Odkrivanje izdelkov in spletno trgovsko prikazovanje

Kaj priporočiti, ko si kupec ogleda več nepovezanih izdelkov

Posted on July 18, 2026 by

Ko si nekdo v eni seji ogleda voziček, komplet zvočnikov in litoželezno ponev, teh signalov ne poskušajte povprečiti v eno samo priporočilo — dobili boste zmešnjavo. Prava poteza je, da prepoznate, da razpršeno brskanje običajno pomeni eno od nekaj točno določenih stvari (kupca za darilo, raziskovalni prvi obisk ali več namer v eni seji), in da se odzovete na najbolj svež, najbolj zavezan signal, ne na celoten zamegljen nabor. V praksi: priporočajte glede na zadnjo stvar, s katero se je resno ukvarjal, nadaljnje sporočilo naj bo dovolj široko, da pokrije več kot eno nit, in oprite se na tisto, kar dejansko veste — pretekle nakupe — namesto na hrup ene zmedene seje. Ta stran govori o tem točno določenem primeru: kaj storiti, ko zgodovina brskanja noče povedati jasne zgodbe.

Pravi problem: signali, ki kažejo v pet smeri

Priporočilni sistemi in nadaljnja e-poštna sporočila predpostavljajo, da ima kupec skladno namero — pogledal je tekaške copate, torej mu pokaži tekaško opremo. Preprosto. A resnična seja pogosto ne izgleda tako. Človek pride z oglasa, tava, klika, kar mu pade v oči, odpre tri izdelke, ki ne delijo ne kategorije, ne cenovnega razreda, ne namena uporabe, in odide.

Zdaj mora vaša avtomatizacija sprejeti odločitev, a nima očitnega odgovora. Naj pokaže dodatke za voziček? Morda kupuje darilo in nima otroka. Naj potisne zvočnike? Morda je bila to zgolj radovednost. Naj priporoči ponev? Mogoče. Trgovina, ki vsako brskanje obravnava enako, bo samozavestno, konkretno priporočilo sprožila proti kupcu, čigar signal je bil vse prej kot konkreten — in običajno bo zgrešila. Še huje, lahko deluje vsiljivo, kot da je trgovina kupca opazovala in glasno napačno ugibala.

To prepoznate, če vaša e-poštna sporočila o opuščenem brskanju dobro konvertirajo pri osredotočenih kupcih in odpovedo pri vseh drugih. Ta ploski segment so razpršeni brskalci. So resničen izsek prometa in večina trgovin zanje nima načrta.

Zakaj običajna rešitev — priporočiti »najboljši« izdelek — ne zadošča

Privzeta poteza, ko so signali nejasni, je umik k priljubljenosti: pokaži najbolje prodajane, pokaži najbolje ocenjene, pokaži, kar kupujejo vsi drugi. Zdi se varno. Večinoma je zapravljeno.

Mreža najbolje prodajanih izdelkov prezre edino stvar, ki ste jo izvedeli — da si je ta človek ogledal voziček, zvočnike in ponev, od katerih nobeden ni vaš najbolje prodajani izdelek. Imeli ste šibek signal in ste ga zavrgli za nič signala, preoblečenega v samozavest. Priljubljeni izdelki niso isto kot pravi izdelek za določenega kupca, in privzeto zatekanje k njim je zbita past; razlogovanje je razloženo v zakaj priljubljeni izdelki niso vedno najboljše priporočilo. Povprečenje je prav tako slabo — razdelite razliko med tremi nepovezanimi izdelki in pristanete na nečem, kar ni relevantno nikomur. Zmešnjava, kot obljubljeno.

Razpršeni brskalec potrebuje drugačno logiko, ne glasnejše različice napačne.

Kje odteka priložnost

Puščanje je tu prikrito, ker se skriva znotraj avtomatizacije, ki izgleda, kot da deluje. Vaš potek za brskanje poroča o spodobni skupni stopnji konverzije, zato nihče ne dvomi vanjo. A to povprečje nosijo osredotočeni kupci. Razpršeni — pogosto četrtina do tretjina brskalcev, v trgovinah, ki sem jih pregledal — dobijo napačno usmerjeno e-poštno sporočilo, ga prezrejo in odpadejo iz poteka brez drugega poskusa.

To so povrnljivi ljudje. Dovolj so se vključili, da so si ogledali več izdelkov. Le urejene namere vam niso izročili. Če jih obravnavate kot izgubljeni primer, vsak dan tiho zavržete kos toplega prometa, znotraj poteka, ki si sam prisodi prehodno oceno.

Praktična rešitev: preberite vzorec, nato izberite strategijo

Razpršeno brskanje ni ena sama stvar. Ugotovite, katero od teh gledate, ker vsaka zahteva drugačen odziv.

1. Branje po svežosti in globini. Vsi ogledi niso enaki. Izdelek, ki ga je odprl zadnjega, na katerem se je zadržal najdlje ali h kateremu se je dvakrat vrnil, je pomembnejši od tistega, ki ga je bežno pogledal na začetku. Ko je seja neskladna, obtežite svežost in globino vključenosti ter priporočajte okoli tega izdelka, preostalo pa obravnavajte kot ozadje. Že to samo reši večino zmedenih sej.

2. Vzorec kupca za darilo. Nepovezani izdelki prek kategorij, cenovnih točk in demografskih skupin so klasičen prstni odtis nekoga, ki nakupuje za druge — enega za očeta, enega za partnerja, enega za prijatelja. Če se vzorec ujema, mu ne prodajajte dodatkov za otroka, ki ga nima; pomagajte mu nakupovati kot obdarovalcu. To je strategija zase: kako personalizirati priporočila za kupce daril.

3. Raziskovalni prvi obisk. Nov obiskovalec brez zgodovine, ki vzorči vaš nabor, ne potrebuje natančnega priporočila — potrebuje orientacijo. Pokažite širino: vaš enega ali dva najmočnejša izdelka iz vsake kategorije, ki se jih je dotaknil, tako da mu nadaljnje sporočilo pomaga ponovno najti trgovino, namesto da ugiba njegove misli.

4. Zatecite se k temu, kar dejansko veste. Če je oseba vračajoči se kupec, ena zmedena seja šteje manj kot njena zgodovina nakupov. Nekdo, ki je dvakrat kupil tekaško opremo in danes brskal naključno, je najverjetneje še vedno tekač. Zasidrajte se na trajnem signalu, ne na hrupnem. Gradnja okoli tega trajnega branja je osrednja ideja v gradnja priporočil izdelkov okoli namere kupca.

Rdeča nit: nehajte siliti eno priporočilo na sejo z več namerami. Bodisi izolirajte najmočnejši signal bodisi razširite odziv, da pošteno pokrije več kot eno nit.

Kaj avtomatizirati

Zgradite nadaljnje nagovarjanje pri brskanju, ki se razveji glede na jasnost signala, namesto da se pretvarja, da je vsako brskanje čisto.

Sprožilec: seja z ogledi izdelkov in brez nakupa, enako kot pri katerem koli poteku brskanja.
Delitev segmenta — ključna poteza: ločite »osredotočene« brskalce (ogledi zgoščeni v eni kategoriji ali lastnosti) od »razpršenih« brskalcev (ogledi čez nepovezane kategorije). Usmerite jih k različnim sporočilom.
Za osredotočene brskalce: standardno e-poštno sporočilo »tukaj je, kar ste si ogledali, plus bližnji sorodniki«.
Za razpršene brskalce: sporočilo, zgrajeno na pravilu svežosti in globine — vidno izpostavite zadnji, najbolj vključeni izdelek, z lažjo drugo vrstico, ki pokrije drugo nit. Za vračajoče se kupce v tej košari preglasite sejo in vodite z izbori na podlagi zgodovine.
Časovnica: v enem dnevu, dokler je katera koli od niti še topla.
Kanal: e-pošta, ker potrebujete prostor za širšo, večnitno postavitev, ki je SMS ne zmore dobro nositi.
Vsebina: en jasen glavni izdelek, zmeren podporni nabor in jezik, ki ne obljublja preveč (»nekaj stvari, ki ste si jih ogledovali« se bere bolje kot lažno samozavesten posamezni izbor).
Cilj: klik nazaj na relevanten izdelek — merjen ločeno za razpršeni segment, tako da dejansko vidite, ali ste rešili tisti del, ki je prej odpovedal.

Dobro branje teh vzorcev je odvisno od tega, kako sploh razlagate podatke o brskanju, kar je veščina zase: kako vedenje pri brskanju uporabiti za boljše odkrivanje izdelkov.

Primer trgovine (ponazoritveni)

Splošna trgovina z izdelki za dom dobi obiskovalca, ki po vrsti odpre: okvir za slike, stiskalnico za česen, dišečo svečo in nato volneno odejo — pravi čas nameni le odeji in se pomika po njenih ocenah. Štirje nepovezani izdelki. Povprečilni sistem bi izpostavil mrežo »priljubljeno za dom« in zgrešil.

Pametnejši potek prebere svežost in globino: odeja je zmagala v seji. Naslednje jutranje e-poštno sporočilo vodi s to odejo, ob njej doda dve deki in usklajeno blazino ter spusti majhno vrstico »pritegnilo je tudi vaše oko« s svečo. Okvir in stiskalnica za česen — bežni pogledi, ne namera — sta izpuščena. Kupec je prišel po prijeten dnevni prostor, ne da bi to prav vedel, in e-poštno sporočilo mu je to povedalo nazaj. (Ponazoritveno — razvejitev zgradite na svojih podatkih o kategorijah in vključenosti.)

Kako izmeriti, ali deluje

Tega ne sodite po skupni številki poteka — prav ta je skrila problem. Namesto tega:

Razdelite konverzijo po segmentu: osredotočeni proti razpršenim. Stopnja klikov in naročil razpršenega segmenta je metrika, ki jo dejansko poskušate premakniti.
Stopnja klikov na priporočeni izdelek pri glavnem izdelku v e-poštnih sporočilih razpršenim brskalcem — je izbor po svežosti in globini zadel?
Prihodek na prejemnika za razpršeno vejo posebej.
Stopnja odjav in prijav neželene pošte na tej veji — napačno usmerjena priporočila zmedenim kupcem so hiter način, da izgubite seznam, zato jo spremljajte.

In preverite, da dodajate naročila, ne pa jih premešate. Priporočilo, ki zgolj prerazporedi zasluge za prodajo, ki je itak prihajala, si svojega mesta ne prisluži — preizkušanje, ali priporočila povečujejo prihodek ali ga zgolj premeščajo pokaže, kako to ugotoviti.

Kako se v to umešča Omnisend

Zgornja logika razvejitve — osredotočeni proti razpršenim, obtežitev svežosti, preglasitev z zgodovino — je zapleteno napeljati ročno, in prav tu si platforma za avtomatizacijo zasluži svoj kruh. Omnisend uporabljam v vseh svojih trgovinah, potem ko sem ga preizkusil proti Klaviyu, in ostal zaradi vizualnih razvejitev avtomatizacije, segmentacije na podlagi vedenja ter e-pošte in SMS-a na enem mestu. Delitev segmenta lahko zgradite na ogledanih kategorijah, izpostavite dinamične bloke izdelkov, vezane na nedavne oglede, in se za znane kupce zatečete k priporočilom na podlagi zgodovine nakupov.

Odkrito povedano: orodje izvaja logiko, ki jo zasnujete, a ne more odločiti, katera od petih niti kupca je bila prava — ta presoja živi v tem, kako nastavite pravila in berete svoje podatke. In nobena razvejitev ne reši nadaljnjega sporočila, če izdelki v ozadju niso vredni vrnitve. Omnisend je partner Shopimationa preko affiliate programa, priporočen iz vsakdanje uporabe; brezplačni paket zadošča za preizkus poteka brskanja z dvema vejama na živem segmentu, preden ga razširite.

Vaš naslednji korak

Odprite svoj potek za opuščeno brskanje in zastavite eno vprašanje: ali razpršeno, večkategorijsko sejo obravnava enako kot osredotočeno? Če jo — in večina potekov jo — zgradite eno samo najbolj dragoceno delitev: osredotočeni proti razpršenim, pri čemer razpršena veja vodi z zadnjim, najbolj vključenim izdelkom. Izmerite to vejo posebej. Nato izostrite celoten pristop z gradnja priporočil izdelkov okoli namere kupca.

Product Discovery and Merchandising

What to Recommend When a Shopper Views Several Unrelated Products

Posted on July 18, 2026 by

When someone views a stroller, a set of speakers, and a cast iron pan in one session, don’t try to average those signals into a single recommendation — you’ll get mush. The right move is to recognize that scattered browsing usually means one of a few specific things (a gift shopper, an exploratory first visit, or several intents in one session) and to respond to the most recent, most committed signal rather than the whole blurry set. In practice: recommend against the last thing they engaged with seriously, keep the follow-up broad enough to cover more than one thread, and lean on what you actually know — past purchases — over the noise of one confusing session. This page is about that specific case: what to do when the browse history refuses to tell a clean story.

The real problem: signals that point in five directions

Recommendation engines and follow-up emails assume a shopper has a coherent intent — they looked at running shoes, so show running gear. Easy. But a real session often looks nothing like that. A person lands from an ad, wanders, clicks whatever catches their eye, opens three products that share no category, no price band, no use case, and leaves.

Now your automation has a decision to make and no obvious answer. Show the stroller’s accessories? They might be buying a gift and have no baby. Push the speakers? That might have been idle curiosity. Recommend the pan? Maybe. The store that treats every browse the same will fire a confident, specific recommendation at a shopper whose signal was anything but specific — and it’ll usually miss. Worse, it can feel intrusive, like the store watched them and guessed wrong out loud.

You recognize this if your browse abandonment emails convert well for focused shoppers and fall flat for everyone else. The flat segment is the scattered browsers. They’re a real slice of traffic, and most stores have no plan for them.

Why the usual fix — recommend the “top” product — falls short

The default when signals are unclear is to retreat to popularity: show the bestsellers, show the highest-rated, show what everyone else buys. It feels safe. It’s mostly wasted.

A bestseller grid ignores the one thing you did learn — that this person looked at a stroller, speakers, and a pan, none of which is your bestseller. You had faint signal and you threw it away for zero signal dressed up as confidence. Popular products aren’t the same as the right product for a given shopper, and defaulting to them is a well-worn trap; the reasoning is laid out in why popular products are not always the best recommendation. Averaging is just as bad — split the difference between three unrelated products and you land on something relevant to no one. Mush, as promised.

The scattered browser needs a different logic, not a louder version of the wrong one.

Where the opportunity leaks

The leak here is subtle because it hides inside an automation that looks like it’s working. Your browse flow reports a decent overall conversion rate, so nobody questions it. But that average is carried by the focused shoppers. The scattered ones — often a quarter to a third of browsers, in stores I’ve looked at — get a mistargeted email, ignore it, and churn out of the flow with no second attempt.

Those are recoverable people. They engaged enough to view multiple products. They just didn’t hand you a tidy intent. Treat them as a lost cause and you’re discarding a chunk of warm traffic every single day, quietly, inside a flow that grades itself a pass.

The practical solution: read the pattern, then pick a strategy

Scattered browsing isn’t one thing. Diagnose which of these you’re looking at, because each wants a different response.

1. The recency-and-depth read. Not all views are equal. The product they opened last, spent the longest on, or returned to twice outranks the one they glanced at first. When a session is incoherent, weight recency and engagement depth and recommend around that item, treating the rest as background. This alone rescues most confusing sessions.

2. The gift-shopper pattern. Unrelated products across categories, price points, and demographics is the classic fingerprint of someone shopping for other people — one for dad, one for a partner, one for a friend. If the pattern fits, don’t sell them accessories for a baby they don’t have; help them shop as a gifter. That’s a whole strategy of its own: how to personalize recommendations for gift shoppers.

3. The exploratory first visit. A new visitor with no history who’s sampling your range doesn’t need a precise recommendation — they need orientation. Show breadth: your strongest one or two items from each of the categories they touched, so the follow-up helps them re-find the store rather than guessing their mind.

4. Fall back to what you actually know. If the person is a returning customer, one messy session matters less than their purchase history. Someone who’s bought running gear twice and browsed randomly today is still, most likely, a runner. Anchor on the durable signal, not the noisy one. Building around that durable read is the core idea in building product recommendations around customer intent.

The through-line: stop forcing one recommendation onto a multi-intent session. Either isolate the strongest signal, or widen the response to cover more than one thread honestly.

What to automate

Build a browse follow-up that branches on signal clarity instead of pretending every browse is clean.

Trigger: a session with product views and no purchase, same as any browse flow.
Segment split — the key move: separate “focused” browsers (views concentrated in one category or attribute) from “scattered” browsers (views spanning unrelated categories). Route them to different messages.
For focused browsers: the standard “here’s what you looked at, plus close relatives” email.
For scattered browsers: a message built on the recency-and-depth rule — feature the last, most-engaged product prominently, with a lighter secondary row covering a second thread. For returning customers in this bucket, override the session and lead with history-based picks.
Timing: within a day, while any of the threads is still warm.
Channel: email, because you need room for a broader, multi-thread layout that SMS can’t carry well.
Content: one clear hero item, a modest supporting set, and language that doesn’t over-claim (“a few things you were looking at” reads better than a falsely confident single pick).
Goal: a click back to a relevant product — measured separately for the scattered segment, so you can actually see whether you’ve solved the part that used to fail.

Reading these patterns well depends on how you interpret browse data in the first place, which is its own skill: how to use browsing behavior to improve product discovery.

A store example (illustrative)

A general homewares store gets a visitor who opens, in order: a picture frame, a garlic press, a scented candle, and then a wool throw — spending real time only on the throw, and scrolling its reviews. Four unrelated products. An averaging engine would surface a “popular in home” grid and miss.

The smarter flow reads recency and depth: the throw won the session. Next morning’s email leads with that throw, adds two blankets and a matching cushion beside it, and drops a small “also caught your eye” row with the candle. The frame and garlic press — glances, not intent — are dropped. The shopper came for a cozy living room without quite knowing it, and the email said so back to them. (Illustrative — build the branching on your own category and engagement data.)

How to measure whether it’s working

Don’t judge this on the flow’s overall number — that’s what hid the problem. Instead:

Split conversion by segment: focused vs scattered. The scattered segment’s click and order rate is the metric you’re actually trying to move.
Click-to-recommended-product rate on the hero item in scattered-browser emails — did the recency-and-depth pick land?
Revenue per recipient for the scattered branch specifically.
Unsubscribe and spam rate on that branch — mistargeted recommendations to confused shoppers are a fast way to lose the list, so watch it.

And check you’re adding orders, not reshuffling them. A recommendation that only reassigns credit for a sale that was coming anyway isn’t earning its place — testing whether recommendations increase revenue or just shift it shows how to tell.

How Omnisend fits

The branching logic above — focused vs scattered, recency weighting, history override — is fiddly to wire by hand, which is where an automation platform earns its keep. I use Omnisend across my own stores after testing it against Klaviyo, and stayed for the visual automation splits, the behavior-based segmentation, and email-plus-SMS in one place. You can build the segment split on viewed categories, feature dynamic product blocks tied to recent views, and fall back to purchase-history recommendations for known customers.

Being straight about it: the tool executes the logic you design, but it can’t decide which of a shopper’s five threads was the real one — that judgment lives in how you set the rules and read your data. And no branching saves a follow-up if the underlying products aren’t worth returning for. Omnisend is an affiliate partner of Shopimation, recommended from daily use; the free tier is enough to test a two-branch browse flow on a live segment before you widen it.

Your next step

Open your browse abandonment flow and ask one question: does it treat a scattered, multi-category session the same as a focused one? If it does — and most flows do — build the single most valuable split first: focused vs scattered, with the scattered branch leading on the last, most-engaged product. Measure that branch on its own. Then sharpen the whole approach with building product recommendations around customer intent.