Blog/Operations
OperationsAugust 8, 2026·12 min read

How to train a Shopify virtual assistant

Most VA hires fail at the handoff, not the hire. The 30-day ramp, the order to give tasks in, and what to write down before day one.

AY
Anand Yadav · Founder, ReccordSOP
·Last reviewed August 8, 2026

The pattern repeats across DTC brands with unnerving consistency. You hire a Shopify VA, spend two weeks answering questions you thought you had already answered, redo half their work yourself, and quietly conclude that virtual assistants do not work for your store.

The hire was almost certainly fine. What went wrong is the handoff: you gave a competent stranger a set of tasks that only exist properly in your head, and no amount of competence compensates for that. This is a documentation problem wearing a hiring problem's clothes.

Here is the version that works: what to write down before they start, the order to hand tasks over in, a 30-day ramp, and the month-three failure that catches almost everyone.

Why the handoff fails, not the hire

Run one test on any task you are about to delegate. Could a competent person who has never seen your store do it correctly from what is currently written down? If the answer is no, you have not delegated the task. You have created a dependency on yourself that now also costs you a salary.

Three things make this worse in a DTC context specifically:

  • The work lives inside other companies' software. A refund is a Gorgias action plus a Shopify action, and the exact sequence matters. Knowing 'process the refund' is not knowing the procedure.
  • The timezone gap turns small gaps into whole days. A question that costs thirty seconds when someone is sitting next to you costs twelve hours when they are not, and a VA who is blocked will either guess or wait. Both are expensive.
  • Most of the difficulty is in the exceptions, not the standard path. Anyone can process a normal refund. The job is knowing what to do when the order is 47 days old, or the customer is abusive, or the item shipped to the wrong address because the customer typed it wrong.

That last one is where most training material stops, which is why most training material fails.

Write down three things before day one

Not your whole operation. Three tasks, the ones you plan to hand over first. Each one needs four parts, and the third is the one everybody skips.

1. The trigger

What starts this task, stated precisely enough to be noticed. Not 'handle refund requests' but 'a Gorgias ticket is tagged refund-request, or a customer email contains the word refund and has an order attached.' A VA cannot act on a trigger they cannot recognize.

2. The steps, with screenshots

Click by click, in the tools you actually use, with a screenshot of each screen. Written steps alone assume the reader already knows where things are. They do not, and asking them to find out is how a 20-minute task becomes a 20-minute task plus four questions.

3. The decision rules

This is the part that separates a document someone can work from and a document that generates questions. Every judgment you make without thinking has to become an explicit rule with a number in it.

  • Refunds under $75 on orders less than 30 days old: process it, no approval needed.
  • Refunds over $75, or orders older than 30 days: draft the response, tag me, do not send.
  • Customer has had two or more refunds in 90 days: escalate regardless of amount.
  • Item arrived damaged with a photo: approve immediately, do not request a return.

Without rules like these, a VA does one of two things: asks you about everything, which is not delegation, or guesses, which is worse. Give the thresholds and you get consistent decisions from someone who has never met your customers.

4. The escalation line

One sentence naming who to ask, on which channel, and how long to wait before proceeding. 'If unsure, message me on Slack and wait' is incomplete without the last part. Say what happens if you are asleep, because for most offshore arrangements you will be.

Free tool: which processes only exist in one person's head?

Rate the tasks your store runs weekly and see which ones nobody else could pick up. That list is the documentation order for the handoff.

The order to hand tasks over

The instinct is to delegate the task you hate most. Resist it, because the task you hate most is usually the one with the most judgment in it, and judgment is the last thing to hand over, not the first.

Hand tasks over in this order:

  1. High frequency, low stakes, reversible. Order tagging, tracking-number replies, review requests, updating inventory notes. Mistakes here are cheap and visible, which is exactly what you want while you are learning how someone works.
  2. High frequency, medium stakes, with a clear rule. Standard refunds under your threshold, address changes before fulfillment, reshipping a confirmed lost parcel. The rule does the deciding, so the VA is executing rather than judging.
  3. Judgment work, drafted for your approval. Escalations, angry customers, exceptions to policy. They write the response, you approve it, for as long as it takes to trust the reasoning. This is training, not review theatre.
  4. Judgment work, independent. Only after you have watched them reason through a few dozen of these and agreed with the calls.

A brand doing 300 orders a week can usually move through the first two tiers inside a month. The third takes a quarter, and some brands never hand it over, which is a legitimate choice rather than a failure.

The 30-day ramp, week by week

Week 1: watch, then do with full review

Record yourself doing each of the three tasks end to end, narrating the decisions as you go, and send the recordings before their first shift. Then they do the same tasks and you review every single one. Expect corrections. The point of week one is not output, it is finding the gaps in your own documentation, which will be more numerous than you expect.

Keep a running list of every question they ask. That list is the missing content in your SOPs, and it is the most valuable artifact of the first week.

Week 2: same tasks, spot checks, add one

Stop reviewing everything and start sampling around one in five. Add a fourth task from tier one or two. If week one produced fifteen questions and week two produces two, the documentation is working. If week two produces fifteen again, stop adding tasks and fix the documents.

Week 3: introduce judgment work, drafted

Hand over one tier-three task in draft-and-approve mode. You are reading their reasoning, not just their output. When you disagree with a call, explain the reasoning rather than correcting the text, and add the rule you just articulated to the document. Most of your decision rules get written in week three, not week zero.

Week 4: independence on tiers one and two

No review on the first tasks, weekly sampling instead. Judgment work stays in draft mode. By the end of the month you should be able to name exactly which tasks you no longer think about, and that number, not hours worked, is what you bought.

The mistake that wastes month one

Handing over ten tasks in week one because the VA seems capable. They are, individually. But ten undocumented tasks at once means ten sets of questions arriving in parallel, which is more management overhead than doing the work yourself, which is how brands conclude that VAs do not work.

Checking work without micromanaging

Time tracking and activity screenshots measure the wrong thing and poison the relationship. What you want is a small number of outcome checks per task.

  • Sample rather than inspect. Five refunds a week reviewed properly tells you more than glancing at forty.
  • Pick one number per task. For refunds, the share processed outside policy. For tagging, the share of orders correctly tagged at end of day. For support, first-response time and reopen rate.
  • Watch reopens and escalations, not volume. A VA closing 60 tickets a day with a 20 percent reopen rate is producing work for you, not removing it.
  • Ask for a recording when something looks wrong. Watching how the task was actually performed, once, ends more disputes than a paragraph of feedback.

The recording point generalizes. When the way a task gets done drifts from the way it was documented, watching a real run beats rereading the document, and it doubles as the raw material for the update.

Month three, when it quietly breaks

The failure almost nobody plans for arrives around week ten, and it does not look like a training problem. It looks like a VA who was doing fine and has started making mistakes.

What actually happened is that the store moved and the documentation did not. Klaviyo redesigned the flow builder, so the screenshots are wrong. You changed the refund window in Slack and never updated the SOP. You switched 3PLs and the receiving steps are different now. The VA is following the document exactly, which is what you asked for, and getting it wrong, which is not their fault.

That gap is drift, and remote teams feel it before anyone else does, because a person in the room notices the button moved while a person following a document does not. Two habits contain it: when the VA hits something the document does not cover, they log it rather than improvising silently, and once a quarter they re-record a task end to end so the two of you can compare the current run against what the SOP claims.

SOP drift: why your documentation is lying to you

The six signals that a procedure has stopped matching reality, and the 20-minute monthly check that catches them before your VA does.

This is the problem ReccordSOP is built for: record the task once and it becomes a step-by-step SOP with screenshots your VA can follow, and when the process changes, re-recording it shows exactly which steps moved. For a remote handoff that is the difference between documentation that ages well and documentation that quietly turns into misinformation.

Free SOP templates for the DTC stack

Ready-made procedures for Klaviyo, Gorgias, Recharge, ShipBob and the rest, plus the cross-tool ones. Copy them as a starting point for the handoff pack.

When it genuinely is the hire

Documentation fixes most VA problems, not all of them. Three signals point at the person rather than the process, and it is worth naming them honestly so you do not spend a quarter rewriting documents for someone who will not follow them:

  • The same mistake recurs after being corrected twice with the rule written down. Once is a documentation gap. Twice is a comprehension or attention problem.
  • No questions at all in week one. This is worse than too many. Nobody understands a new store's exceptions immediately, so silence usually means guessing.
  • Work that is marked done but not actually finished. Tickets closed without resolution, orders tagged without the underlying check. This is the hardest to detect and the most damaging, and it is why the outcome metric matters more than the activity log.

Everything else is almost always fixable by writing the thing down properly. Before you conclude that a VA cannot do the job, reread what you gave them and ask whether you could do the task from that document alone, in a different timezone, without being able to ask a question.

Build a DTC SOP library that doesn't go stale

Once the first three procedures exist, this is how the rest of the library gets built without stalling in week two.

Frequently asked questions

How do I create training materials when I hire a virtual assistant for Shopify?

Start with only the three tasks you plan to hand over first, not your whole operation. For each one, document four things: the trigger that starts it, the click-by-click steps with a screenshot of every screen, the decision rules with actual numbers in them (refunds under $75 go through, over $75 get drafted for approval), and who to ask when something is unclear. Screen recording each task while narrating your decisions is the fastest way to produce this, because it captures the judgment calls you would forget to write down.

How long does it take to train a Shopify VA?

Roughly 30 days to full independence on routine work, and about a quarter before handing over judgment work like escalations and policy exceptions. The realistic week-one experience is that you spend more time than the task would have taken you, because week one is really about finding the gaps in your documentation. That investment is what buys the following eleven months.

What tasks should I give a Shopify VA first?

High frequency, low stakes, and reversible: order tagging, tracking-number replies, review requests, inventory note updates. Then move to medium-stakes work that has a clear rule attached, such as standard refunds under a set threshold. Save escalations, angry customers, and policy exceptions for last. The common mistake is delegating the task you hate most first, which is usually the one requiring the most judgment.

How do I train a Shopify VA to run my store without me?

In stages, and the last stage is not documentation, it is decision authority. Written procedures get someone to competent execution. Running the store without you also requires explicit spending and policy limits (what they can approve alone, what needs you), a named escalation path for the situations no procedure covers, and a review cadence where you check outcomes rather than activity. Most brands reach independent routine operations and deliberately keep judgment work, which is a reasonable end state rather than a failure.

My VA was doing well and started making mistakes. What changed?

Usually the store changed and the documentation did not. A tool shipped a redesign so the screenshots no longer match, a policy shifted in a Slack thread that never reached the SOP, or you switched fulfillment partners. The VA is following the document exactly and getting it wrong. Remote workers hit this before in-house staff do, because someone in the room notices the button moved and someone following a document does not. Re-record the task and compare it against what the procedure claims.

AY
Anand YadavFounder, ReccordSOP

I built ReccordSOP after watching too many DTC ops teams lose months to undocumented workflows. These SOPs are battle-tested with Shopify operators running $1M to $50M brands.

Last reviewed August 8, 2026

Related reading