped create. Fund the browse road, do not demolish it.