AfterDesk
EN
Apply

The toolkit

E-commerce store operations

Online stores run on unglamorous back-office work: listings that stay consistent, stock counts that match reality, orders that hit exceptions, returns, refunds and reviews. This course teaches those shapes on any platform, so you can open an unfamiliar store admin and know what to check first. You finish able to run a week of store operations and prove in writing what you did.

9
lessons
~59
minutes
12
exam questions

Free · No paid tier · No certificate fee

After this course

Everything, and what is in it.

The Back Office of a Store

~6 min

What store operations actually is

Store operations is everything that happens after the marketing and before the customer is happy. Someone keeps the catalogue accurate, watches stock, pushes orders out the door, answers where is my order, processes returns, and reads reviews. None of it is glamorous and all of it is load-bearing. A wrong stock number sells something that cannot be shipped. A missing image kills a product's sales. An unanswered return becomes a chargeback. Store owners hire help for this work because it is constant and detailed, and because it can be described precisely enough to hand over. That is why it shows up as paid work, and why doing it well makes you the person a store keeps coming back to.

The four systems in every store

Whatever platform a store runs on, the same four systems sit underneath, wearing different names. The catalogue holds products, variants, prices and images. Inventory holds how many of each exist, and where. Orders hold what was bought, by whom, and what has shipped. Customers hold the people, their addresses and their history. Every task you get touches one or two of these, and most confusion comes from not knowing which. A product with the wrong price is a catalogue problem. A product that sold when it was gone is an inventory problem. A parcel that never moved is an order and shipping problem. Name the system first; the fix is usually obvious once you have.

Learn the shape, not the buttons

You will be dropped into store admins you have never seen. Do not try to memorise menus. Ask any system the same six questions: where do I list all records, how do I filter and sort them, how do I find one, how do I edit one, how do I edit many at once, and how do I export to a spreadsheet. Nearly every admin answers all six, and once you can answer them you can work. Find the export first. A store's own admin is often clumsy, while an export into a free spreadsheet gives you sorting, counting and comparison it may not offer. Remember that the export is client data and stays where the brief says it stays.

Order data is the sensitive part

Catalogue work is mostly public information. Order and customer work is not. An order record carries a real person's name, home address, phone number, email and what they bought, and in some stores the last digits of a card. Treat all of it the way you would want your own address treated in a stranger's hands. Never export customer data anywhere except the file the brief asks for, never paste order details into a translator, a formatting site or an AI tool, and never keep a copy after the work is approved. On AfterDesk this is not a preference: client data stays inside the task, and a tool that needs it uploaded is a tool you cannot use. If you ever see full card numbers, stop and flag it.

Remember

  • Store operations is the constant, detailed work after marketing and before a happy customer.
  • Four systems in every store: catalogue, inventory, orders, customers. Name the system before fixing.
  • Learn where any admin lists, filters, edits, bulk-edits and exports. Skip the menus.
  • Order records hold real names and addresses. They never leave the task.

Listings That Hold Up

~7 min

The anatomy of a listing

A listing is not a description with a price. It is a record with a dozen fields, and a store is judged on whether those fields are complete and consistent across hundreds of products. The usual set: title, description, one or more images, price and compare-at price, product code, variants, stock quantity, weight and dimensions for shipping, category, tags, and the spec fields the store filters on, such as size, material, colour or voltage. Missing fields have visible consequences. No weight means shipping is charged wrong. No category means the product is unfindable. No spec values mean it never appears when a shopper filters. Before adding anything, open three listings the client is happy with and write down every field they fill. That list is your template.

Titles follow a house pattern

Store titles are not creative writing. They follow one pattern applied to every product, so a shopper scanning a category page compares like with like. A common shape is brand, then model or product name, then the attribute that distinguishes it, then size or quantity. Whatever the store already does, copy it exactly: capitalisation, the separator it uses, whether units are written as ml or mL. If the existing titles disagree with each other, that is worth flagging rather than silently picking a winner. Write the pattern down as a sentence at the top of your working sheet and check every title against it before delivery. Consistency is the deliverable here, and a cleverer title that breaks the pattern is a defect.

Product codes and variants

A product code, usually called a SKU, is the store's internal name for one exact sellable thing. A blue shirt in medium and the same shirt in large are two codes, not one. Good codes are unique, stable, and readable by a human in a warehouse: TSH-BLU-M tells you more than 100294. Three rules matter. Never reuse a retired code for a new product, because old orders still point at it. Never change a code that has sales history unless the brief says to. Never invent one that breaks the store's existing pattern. Variants are the same product in different options, and every combination the store sells needs its own row with its own code, price and stock. A combination that does not exist should not be listed as available.

Images to the store's standard

Most stores have an image standard even if nobody wrote it down. Look at the listings the client likes. How many images, what the first one shows, whether the background is white, whether they are square, roughly what size. Match it. Free tools cover the usual work: GIMP crops, resizes and exports at a set pixel size, and handles batches. Name files predictably, usually by product code, so an image can always be traced back to its product. Write alt text that describes the product plainly for someone who cannot see it, not a list of keywords. Two hard limits. Never take images from another store or a search engine, because they belong to someone else. Never present a photo of one variant as another.

Check it the way QC will

Before delivery, stop being the person who wrote the listings and become the person checking them. The fastest method is a sweep by field, not by product. Sort your sheet by title and scan for pattern breaks. Sort by product code and look for duplicates or blanks. Filter for empty image cells, empty weights, missing categories. Errors cluster in columns, so column by column finds what product by product misses. Then open two or three finished listings the way a shopper sees them and read them cold. If the brief asked for forty and you delivered thirty-eight because two spec sheets were unusable, say so in your delivery note with the two product names. A stated gap is a decision. A silent one is a rejection.

Remember

  • A listing is a record of a dozen fields, not a description with a price.
  • Copy the store's title pattern exactly. Consistency beats cleverness.
  • One sellable thing, one product code. Never reuse or silently change one.
  • Match the store's image standard, and never take images from another site.
  • Check by column, not by product. Errors cluster in fields.

Product Copy Without Invented Claims

~6 min

Start from the spec sheet

Product copy is a translation job, not an invention job. Your source is the spec sheet, the manufacturer's data, or whatever the client supplied, and everything you write has to trace back to it. Read the whole source before writing a word, then list the facts you actually have: dimensions, materials, capacity, compatibility, what is in the box, warranty length. Now sort them by what a buyer decides on. For a backpack that is capacity, laptop size, weight and whether it fits in a cabin bag. For a cable it is length, connector types and speed. Your paragraphs should follow the buyer's order of questions, not the spec sheet's order of rows. If a decisive fact is missing from the source, that is a finding, not a gap for you to fill.

Feature first, then what it means

The pattern that works is a fact followed by its consequence for the buyer. The bag holds 28 litres, which covers a change of clothes for three days. The panel is 6 mm aluminium, so it does not flex when you lean on it. USB-C at 100 watts, enough to charge a laptop while you use it. The fact comes from the spec sheet, and the consequence has to be an ordinary, checkable inference from it, not a promise. Best in its class, revolutionary, and you will love it are not consequences, they are noise, and a careful store owner cuts them. Keep sentences short and readable on a phone, because most shoppers are on one. Two tight paragraphs plus a spec list beats a page nobody reads.

The line you never cross

You may never add a claim the source does not make. Water resistant is not waterproof. Antibacterial, hypoallergenic, organic, certified, food safe, medical grade and lifetime warranty are all statements a business can be held to, and in many countries they are regulated. If the spec sheet does not say it, it does not go in the copy, even when a similar product elsewhere is sold that way. This applies hardest to anything eaten, applied to skin, or used by children: never write that a product treats, prevents or cures anything. If the client wants such wording, it has to come from them in writing, and the authority on what is allowed is the regulator in the country of sale, not us and not you. This is not legal advice; it is where the work stops.

When the source is silent

Work with three buckets. Known facts, straight from the source. Reasonable consequences, which you may write. Unknowns, which you may not. Keep the unknowns in a visible list as you go: no weight given, strap material not stated, compatibility with the older model unclear. Deliver that list with the copy. On AfterDesk it goes in your delivery note, plainly: fifteen descriptions written, three fields I could not source, listed here. The operator can go back to the client in a minute. What you must never do is close a gap by finding a similar product elsewhere and borrowing its numbers, because eventually you borrow a number that is wrong and it gets published under the client's name. An honest gap is cheap. A wrong specification is not.

Remember

  • Everything you write traces back to the source the client supplied.
  • A fact plus a checkable consequence. Superlatives are noise.
  • Never add a claim the spec sheet does not make.
  • Health and safety wording comes from the client and the regulator, never from you.
  • Missing facts go in a list for the client, never invented.

Inventory and the Reorder Point

~7 min

The number has to match the shelf

Every stock number in a store is a claim about the physical world, and it is wrong more often than owners think. Returns get restocked twice. Damaged units are thrown out without an adjustment. A warehouse pick is missed. A second sales channel sells the last unit a minute before the website does. The consequences are not symmetric. Stock recorded too high causes oversells, which means cancelling on a customer who already paid, and that costs the store far more than the item. Stock recorded too low is quieter but expensive: the product looks unavailable and simply stops selling. Your job in inventory work is to make the recorded number and the real number agree, and to make the disagreements visible when they cannot be resolved from your desk.

Sales rate, lead time, reorder point

The reorder point is the stock level at which you must order more so you do not run out while the next shipment is in transit. It has three inputs: how many you sell a day, how many days the supplier takes to deliver, and how much cushion you want. The arithmetic is average daily sales multiplied by lead time in days, plus safety stock. If a product sells 8 a day, the supplier takes 21 days, and the store wants 10 days of cushion, that is 8 times 21, which is 168, plus 80, giving a reorder point of 248 units. When stock falls to 248, it is time to order, not when it reaches zero. Calculate the daily rate over a recent, representative period, and state which period you used.

Safety stock is a judgement

Safety stock covers the days when sales spike or the supplier is late, and choosing it is a business decision, not a formula you settle alone. A cheap, slow, always-available item needs almost none. A bestseller shipped by sea from another continent needs weeks of it. If the brief gives you a rule, apply it exactly. If it does not, use a stated assumption rather than a silent one: I used ten days of cover for every product because the brief did not specify, and here is the same list recalculated at seven and fourteen days. That hands the client a decision instead of a mystery. Watch for the things that break an average: a promotion week, a one-off bulk order from a single buyer, a product that only sells in December.

Counting, adjusting, and the log

Nobody counts a whole warehouse often. The practical method is cycle counting: count a slice regularly, prioritising fast movers and high value items, so everything is verified over a cycle instead of once a year in a panic. When a count disagrees with the system, do not simply type in the new number. Record the product, the counted quantity, the system quantity, the difference, the date, and, where you know it, the likely cause. A pattern in those causes is worth more than any single correction, because it points at the process that keeps breaking. Adjustments change what a store believes it owns, so they need a trail. If you were not asked to make adjustments, do not make them. Deliver the discrepancy list and let the client decide.

Remember

  • Stock too high causes oversells and cancellations. Stock too low makes products invisible.
  • Reorder point equals average daily sales times lead time, plus safety stock.
  • Safety stock is a business decision. State your assumption instead of hiding it.
  • Cycle count regularly, and log every adjustment with quantity, date and likely cause.

Orders and the Exceptions

~7 min

The clean orders come first

An ordinary order needs almost nothing from you. Payment captured, items in stock, address clean, pack, ship, done. On a good day you push a hundred of those through in batches and the whole thing takes an hour. The time goes somewhere else. In most stores a small share of orders, often somewhere between one in twenty and one in ten, has something wrong with it, and that minority consumes most of your shift. So process the clean ones first, in one uninterrupted pass, and move the odd ones into a separate list as you meet them. Working straight down the list and stopping to solve each problem as it appears is the slowest possible method, because every stop breaks the rhythm of the easy work.

The exceptions that eat the day

The list is short and it repeats. An address that is incomplete, undeliverable, or obviously a typo. An item that turns out to be out of stock after the sale. Two identical orders from the same customer minutes apart. An order that has to ship in two parcels from two locations. A destination that is a post box, a remote area, or another country with customs paperwork. A payment the processor has flagged for review. A customer who wants to change or cancel after paying but before shipping. A price on the site that was plainly wrong. Learn to recognise all eight on sight, because recognising them is most of the speed. For each, the store either already has a rule or it does not, and finding out which is the first step.

Rules first, judgement second

Sort every exception into three boxes before touching it. Things you may fix yourself, such as correcting Steet to Street or adding a house number the postcode confirms. Things you do by a stated rule, such as the store's policy of shipping the in-stock part of an order and cancelling the rest. Things that are never yours, which means anything that moves money or promises something to a customer: refunds, cancellations, free replacements, upgraded shipping, discount codes. If the store has no rule for a case, do not invent one under time pressure. Park the order, keep working, and put the case in front of whoever decides, with what you found and what the options are. On AfterDesk that is your delivery note; there is nobody else to ask.

Write it on the order

Every store's order record has somewhere to write an internal note, and it is the only memory that survives you. Write what you found, what you did, and when. Address corrected from the customer's reply, 14 March. Held for stock, supplier says the sixth. One line, on the record, at the moment it happens. Do not keep this history in your head, in a personal notebook, or in your own email, because the person picking the order up tomorrow will have none of those. When you deliver a batch of order work, your summary is a count and a list: eighty-two processed, six held, with the reason for each. That is a deliverable someone can act on without asking you a single question.

Remember

  • Process clean orders in one pass and quarantine the exceptions into a separate list.
  • Eight exception types cover most days. Recognising them on sight is the speed.
  • Anything that moves money or promises a customer is never your call alone.
  • Write what you did on the order record, at the moment you do it.

Shipping and Where Is My Order

~6 min

Tracking numbers are a promise

The tracking number is where a store's promise becomes checkable. Three things have to be true: the number is real and scanning, it is attached to the right order, and the customer has been told it exists. Mismatched tracking is a common and damaging mistake, because a customer watching someone else's parcel believes theirs is arriving and is angrier when it does not. When you enter or import tracking numbers in bulk, verify a sample against the orders by hand, and check that the carrier on the label matches the one recorded on the order. If a shipment goes out without a number, write that on the order rather than leaving the field empty and hoping nobody asks.

Reading a tracking history

Carrier statuses differ in wording but not in meaning. Label created with no scan means the parcel never entered the network. It may still be on a packing table, and after a couple of business days that is a warehouse question, not a carrier question. Scans that stop for several days usually mean a missorted or stuck parcel, and most carriers want an enquiry opened rather than a wait. Delivered with nothing received is often a neighbour, a reception desk, or a scan made a street early, and the first step is asking the carrier for the delivery detail or photo. Returned to sender means a bad address or failed attempts, and it needs a reshipping decision before anything else. Read the whole history, not the last line.

The where-is-my-order reply

The customer wants one thing: to know what happens next and when. A reply that works has four parts and fits in a short paragraph. Acknowledge the order by number. State where it is in plain words, not a copied carrier code. State what you are doing and by when: we have opened an enquiry with the carrier and will come back to you by Thursday. Give one clear next step if there is one. Apologise once, briefly, and never twice in the same message. Do not promise a delivery date the carrier has not given you, and never write that it is on its way when the tracking says nothing has moved. On AfterDesk you draft these and the client sends them, so write them ready to send.

Lost, late, damaged, and duties

Every carrier has a window for reporting a lost or damaged parcel, and missing it usually ends the claim, so the date matters more than the wording. Gather what the claim needs before writing anything: order number, tracking number, ship date, declared value, photos of the damage and the packaging. What the customer is offered, a reshipment, a refund, or nothing, is the store's decision and not yours. International orders bring customs duties and taxes, usually paid by the buyer on arrival unless the store has chosen otherwise. Rates and thresholds change by country and by year, so never quote a figure from memory. Point to the destination country's official customs authority and tell the client the answer has to come from there. That is not tax advice, and neither is anything else you write.

Remember

  • Verify tracking numbers against orders. Wrong tracking is worse than none at all.
  • Read the whole tracking history; the last line rarely explains the problem.
  • A good reply gives the fact, the action being taken, and a date.
  • Report loss or damage inside the carrier's window, with the evidence gathered first.
  • Duties change by country and year. Point to the official customs source.

Returns and Refunds

~7 min

The policy is your instruction

Every returns question is answered by the store's policy first and by your opinion never. Before you touch a return queue, take five things from that policy and keep them beside you: how many days the customer has, what condition the item must be in, who pays return shipping, what is excluded, and what proof is required. Exclusions are where disputes come from, so read them twice: personalised items, opened hygiene products, perishable goods, clearance stock. If the policy is silent on the case in front of you, that silence is the finding. Do not decide it, and do not tell a customer what you assume the policy means. Ask the client through your delivery note and write the answer down, so the next person inherits a rule instead of the same question.

The workflow end to end

A return has the same eight steps everywhere. The customer requests it. Someone checks it against the policy and authorises or declines. An authorised return gets a reference number and usually a label. The parcel arrives and is matched to that reference. Someone inspects the item and records its condition. A decision follows: refund, exchange, repair, or refuse. The stock is either put back on the shelf or written off. The refund or exchange is processed and the case is closed with the reason recorded. The two steps most often skipped are matching the arriving parcel to its reference, which is how unidentified boxes pile up in a corner, and recording the reason, which is the only data that ever tells a store why its returns happen.

Inspect it and write it down

Inspection is where a return becomes a number. Open the parcel, check the item is what the return was authorised for, and record its condition in words anyone can check later: unopened, opened but complete and unmarked, used with scuffing on the base, missing the charger. Photograph anything that is not perfect, before repackaging. Then apply the policy rather than your sympathy, and where the policy allows a partial refund for a missing accessory or visible wear, state the deduction and the reason on the record. Two things to resist: deciding a customer is lying because the item looks used, and deciding they are honest because their note is polite. Neither belongs in the file. Describe what is in front of you and let the person with authority decide.

Refunds move money

A refund goes back to the payment method that paid, in the currency it paid in. That is not a preference. It is how disputes stay resolvable, and it is the best defence against a common fraud: a request to send the money to a different card, account or wallet. Treat any such request as a stop, not a task, however plausible the reason, and treat a return address that changes mid-conversation the same way. You should never need a customer's bank details, never ask for a full card number, and never accept one if it is sent to you. Unless the brief explicitly authorises you to issue refunds, your work ends at a recommendation with the evidence attached, and someone with the authority presses the button.

Remember

  • Take five facts from the policy: window, condition, shipping, exclusions, proof required.
  • Match every arriving parcel to its return reference, and record the reason.
  • Describe condition in checkable words and photograph anything that is not perfect.
  • Refunds go back to the original payment method. Any other request is a stop.
  • Never handle bank or card details. Recommend, and let the authorised person execute.

Reviews and the Bad One

~6 min

Who you are really answering

A public review has two audiences and you only really speak to one of them. The reviewer may never read your reply. The hundred people who read the review while deciding whether to buy certainly will, and what they are judging is not the complaint, it is how the store behaves when something goes wrong. That changes what a good reply looks like. Calm, specific, short, and it never argues. A store that answers a two-star review with a defensive paragraph loses more sales than the review did on its own. A store that answers plainly and fixes the problem often gains from it, because shoppers now know what happens if their own parcel goes missing.

How to answer a bad review

Four moves, in order, in about five sentences. Thank them and name the specific problem so it is obvious you read it: you ordered the 40 centimetre model and received the 30. Say what went wrong, if you know and it does not expose anyone by name. Say what is being done about it, both for them and in general. Give one route to reach the store, the normal support channel, and stop. Never dispute the facts in public, never repeat the customer's order details or address, never offer anything in exchange for changing a review, and never use a reply to sell. If the review contains a factual error that matters, correct it once, neutrally, and leave it there.

Nothing about reviews is invented

You will occasionally be asked, directly or by hint, to write reviews, to post one under another name, to have friends post them, or to bury a bad one under new ones. The answer is no, every time. Fake reviews break the rules of every platform that hosts them, they are illegal in a growing number of countries, and stores are removed for them. The same goes for review gating, meaning a system that quietly routes happy customers to the public form and unhappy ones somewhere private. If a brief asks for anything in this territory, stop and flag it rather than complying. On AfterDesk, say so plainly in your delivery note. Refusing protects the client from a penalty they may not have considered.

Reviews are cheap data

Read a store's reviews with a spreadsheet open and one column for the theme, and the pattern shows up within fifty entries. Smaller than I expected, four times, is not a customer problem. It is a listing with no dimensions and no photo showing scale. Arrived in nine days, repeatedly, means the shipping estimate on the product page is a story the store tells itself. Instructions unclear points at a leaflet, not at the product. This is the most useful thing you can hand a store owner: not the reviews, which they have already seen, but the count by theme and the specific listing or process each theme points at. It turns complaining into a work list.

Remember

  • You reply for the future readers, not the reviewer. Calm, specific, short.
  • Answer in four moves, and never argue in public or repeat order details.
  • Never write, buy, or gate reviews. Flag the request instead of complying.
  • Tag review themes; each pattern points at one listing or process to fix.

The Weekly Store Health Check

~7 min

Why a weekly sweep exists

Store problems do not announce themselves. A product goes out of stock on a Tuesday and nobody notices until a customer asks. An image fails to upload and one listing sits blank for a month. Six orders never left the warehouse because their payment status was odd. None of these produce an alert. They produce a slow leak. A weekly health check is a fixed set of questions asked in the same order every week, so leaks are caught in days instead of quarters. It takes about an hour once you have built it, and it is one of the easiest recurring deliverables to justify, because the client gets a list of real problems every week that they did not have to find.

Catalogue and stock checks

Export the catalogue and ask five questions. Which live products have no image, no description, or no price. Which product codes are duplicated or blank. Which listings break the title pattern or are missing a spec field the store filters on. Which products are below their reorder point, and which will cross it before the next check given their sales rate. Which products have had no sales for a long stretch and are still occupying stock and attention. A free spreadsheet such as Google Sheets answers all of this with sorting, filters and a few counts, and you can rebuild the same sheet every week in minutes if the column layout never changes. Record the date and time of the export, because a stock number without a timestamp means nothing.

Orders, returns and reviews

Then ask five more. Which paid orders are still unfulfilled beyond the store's normal window. Which shipped orders have tracking that has not scanned in days, or no tracking at all. Which authorised returns have not arrived, and which arrived parcels are still uninspected. Which refunds are approved but not yet processed. Which reviews from the past week are unanswered, and what the week's complaint themes were. Each of these is a queue, and queues fail by growing quietly. What you are looking for is not only the number but the direction and the age. Eleven unfulfilled orders is fine if it was fourteen last week and the oldest is two days old. It is an emergency if the oldest is three weeks.

Deliver a report, not a feeling

The output is a short document and one spreadsheet, never a paragraph saying things look fine. Give each check a count, last week's count beside it, and the rows behind it on their own tab. At the top, the three to five things you would fix first and why. Keep the wording factual: 23 live products have no image, up from 19; the oldest unfulfilled order is 12 days old; three of this week's five reviews mention sizing on the same product. When this goes to an operator, your delivery note names the biggest finding in one sentence so it is not buried in a tab. Then delete your working copies once the work is approved, because they are full of the store's customers.

Remember

  • Errors accumulate silently. A weekly sweep catches leaks in days, not quarters.
  • Ask the same questions in the same order every week.
  • Catalogue checks: missing fields, duplicate codes, pattern breaks, reorder points, dead stock.
  • Queue checks: unfulfilled orders, silent tracking, pending returns and refunds, unanswered reviews.
  • Report each count against last week, with the rows behind it and the top fixes.

Sit the exam

The courses are free. The work is real.

Twelve scenario questions. Pass at ten. Three attempts a day. The bar is the point.

No paid tier. No certificate fee. No upsell. Not now, not later.

Create a free account

You can start the first course tonight.

The curriculum

Everything, and what is in it.