Algeria AgencyE-commerce
Running an online shop: the day, the parcel, the return
A shop is chosen once and run every day. The second job is the one that decides whether the first was worth doing.
Choosing a platform is a decision. Running a shop is a routine, and the routine is what decides the outcome. Two businesses on the same solution, with the same catalogue and the same advertising budget, end up with very different figures because one has a cut-off time and the other does not, and because one hands parcels over at two in the afternoon and the other whenever it remembers.
This article describes that routine. It begins where the companion article stops: that one compares platforms, describes the catalogue and the funnel, and says what to check before signing. None of it is repeated here.
It is written for whoever actually packs the orders — often the owner, at the start — and for whoever picks the tool knowing they will be the one using it. The vocabulary is the back office, because that is where the day happens, rather than the shop front.
It carries no chart. The first section explains why the rates circulating on this subject cannot be compared with each other, and why publishing one more would add to the problem rather than solve it.
The day of a shop, and the hour that cuts it
An online shop has no opening hours, but it necessarily has a cut-off: the moment after which an order will leave tomorrow. That hour exists whether you choose it or not. If you do not choose it, the carrier pickup chooses it for you, and it moves every day.
Choosing it and writing it down changes two things at once. It becomes a promise you can keep, shown on the product page and in the order confirmation, and it becomes a routine: until that hour you sell, after it you pack. A shop that packs continuously stops at every notification and packs badly.
The cut-off also fixes what shipped within twenty-four hours means. Without it the phrase is an average nobody can check, yourself included. With it, it is a rule: ordered before the hour, handed over the same day; ordered after, handed over tomorrow. A buyer understands that immediately, because everything else they deal with works that way.
There is no right hour in the abstract. There is the hour that leaves you time to pack before the carrier arrives, minus a margin for the orders that go wrong. Take the pickup time, subtract two hours, and publish it. It is the cheapest operating decision on this page.
Why the rates that circulate here cannot be compared
On this subject, unlike payment, numbers are not scarce: everyone quotes a refusal rate, a return rate, an average delay. The problem is not that they are rare, it is that none of them says what it was calculated on, and the choice of denominator moves the result by a factor of three.
Take refusal at delivery. Measured against orders received, it includes the ones you never shipped because the number was wrong. Measured against parcels handed to the carrier, it excludes them. Measured against delivery attempts, it counts a parcel presented twice as two. Three defensible definitions, three very different numbers, and the same shop behind all of them.
A rate published without its definition is therefore an invitation to compare yourself with something you do not know. A merchant who reads fifteen per cent and measures twenty concludes they are doing badly, when they may simply be counting differently. It is a number that produces a decision, which makes it more harmful than an absent one.
The useful answer is not a better rate, it is yours, defined by you and stable over time. Write the definition once, measure it weekly, and compare yourself with yourself. A consistent internal series beats an outside benchmark whose contents you cannot see, and it is the only ground on which comparison means anything.
The state of an order is not its payment status
Two different pieces of information constantly end up in one column, and it is the most common operating mistake. An order has a state — received, packed, handed over, delivered, cancelled — and a payment status — unpaid, authorised, collected, refunded. The two advance separately and there is no fixed correspondence between them.
Merged, they produce situations with no way out. An order paid but never packed stays invisible in a list sorted by payment. An order delivered on cash terms looks unpaid for several days and nobody chases it. A refunded order stays marked delivered, which is true and stops being true if the parcel comes back.
So separate the two columns from day one, even on a spreadsheet. The packing list is read on the state, the accounts are read on the payment, and nobody has to interpret a value to know what they should be doing.
The rule that follows from that separation is the one holding the whole operation together: at any moment every order must sit in exactly one work queue, and it must be possible to say who has it. An order in no queue is a forgotten order, and it never announces itself.
When stock should drop, and by how much
The question looks technical and decides your stockouts. Stock can be decremented at the order, at the payment, or at the handover to the carrier. All three are defensible and none gives the same trouble.
Decrementing at the order protects the buyer who has just bought and exposes you to orders never paid, which tie up available goods. Decrementing at payment inverts the problem exactly: nothing is tied up, and two people can buy the last item, which produces the call nobody enjoys making.
The compromise that works for most shops is a reservation with an expiry: the order holds the item for a short window, and the hold falls away if payment does not arrive. Something then has to actually release the expired holds, and that is the part that gets forgotten — stock that never releases is worse than no reservation at all.
Whatever you choose, one rule is not negotiable: one stock truth, living in one place. A shop whose stock is kept both in the software and in a notebook does not have two sources, it has none, because at the first divergence nobody knows which to believe.
Packing: the movement repeated eighty times
Packing is the only item on this page whose cost is proportional to success. Everything else is done once; this is redone at every order, and a minute saved per parcel is an hour saved over sixty orders.
Two settings produce most of the gain. The first is the order of the packing list: sorted by location rather than by order time, it walks the stockroom once instead of once per order. The second is batching: packing eight parcels in one pass instead of eight passes of one parcel.
The third setting is not a setting, it is a discipline: never pack and answer messages at the same time. The interruption costs more than the message does, and it is nearly always during an interruption that the wrong item goes into the right box.
Finally, the check happens as the box is closed, by whoever packed it, against the list rather than from memory. A separate check by somebody else looks more serious and costs twice as much; the check that actually catches errors is the one that happens while the carton is still open.
The parcel, the label, and what the carrier really needs
The carrier does not need to know what you sell; it needs an address it can find and a number somebody answers. Almost every delivery failure that is your fault comes down to one of those two, and both are corrected before shipping rather than after.
A useful address here is not a postal address, it is a route: the commune, the neighbourhood, and a landmark the driver will recognise. A free-text address field produces whatever people type; a separate landmark field produces what the driver needs, because it asks the question explicitly.
The phone number is verified at the moment of the order, not at the moment of failure. A number wrong by one digit is indistinguishable from a good one on screen, and it only reveals itself once the parcel has already gone and already cost money. A simple confirmation message settles nearly all of this category.
Keep proof of handover, in whatever form. The moment a parcel stops being your responsibility is a boundary, and it is the only thing you will have to show if the parcel never arrives. A book the driver signs is enough; what is not enough is remembering.
Refusal at the door, and what to do about it
A parcel refused at delivery has cost the packing, the box, the outward trip and the return, and it comes back holding goods that are no longer quite new. It is the most painful item in a shop that delivers against payment, and it is treated upstream, never on the return.
The cause is rarely regret. Most often it is absence: the person is not there, does not recognise the call, or expected the parcel on another day. Those are problems of information and timing rather than of desire, and they respond very well to a message announcing the visit.
The second cause is delay. A parcel arriving long after the purchase reaches somebody whose appetite has passed, or who has bought elsewhere meanwhile. The refusal is then the consequence of a delay you let slip, and it is corrected under packing rather than under delivery.
Treat refusals as a list to reread every week, not as isolated incidents. Three refusals in the same commune in the same week are not three difficult buyers, they are a routing or address problem, and it is fixed once for every order that follows.
Going back into stock is not automatic
A returned parcel is the easiest order to lose, because it no longer has a buyer chasing it. It arrives in a box put down in a corner and stays there: the goods are nowhere, neither sold nor available, and they are missing at the moment somebody orders them.
A return is therefore a work queue like any other, with the same obligations: a dedicated physical place, a moment in the week, and a person who opens it. Three movements are enough — check the condition, put it back or set it aside, and close the order — but none of the three happens on its own.
Putting it back demands an honest decision about the condition. A product that comes back damaged and goes out as new produces a second return and a buyer who will not come back; better to take it out of sellable stock and move it another way. That is a loss recognised once instead of a loss deferred twice.
Closing the order in the software is the part that gets skipped and the one that distorts everything else. A refunded order still marked delivered inflates your sales, hides a missing item in stock, and makes the end of month reconciliation inexplicable. The movement takes ten seconds and it is the only one that leaves the figures right.
Messages: three templates cover almost everything
Customer service in a shop that is working is repetitive: the same four questions come back, and whoever answers rewrites them slightly differently each time. That is not a motivation problem, it is a missing template.
Three templates are enough at the start. The confirmation, restating what was ordered and when it will leave. The visit notice, warning that the parcel is coming and asking somebody to confirm they will be there. And the delay reply, saying what is happening, what you are doing, and when you will come back to them.
These are not automatic replies and must not become them. They are a starting point the person completes, which keeps the message personal while removing the part that is tiring to rewrite. The difference shows on the third message of a busy day rather than on the first.
The message that pays back most is the delay one, and it is the one never sent. A buyer warned of a delay waits; a buyer who discovers the delay by noticing it writes, then cancels. Sending it costs a minute; not sending it costs the order.
What the back office owes you every Monday
A shop is steered on very few numbers and they fit on one sheet. Orders received, orders handed to the carrier, orders delivered, orders refused, orders returned, and money actually collected. Six lines, the same week, read side by side.
Their value comes from the gaps between them rather than from their level. Received minus handed over is what you are failing to pack. Handed over minus delivered is transport and addressing. Delivered minus collected is payment. Each gap points at one precise place, and none of them shows on a turnover figure.
Track them weekly rather than daily. A day is too short to mean anything in a business of this size: one quiet Monday means nothing, four quiet Mondays mean something. The weekly rhythm is also the one at which a correction can be decided and then observed.
If your tool does not produce those six numbers, that is not a reason to change tool, it is a reason to count them by hand for a month. The manual month teaches you what you want automated, and it avoids installing a dashboard showing twenty indicators of which none is this one.
The days that are not like the others
A few days a year the volume doubles or triples: a promotion, a holiday, a product that spreads. Those days are not handled with the usual routine, and the classic mistake is discovering the problem while it is happening.
The bottleneck is almost never the site. It is packing and boxing, which is to say the number of parcels one pair of hands can close in a day. That limit can be known in advance, it is measured on an ordinary day, and it does not stretch by twenty per cent because the day matters.
Preparing for a peak is therefore physical before it is anything else: boxes counted in advance, labels ready, a second person warned, and the carrier told there will be more parcels. None of those is a shop setting, and each is arranged the week before.
Finally, plan to close deliberately whatever will not keep up: withdraw a product you hold only a few units of, or announce a longer delay during the promotion. A longer delay announced and kept beats the usual delay promised and missed across a whole week of orders, because the second option also damages buyers who asked for nothing.
What we do, and what we refuse
Our share is the back office and what surrounds it: order states kept apart from payment status, the packing list in stockroom order, a stock hold that really expires, the three message templates, and the Monday sheet of six numbers. These are settings and small pieces of development, not a project.
We refuse to promise a refusal rate, and for a different reason from scarcity: the number depends on a definition you choose and on delivery rounds that are not ours. We can help you define it, measure it and bring it down; we cannot tell you in advance where it will land, and a supplier quoting you a number here is quoting a definition they have not shown you.
Nor do we take on the operation itself. We do not box your parcels, we do not answer your buyers and we do not hold your stock. That is a stated limit of competence: those three are done better inside the business, by people who know the products, and a supplier who takes them on becomes one more intermediary between you and the buyer.
And the part that pays back fastest needs no development at all: choose a cut-off and publish it, add a landmark field beside the address, and send the delay message. Those three happen this week, without us, and if you keep only three things from this page, keep those.
Frequently asked questions
At what moment should stock be decremented?
The compromise that suits most shops is a hold placed at the order with a short expiry if payment does not arrive. Decrementing at payment lets two buyers take the last item; decrementing at the order ties up goods for orders never paid. The forgotten part is the mechanism that releases expired holds: without it, the hold is worse than no hold.
How do you reduce parcels refused at delivery?
Three movements cover most of it: verify the phone number at the moment of the order rather than at the moment of failure, ask for a landmark beside the address in a separate field, and announce the driver visit. The cause is almost always absence or delay rather than regret, and both respond to information sent in time.
Does a small shop need a cut-off time?
Yes, and it is more useful the smaller the shop is. Without one, shipped within twenty-four hours is an average nobody can check, yourself included. With one, it is a rule a buyer understands immediately. Take your carrier pickup time, subtract two hours, and publish the result.
What should be done with parcels that come back?
Treat them as a work queue, with a dedicated place, a moment in the week and somebody who opens it: check the condition, put it back or set it aside, then close the order in the software. That last movement is the one people skip and the one that distorts everything: a refunded order still marked delivered inflates your sales and hides a missing item.
Which numbers should be followed every week?
Six: orders received, handed to the carrier, delivered, refused, returned, and money actually collected. Their value is in the gaps between them rather than in their level — received minus handed over points at packing, handed over minus delivered at transport and addressing, delivered minus collected at payment. None of those three problems shows on a turnover figure.
Why does this article give no benchmark refusal rate?
Because a refusal rate depends on its denominator — orders received, parcels handed over, or delivery attempts — and no published figure states which. Three defensible definitions give three numbers that can differ threefold for the same shop. A number you cannot define invites a false comparison, which is more harmful than an absence.
Where we come in
Choosing a cut-off hour from your carrier’s collection time pays back faster than any development, and it is decided today.
- We separate an order’s state from its payment’s state, which are not the same thing.
- We surface the six numbers you count by hand today.
- We leave operations with you: we do not pack and we do not deliver.
No refusal rate will be quoted: it depends on your customers and your delivery area, not on the shop we would build.
Read next
E-commerce: what you are shown is the shopfront, what you are buying is the back office
Every demonstration shows the shop. All the cost and all the risk sit behind it, in the part nobody draws.When it works: what breaks next, and in what order
Growth breaks a business before it breaks a campaign. The constraint does not disappear, it moves — and never to where you are watching.The month renting gets more expensive than building
Past a certain volume, buying attention every month costs more than making the thing that receives it. How to spot that moment, and what breaks if you build too early.
Let us talk about your project
A free audit, no commitment: we look at your online presence and tell you what is holding it back.