A/B-testiranje in optimizacija

Kako testirati kratke in dolge emaile spletne trgovine

Posted on July 18, 2026 by

Kratke proti dolgim testirajte tako, da najprej odločite, kje naj se zgodi odločitev za nakup. Kratek email potisne odločitev na vaše spletno mesto: ena ideja, ena povezava, minimalno branja. Dolg email skuša skleniti prodajo kar v nabiralniku: koristi, ugovori, družbeni dokaz, nato povezava. Obe različici istega emaila zgradite okoli te razlike, občinstvo razdelite 50/50 in presojajte po prihodku na prejemnika — nikoli zgolj po klikih, ker kratki emaili skoraj vedno pobirajo klike, hkrati pa pošiljajo manj prepričane obiskovalce. Pričakujte, da bo odgovor odvisen od vrste emaila in cenovnega razreda: opomnik o vrnitvi na zalogo za 15 € redko potrebuje 400 besed, odločitev za prvi nakup za 900 € pa se redko sklene z dvovrstičnim sporočilom. Ta članek obravnava, kako ta test izvesti čisto in rezultat prebrati pošteno.

Težava: dolžino odloči navada, ne dokazi

Večina trgovin dolžine svojih emailov ni nikoli izbrala. Prišla je iz predloge ali od tistega, ki je pred tremi leti napisal prvo kampanjo, in vsak email odtlej je podedoval to obliko. Medtem lastnik sumi, da so emaili predolgi — ali pretanki — na podlagi ničesar drugega kot občutka in kakšne odjave, ki je zbodla.

Vložek je resničen na obeh straneh. Predolg, in bralci na telefonih odnehajo dva odstavka noter, nikoli ne dosežejo gumba. Prekratek, in oklevajoč kupec klikne skozi neprepričan, pristane na strani izdelka, ki ne odgovori na nobenega od njegovih ugovorov, in odide. Oba neuspeha izgledata enako v vašem poročilu o prihodku: email, ki podpovprečno deluje iz razlogov, ki jih ne vidite.

Zakaj je “naj bo kratko” slab privzeti nasvet

Najbolj ponavljano pravilo emailnega marketinga — nihče ne bere dolgih emailov — je napol resnično in slabo naciljano. Nihče ne bere dolgih dolgočasnih emailov. Ljudje z veseljem preberejo 600 besed o nečem, kar dejavno razmišljajo kupiti, še zlasti, ko je nakup drag, nepoznan ali darilo, pri katerem se bojijo, da bi zgrešili.

Dolžina deluje kot predkvalifikacija. Dolg email filtrira svoje klike: bralec, ki se prebije skozi vašo razlago velikosti, zgodbo o materialih in politiko vračil ter nato klikne, je veliko bliže nakupu kot nekdo, ki je pritisnil gumb štiri sekunde v dvovrstičnem emailu. Manj klikov, toplejši kliki. To pomeni, da vam kratka različica, ki zmaga v stolpcu klikov, pove skoraj nič — in če zmagovalce kronate po stopnji klikov, boste sistematično krajšali vsak email v svojem programu, medtem ko prihodek tiho hodi vštric.

Kako odločiti, kaj “kratko” in “dolgo” pomenita, preden testirate

Test dolžine se sesuje v godljo, ko se različici razlikujeta v sporočilu, ne le v dolžini. Ohranite disciplino: ista ponudba, isti izdelki, ista zadeva, isti en poziv k dejanju, ista ciljna stran. Dolga različica podaljša kratko — doda razloge, odgovori na ugovore, morda navedek stranke — namesto da bi govorila druge stvari. Če hkrati spremenite argument in dolžino, ste prekršili pravilo ene spremembe in rezultat vam ne bo povedal, kaj je bilo pomembno; kako testirati po eno spremenljivko naenkrat pojasnjuje škodo, ki jo to naredi.

Dve mejni opombi, saj ima ta test bližnje sosede. Če je vaša “kratka” različica oskubljeno, osebno izgledajoče sporočilce, “dolga” pa polna oblikovana predloga, dejansko izvajate test formata — to je kako testirati preproste tekstovne emaile v primerjavi z oblikovanimi in odgovarja na drugo vprašanje. Tu naj bosta obe različici v istem vizualnem slogu. In pri tem se uprite spreminjanju besedila gumba med različicama; pozivi k dejanju so svoj poskus, obravnavan v A/B-testiranju pozivov k dejanju v emailih spletne trgovine.

Praktično pravilo za same različice: naj bosta pomembno različni. Testiranje 120 besed proti 150 ne testira ničesar. Nekaj takega kot 80–120 besed proti 350–500 da formatoma prostor, da se dejansko različno vedeta (ilustrativni razponi — prilagodite kompleksnosti svojega izdelka).

Kje izvesti prvi test dolžine

Ni vsak email vreden testiranja. Prednost dajte tam, kjer dolžina verjetno spremeni izid:

Dragi ali premišljeni nakupi. Če je vaše povprečno naročilo 150 € ali več ali izdelek potrebuje razlago — nega kože z aktivnimi sestavinami, tehnična oprema, pohištvo — ima dolga različica pravo priložnost.
Prvi email opuščene košarice. Stranka je izdelek že izbrala; odprto vprašanje je, ali je treba odgovoriti na oklevanje (stroški dostave, vračila, velikosti) ali zadostuje hiter sunek.
Email dobrodošlice. Novi naročniki še ne poznajo vaše zgodbe. To je edino mesto, kjer daljši email nosi informacije, ki jih nobeno drugo sporočilo ne bo prišlo dostaviti.

Testiranja dolžine ne izvajajte na transakcijsko obarvanih sporočilih — potrditvah dostave, opomnikih za dopolnitev potrošnega blaga. Ta imajo eno nalogo in kratka oblika ji ustreza; testirajte najprej drugje.

Izvajanje testa, ne da bi se zavajali

Razdelite 50/50 znotraj istega pošiljanja. Isti segment, isti trenutek. Vsaka druga ureditev testira čas ali občinstvo, ne dolžine.
Presojajte po prihodku na prejemnika, s stopnjo klik-do-naročila kot diagnostiko spodaj. Vzorec, ki ga iščete: če dolga dobi manj klikov, a veliko višji delež kliknjevalcev, ki kupijo, dodatne besede opravljajo pravo prodajo. Če dolga dobi manj klikov in enako stopnjo klik-do-naročila, so besede zgolj trenje — izrežite jih.
Pustite ga teči do pravega vzorca. Testi, presojani po naročilih, potrebujejo potrpljenje. Kampanja, razdeljena med 10.000 ljudi pri 1,5-odstotni stopnji naročil, vam da približno 75 naročil na krak za primerjavo (ilustrativno); razlike peščice naročil so šum. Vse presenetljivo ponovite, preden ukrepate, in glejte kako dolgo naj traja A/B-test emaila spletne trgovine za razlogovanje o trajanju.
Preverite mobilni predogled dolge različice. Na telefonu je 450 besed zid. Če dolga različica zakoplje svoj gumb pod petnajst drsanj s palcem, gumb premaknite gor — poziv k dejanju sredi emaila plus enega na koncu — sicer test postane “gumb, ki ga nihče ni našel, proti gumbu, ki so ga vsi našli”.

Primer trgovine (ilustrativno)

Trgovina, ki prodaja ergonomske pisarniške stole (povprečno naročilo okoli 520 €), je testirala svoj prvi email opuščene košarice. Kratka različica: “Vaš stol je še tu — blagajna vzame minuto”, slika, gumb, približno 60 besed. Dolga različica: isti uvod, nato trije kratki bloki, ki odgovarjajo na dejanska prednakupna oklevanja, potegnjena iz vprašanj podpore — brezplačni 30-dnevni domači preizkus, sestavljanje v desetih minutah, kaj krije garancija — okoli 380 besed, isti gumb.

Po štirih tednih (ilustrativne številke): kratka je zmagala pri klikih, 11 % proti 8 %. Dolga je zmagala pri naročilih, 2,9 % proti 2,1 %, in prihodek na prejemnika je sledil. Trgovina je obdržala dolgo različico, nato pa je — kot ločen nadaljevalni test — poskusila premakniti blok o preizkusnem obdobju nad sliko izdelka, ker so vprašanja podpore nakazovala, da je bil to najmočnejši ugovor. To je vzorec, vreden posnemanja: test dolžine izbere smer, manjši testi optimizirajo znotraj nje. Ko metrike kažejo v različne smeri, kot so tukaj, izbira zmagovalnega emailnega testa obravnava, kako se odločiti brez racionalizacije.

Kako to početi v Omnisendu

V mojih trgovinah to teče skozi Omnisend in nastavitev je enostavna. Kampanje: funkcija A/B razdeli pošiljanje; obe dolžini zgradite kot ločeni različici in prihodek na različico v poročilu preberite sami, namesto da sprejmete zmagovalca po stopnji odpiranja. Tokovi: v avtomatizacijo opuščene košarice ali dobrodošlice spustite korak razdelitve A/B, ena veja na dolžino, in prihodek vej primerjajte v poročilu toka — vedoč, da se testi tokov polnijo počasi, saj v poskus vstopajo le novi sprožilci. Naloga orodja tukaj je čista razdelitev in številka prihodka na vejo. Odločitev, kaj naj dolga različica pove — kateri ugovori si zaslužijo tistih dodatnih 300 besed — še vedno prihaja iz vašega nabiralnika podpore in ocen izdelkov, ne iz kakšne nadzorne plošče.

Vaš naslednji korak

Priklopite svoj prvi email opuščene košarice in preštejte njegove besede. Nato odprite svojih zadnjih deset pogovorov s podporo in naštejte vprašanja, ki jih ljudje postavljajo pred nakupom. Če email ne odgovarja na nobenega, se je vaša dolga različica pravkar napisala sama — zgradite jo ta teden, razdelite 50/50 in pustite, da prihodek na prejemnika izbere zmagovalca v enem mesecu.

A/B Testing and Optimization

How to Test Short vs Long Ecommerce Emails

Posted on July 17, 2026 by

Test short vs long by first deciding where the buying decision should happen. A short email pushes the decision to your website: one idea, one link, minimal reading. A long email tries to close the sale in the inbox: benefits, objections, social proof, then the link. Build both versions of the same email around that distinction, split your audience 50/50, and judge on revenue per recipient — never on clicks alone, because short emails almost always win clicks while sending less-convinced visitors. Expect the answer to differ by email type and price point: a €15 restock reminder rarely needs 400 words, and a €900 first-purchase decision rarely closes off a two-line note. This article covers how to run that test cleanly and read the result honestly.

The problem: length is decided by habit, not evidence

Most stores never chose their email length. It came from the template, or from whoever wrote the first campaign three years ago, and every email since has inherited the shape. Meanwhile the owner suspects the emails are too long — or too thin — based on nothing but gut feel and the occasional unsubscribe that stung.

The stakes are real on both sides. Too long, and readers on phones bail two paragraphs in, never reaching the button. Too short, and a hesitant buyer clicks through unconvinced, lands on a product page that answers none of their objections, and leaves. Either failure looks identical in your revenue report: an email that underperforms for reasons you can’t see.

Why “keep it short” is bad default advice

The most repeated rule in email marketing — nobody reads long emails — is half true and badly aimed. Nobody reads long boring emails. People happily read 600 words about something they’re actively considering buying, particularly when the purchase is expensive, unfamiliar, or a gift they’re nervous about getting wrong.

Length works as pre-qualification. A long email filters its own clicks: the reader who makes it through your sizing explanation, your materials story, and your returns policy and then clicks is far closer to buying than someone who tapped a button four seconds into a two-line email. Fewer clicks, warmer clicks. Which means the short variant winning the click column tells you almost nothing — and if you crown winners on click rate, you will systematically shorten every email in your program while revenue quietly goes sideways.

Deciding what “short” and “long” mean before you test

A length test collapses into mush when the two variants differ in message, not just length. Keep the discipline: same offer, same products, same subject line, same single call to action, same destination page. The long version extends the short one — it adds reasons, answers objections, maybe a customer quote — rather than saying different things. If you change the argument and the length together, you’ve broken the one-change rule and the result won’t tell you which mattered; how to test one variable at a time explains the damage that does.

Two boundary notes, since this test has close neighbors. If your “short” variant is a stripped, personal-looking note and your “long” one is a full designed template, you’re actually running a format test — that’s how to test plain-text vs designed emails, and it answers a different question. Keep both variants in the same visual style here. And resist tweaking the button copy between variants while you’re at it; CTAs are their own experiment, covered in A/B testing calls to action in ecommerce emails.

A practical rule of thumb for the variants themselves: make them meaningfully different. Testing 120 words against 150 is testing nothing. Something like 80–120 words against 350–500 gives the formats room to actually behave differently (illustrative ranges — scale to your product’s complexity).

Where to run the first length test

Not every email is worth testing. Prioritize where length plausibly changes the outcome:

High-value or considered purchases. If your average order is €150+ or the product needs explaining — skincare with active ingredients, technical gear, furniture — the long variant has a genuine shot.
The first abandoned-cart email. The customer already chose the product; the open question is whether hesitation needs answering (shipping costs, returns, sizing) or whether a quick nudge suffices.
The welcome email. New subscribers don’t know your story yet. This is the one place a longer email carries information no other message will get to deliver.

Skip length testing on transactional-adjacent messages — shipping confirmations, replenishment reminders for consumables. Those have one job and a short form fits it; test elsewhere first.

Running the test without fooling yourself

Split 50/50 within the same send. Same segment, same moment. Any other arrangement tests timing or audience, not length.
Judge on revenue per recipient, with click-to-order rate as the diagnostic underneath. The pattern to look for: if long gets fewer clicks but a much higher share of clickers buying, the extra words are doing real selling. If long gets fewer clicks and the same click-to-order rate, the words are just friction — cut them.
Let it run to a real sample. Order-judged tests need patience. A campaign split to 10,000 people at a 1.5% order rate gives you roughly 75 orders per arm to compare (illustrative); differences of a handful of orders are noise. Rerun anything surprising before acting on it, and see how long should an ecommerce email A/B test run for the duration reasoning.
Check the mobile preview of the long variant. On a phone, 450 words is a wall. If the long version buries its button below fifteen thumb-scrolls, move the button up — a mid-email CTA plus one at the end — or the test becomes “button nobody found vs button everybody found.”

A store example (illustrative)

A store selling ergonomic office chairs (average order around €520) tested its first abandoned-cart email. Short variant: “Your chair’s still here — checkout takes a minute,” image, button, roughly 60 words. Long variant: the same opener, then three short blocks answering the actual pre-purchase hesitations pulled from support tickets — free 30-day home trial, assembly in ten minutes, what the warranty covers — around 380 words, same button.

After four weeks (illustrative figures): short won clicks, 11% against 8%. Long won orders, 2.9% against 2.1%, and revenue per recipient followed. The store kept the long version, then — as a separate follow-up test — tried moving the trial-period block above the product image, since support tickets suggested it was the strongest objection. That’s the pattern worth copying: the length test picks the lane, and smaller tests optimize inside it. When metrics point in different directions, as they did here, choosing a winning email test covers how to decide without rationalizing.

Doing this in Omnisend

In my own stores this runs through Omnisend, and the setup is straightforward. Campaigns: the A/B feature splits the send; build both lengths as separate versions and read revenue per variant in the report yourself rather than accepting an open-rate winner. Flows: drop an A/B split step into the abandoned-cart or welcome automation, one branch per length, and compare branch revenue in the flow report — knowing that flow tests fill slowly, since only new triggers enter the experiment. The tool’s job here is the clean split and the per-branch revenue number. Deciding what the long version should say — which objections deserve those extra 300 words — still comes from your support inbox and product reviews, not from any dashboard.

Your next step

Pull up your first abandoned-cart email and count its words. Then open your last ten support conversations and list the questions people ask before buying. If the email answers none of them, your long variant just wrote itself — build it this week, split 50/50, and let revenue per recipient pick the winner in a month.