Naročniški in potrošni izdelki

Kako predvideti, kdaj je kupec pripravljen ponovno naročiti potrošni izdelek

Posted on July 19, 2026 by

Ponovni nakup napoveš tako, da izračunaš, kako dolgo kupcu zaloga dejansko zdrži, in ga nato dosežeš tik preden mu zmanjka. Pri večini potrošnih izdelkov za to potrebuješ tri podatke: koliko je kupil, kako hitro to količino porabi in kdaj je nazadnje naročil. Velikost pakiranja pomnoži z realno stopnjo porabe, prištej datum nakupa in dobiš oceno datuma, ko bo temu konkretnemu človeku izdelka zmanjkalo. Doseži ga malo pred tem, ne na koledarju, ki si ga izbral za vse. Preostanek tega vodnika govori o tem, kako vsakega od teh treh podatkov ujeti pravilno, saj je naivna različica — “vsi ponovno naročijo vsakih 30 dni” — točno tam, kjer večina trgovin izgubi ponovljeno prodajo.

Situacija, ki jo to rešuje

Prodajaš nekaj, kar ljudje porabijo in kupijo znova: kavo, prehranska dopolnila, hrano za hišne ljubljenčke, kozmetiko, filtre, čistila za dopolnjevanje, kontaktne leče. Prvo naročilo je bilo lahko — oglas, akcija, dobra stran izdelka so opravili svoje. Drugo naročilo je tam, kjer je denar, in ta pušča. Nekateri kupci ponovno naročijo sami. Mnogi ne, ne zato, ker jim izdelek ne bi bil všeč, ampak ker jim je zmanjkalo v torek, nameravali so ponovno naročiti, pa so v trgovini pograbili nadomestek. Ti tega nisi nikoli izvedel. Videl si samo kupca, ki je enkrat kupil in potem utihnil.

Ta tiha vrzel je celoten problem. Potrošni izdelek ima drugo prodajo vgrajeno vase, in če te ni tam v trenutku, ko je kozarček skoraj prazen, bo tam konkurent ali polica v supermarketu.

Zakaj “opomni vse pri 30 dneh” ni dovolj dobro

Očitna rešitev je časovni opomnik: vsak, ki je kupil, dobi e-pošto “čas je za dopolnitev zaloge” fiksno število dni pozneje. Bolje je kot nič in raje vidim, da to poženeš, kot da čakaš na nekaj popolnega. A obravnava intenzivnega uporabnika in redkega uporabnika kot isto osebo, kar pa nista.

Vzemi proteinski prašek, ki enemu človeku zdrži tri tedne, drugemu pa dvanajst. Obema pošlji opomnik na 30. dan in za prvega kupca zamujaš — ta je že ostal brez in kupil drugje — pri drugem pa si prezgoden, saj nadleguješ nekoga, ki ima še dve tretjini lončka. Prezgodnji opomniki se ignorirajo ali, še huje, se izučijo kot šum. Prepozni pridejo, ko je odločitev že sprejeta. En sam interval je lahko pravilen le za povprečnega kupca, povprečni kupec pa je statistična izmišljotina. Obsežnejšo različico te utemeljitve najdeš v zakaj fiksni opomniki za ponovni nakup ne delujejo, ko se poraba razlikuje, na kratko pa: poraba se preveč razlikuje, da bi se ji prilegala ena sama številka.

Trije podatki in kje se vsak skriva

Napovedovanje je aritmetika, ne čarovnija. Potrebuješ tri stvari.

Koliko so kupili. To je podatek, ki ga ljudje preskočijo, in je najbolj uporaben. Nekdo, ki kupi trimesečno pakiranje, ne sme biti na isti uri kot nekdo, ki kupi enkratno. Količina naročila in velikost pakiranja že sedita v tvojih podatkih o naročilih. Dobra uporaba teh podatkov je svoja tema — uporaba količine naročila za oceno naslednjega datuma nakupa gre globlje, kot lahko jaz tukaj.

Kako hitro se porabi. Količino pomnoži s stopnjo porabe. Za veliko izdelkov lahko smiselno izhodiščno stopnjo dobiš z embalaže ali iz običajne rabe — steklenička s 60 kapsulami, ki jih jemlješ dvakrat na dan, zdrži približno 30 dni; 250-mililitrski serum z odmerjanjem na potisk zdrži kake dva meseca. Začni tam, potem pa naj to popravi resnično vedenje (več o tem spodaj).

Kdaj so nazadnje naročili. Ura se začne ob dostavi, ne ob časovnem žigu naročila, če se ta dva razlikujeta za več kot dan ali dva. Pri naročninah in počasnih pošiljateljih ta razlika šteje.

Sestavi jih skupaj in ocena za enega kupca je: datum zadnjega naročila + (količina ÷ stopnja porabe). To je datum, ko mu zmanjka. Tvoja naloga je, da sporočilo dostaviš malo pred njim.

Naj resnično vedenje popravi tvojo prvo domnevo

Tvoja stopnja porabe, izračunana z embalaže, je hipoteza, ne dejstvo. Najboljši signal za to, kako hitro kupec dejansko porabi nekaj, je, kako hitro je nazadnje ponovno naročil. Če je tvoj model ugibal 30 dni, oseba pa se je vrnila pri 24, ga porablja hitreje od povprečja — skrajšaj njeno naslednjo napoved. Če se je vrnila pri 45, jo podaljšaj. Po drugem ali tretjem naročilu ne ugibaš več z embalaže; bereš resnični ritem tega človeka.

Zato sta napoved za prvo naročilo in napoved za ponovnega kupca dve različni igri. Pri prvem kupcu se opreš na velikost pakiranja in embalažo. Pri vračajočem se kupcu se opreš na njegovo lastno zgodovino, ki je veliko natančnejša. Metrike za dopolnjevanje, ki napovedujejo naslednje naročilo, se je vredno naučiti, če želiš to izpeljati natančno namesto na pol.

En pošten pridržek: nekateri ljudje se založijo, nekateri pustijo, da stvari povsem poidejo, velikost gospodinjstva pa spreminja porabo na načine, ki jih ne vidiš. Napoved obravnavaj kot dobro stavo, ne kot gotovost. Poskušaš biti približno pravilen ob pravem času, ne natančno pravilen.

Izdelan primer (ilustrativno)

Recimo, da prodajaš pasjo hrano, ki pride v 2-kilogramski in 7-kilogramski vreči. Kupec kupi 7-kilogramsko vrečo prvega v mesecu za srednje velikega psa, ki poje približno 200 g na dan. Ta vreča zdrži 35 dni. Datum, ko mu zmanjka, pade okoli 5. naslednjega meseca. Opomnik želiš v njegovem nabiralniku okoli 30. dne — pet dni rezerve, da naroči in prejme, preden je skleda prazna.

Zdaj drugi kupec kupi isto 7-kilogramsko vrečo, a ima dva psa. Poraba se podvoji, vreča zdrži ~17 dni, opomnik na 30. dan pa bi ga dosegel skoraj dva tedna po tem, ko mu je zmanjkalo. Isti izdelek, isto pakiranje, popolnoma drugačna ura. Če bi obema poslal isto generično e-pošto na 30. dan, bi enega zadel, drugega pa izgubil. Številke tukaj so ilustrativne — vstavi svoje izdelke in odmerke — a oblika velja za vse, kar ljudje porabijo.

Kaj avtomatizirati

Aritmetiko spremeni v tok. Sestavni deli:

Sprožilec: naročilo, ki vsebuje potrošni izdelek (te izdelke označi, da tok ve, katera naročila štejejo).
Časovnica: zamik, nastavljen na ocenjeni datum pomanjkanja minus rezerva, kjer ocena uporablja velikost pakiranja in, pri ponovnih kupcih, njihov lastni predhodni interval — ne ene same trdo zakodirane številke za celotno trgovino.
Segment: razdeli vsaj na intenzivne in redke uporabnike, ko imaš drugo naročilo za primerjavo, da se zamik prilagodi. To pravilno izpeljati pomeni kako segmentirati kupce po dejanskem vedenju ponovnega naročanja.
Kanal: za večino e-pošta; SMS dodaj le, ko je časovnica napeta in izdelek res spada v kategorijo “ostati brez pomeni trpeti” (kontaktne leče, izdelki blizu zdravil, hrana za ljubljenčke).
Vsebina: hiter, uporaben pomik, ne trda prodaja. Kaj naj to sporočilo dejansko pove, je tema zase — glej kaj poslati, preden kupcu izdelka zmanjka.
Cilj: ponovni nakup, s čistim izhodom, tako da vsak, ki kupi, neha prejemati opomnike.

Bistvo avtomatizacije je, da ponovno izračuna za vsako osebo posebej, namesto da se sproži po skupnem koledarju.

Kako izmeriti, ali je napoved kaj vredna

Spremljaj štiri stvari:

Stopnjo ponovnih nakupov med kupci, ki so vstopili v tok, v primerjavi s tistimi, ki niso. To je semafor.
Čas do ponovnega nakupa glede na tvoj napovedani datum pomanjkanja. Če se ponovni nakupi zgostijo teden dni po tvojem opomniku, pošiljaš prezgodaj; če se zgostijo pred njim, prepozno. Prilagodi rezervo.
Prihodek na prejemnika v toku, da vidiš, ali zasluži več, kot stane zaradi utrujenosti seznama.
Odjave in pritožbe zaradi neželene pošte pri pošiljanjih za dopolnjevanje. Skok pomeni, da je tvoja časovnica napačna in se ljudje počutijo nadlegovane.

Lepota toka, ki temelji na napovedi, je, da ti pove, kdaj se moti: vrzel med napovedanimi in dejanskimi datumi ponovnega nakupa je gumb, ki ga lahko vsako četrtletje naravnaš.

Kje se umesti Omnisend

Vsako spodobno orodje za e-pošto za e-trgovino zna poslati zakasnjeno e-pošto po nakupu. Za to pa želiš platformo, ki ti omogoča razvejanje glede na oznake izdelkov in zgodovino kupca, tako da se zamik in vsebina spreminjata s kupcem, namesto da ostajata fiksna. V svojih trgovinah uporabljam Omnisend — nanj sem prešel, potem ko sem ga preizkusil proti Klaviyu — ker je bila izgradnja toka, ki se sproži ob označenem potrošnem izdelku, deli po segmentu kupcev in v opomnik vstavi nazadnje kupljen izdelek, stvar vlečenja blokov naokoli, ne prijave na podporo.

Pošteni pridržki: nobeno orodje ne pozna resnične porabe kupca ob prvem naročilu; ve samo, kaj je kupil. Natančnost pride po drugem naročilu, ko model nahraniš z njegovim dejanskim intervalom. In opomnik ne more rešiti izdelka, ki ljudem ni bil všeč. Omnisend je partner Shopimation prek pridruženega programa in ga priporočam iz vsakodnevne uporabe — brezplačna raven je dovolj, da na majhnem segmentu zgradiš in preizkusiš en tok za dopolnjevanje, preden ga razširiš.

Tvoj naslednji korak

Vzemi svoje tri najboljše potrošne izdelke in za vsakega zapiši velikost pakiranja in realno dnevno porabo z embalaže. To ti da prvo oceno datuma pomanjkanja, okoli katere lahko že danes zgradiš tok. Nato se odloči, kaj ta prvi opomnik dejansko pove, tako da prebereš kaj poslati, preden kupcu izdelka zmanjka, in ko boš imel podatke za drugo naročilo, zategni celoto z resničnimi segmenti.

Subscription and Consumable Products

How to Predict When a Customer Is Ready to Reorder a Consumable Product

Posted on July 18, 2026 by

You predict a reorder by working out how long the customer’s supply actually lasts, then reaching them shortly before it runs out. For most consumable products that means three inputs: how much they bought, how fast that quantity gets used, and when they last ordered. Multiply the pack size by a realistic usage rate, add the purchase date, and you get an estimated run-out date for that specific person. Reach them a little before it, not on a calendar you picked for everyone. The rest of this guide is about getting each of those three inputs right, because the naive version — “everyone reorders every 30 days” — is where most stores lose the repeat sale.

The situation this solves

You sell something people finish and buy again: coffee, supplements, pet food, skincare, filters, cleaning refills, contact lenses. The first order was easy — an ad, a promo, a good product page did the work. The second order is where the money is, and it’s leaking. Some customers reorder on their own. Plenty don’t, not because they disliked the product, but because they ran out on a Tuesday, meant to reorder, and grabbed a substitute at the shop instead. You never heard about it. You just saw a customer who bought once and went quiet.

That silent gap is the whole problem. A consumable has a built-in second sale baked into it, and if you’re not there at the moment the jar is nearly empty, a competitor or a supermarket shelf will be.

Why “remind everyone at 30 days” isn’t good enough

The obvious fix is a timed reminder: everyone who bought gets a “time to restock” email a fixed number of days later. It’s better than nothing, and I’d rather you ship that than wait for something perfect. But it treats a heavy user and a light user as the same person, and they aren’t.

Take a protein powder that lasts one person three weeks and another twelve. Send both a reminder at day 30 and you’re late for the first customer — they already ran out and rebought elsewhere — and early for the second, nagging someone with two-thirds of a tub left. Early reminders get ignored or, worse, trained as noise. Late ones arrive after the decision’s been made. A single interval can only be right for the average customer, and the average customer is a statistical fiction. There’s a fuller version of this argument in why fixed replenishment reminders fail when usage varies, but the short version is: usage varies too much for one number to fit.

The three inputs, and where each one hides

Prediction is arithmetic, not magic. You need three things.

How much they bought. This is the input people skip, and it’s the most useful one. Someone who buys a three-month pack should not be on the same clock as someone who buys a single. Order quantity and pack size are sitting in your order data already. Using them well is its own topic — using order quantity to estimate the next purchase date goes deeper than I can here.

How fast it gets used. Multiply quantity by a consumption rate. For a lot of products you can get a sensible starting rate from the label or common use — a 60-capsule bottle taken twice daily lasts about 30 days; a 250ml serum with a pump dose lasts roughly two months. Start there, then let real behavior correct it (more on that below).

When they last ordered. The clock starts at delivery, not at the order timestamp, if the two differ by more than a day or two. For subscriptions and slow shippers that gap matters.

Put them together and the estimate for one customer is: last order date + (quantity ÷ usage rate). That’s the run-out date. Your job is to land a message a bit before it.

Let real behavior correct your first guess

Your label-based usage rate is a hypothesis, not a fact. The best signal for how fast a customer actually goes through something is how fast they reordered last time. If your model guessed 30 days and they came back at 24, they use it faster than average — shorten their next prediction. If they came back at 45, lengthen it. After the second or third order you’re no longer guessing from a label; you’re reading that person’s real rhythm.

This is why the first-order prediction and the repeat-customer prediction are different games. For a first-time buyer you lean on pack size and label. For a returning buyer you lean on their own history, which is far more accurate. The replenishment metrics that predict the next order are worth learning if you want to make this rigorous rather than rough.

One honest caveat: some people stock up, some let things run dry, and household size changes consumption in ways you can’t see. Treat the prediction as a good bet, not a certainty. You’re trying to be roughly right at the right time, not exactly right.

A worked example (illustrative)

Say you sell a dog food that comes in a 2kg and a 7kg bag. A customer buys the 7kg bag on the 1st of the month for a medium dog eating about 200g a day. That bag lasts 35 days. Run-out lands around the 5th of the following month. You want your reminder in their inbox around day 30 — five days of runway to order and receive before the bowl’s empty.

Now a second customer buys the same 7kg bag but has two dogs. Consumption doubles, the bag lasts ~17 days, and a day-30 reminder would reach them nearly two weeks after they ran out. Same product, same pack, completely different clock. If you’d sent both the same generic day-30 email, you’d have nailed one and lost the other. The numbers here are illustrative — plug in your own products and doses — but the shape holds for anything people consume.

What to automate

Turn the arithmetic into a flow. The pieces:

Trigger: an order containing a consumable product (tag those products so the flow knows which orders qualify).
Timing: a delay set to estimated run-out minus a buffer, where the estimate uses pack size and, for repeat buyers, their own prior interval — not one hard-coded number for the whole store.
Segment: split at least into heavy vs. light users once you have a second order to compare, so the delay adapts. Doing this properly is how to segment customers by real reorder behavior.
Channel: email for most; add SMS only when timing is tight and the product is genuinely run-out-or-suffer (contact lenses, medication-adjacent, pet food).
Content: a quick, useful nudge, not a hard sell. What that message should actually say is a topic on its own — see what to send before a customer runs out.
Goal: the reorder, with a clean exit so anyone who buys stops getting reminded.

The point of the automation is that it recalculates per person instead of firing on a shared calendar.

How to measure whether the prediction is any good

Watch four things:

Reorder rate among customers who entered the flow versus those who didn’t. This is the scoreboard.
Time-to-reorder relative to your predicted run-out date. If reorders cluster a week after your reminder, you’re sending too early; if they cluster before it, too late. Nudge the buffer.
Revenue per recipient on the flow, so you can see it’s earning more than it costs in list fatigue.
Unsubscribes and spam complaints on replenishment sends. A spike means your timing is off and people feel nagged.

The beauty of a prediction-based flow is that it tells you when it’s wrong: the gap between predicted and actual reorder dates is a dial you can tune every quarter.

Where Omnisend fits

Any decent ecommerce email tool can send a delayed post-purchase email. What you want for this is a platform that lets you branch on product tags and customer history, so the delay and content change with the buyer instead of staying fixed. I use Omnisend in my own stores — I moved to it after testing it against Klaviyo — because building a flow that triggers on a tagged consumable, splits by customer segment, and inserts the last-bought product into the reminder was a matter of dragging blocks around, not filing a support request.

Honest limits: no tool knows a customer’s real usage on order one; it only knows what they bought. The accuracy comes after the second order, once you feed the model their actual interval. And a reminder can’t rescue a product people didn’t like. Omnisend is an affiliate partner of Shopimation, and I recommend it from daily use — the free tier is enough to build and test one replenishment flow on a small segment before you roll it wider.

Your next step

Pull your top three consumable products and, for each, write down the pack size and a realistic per-day usage from the label. That gives you a first run-out estimate you can build a flow around today. Then decide what that first reminder actually says by reading what to send before a customer runs out, and once you’ve got a second order’s worth of data, tighten the whole thing with real segments.