The post-purchase operations playbook

The reactive tax

Every order is a promise. You promised a thing would arrive, in a condition, by a date. Post-purchase operations is the work of keeping that promise, and the reactive tax is what you pay when the customer is the one who discovers you broke it.

For a decade the maths let brands ignore this. Cheap capital and low acquisition costs meant a customer lost to a bad delivery could be replaced for $20, so operations was run as a cost center: lean, invisible, and reactive by design. That era is over. With acquisition costs now sitting at $50 to $100 and above, you cannot afford to lose a customer you paid $80 to acquire because they did not know where their parcel was.

The bill is bigger than most operators think. Australian e-commerce generates roughly 192 million support contacts a year. If 30 to 40% of those are "where is my order" tickets, that is close to 77 million preventable tickets, and at a blended $10 to $15 per ticket the industry spends up to $1.7 billion answering questions that should never have been asked (Source: Keeyu analysis).

The churn is worse than the tickets. Australian e-commerce loses an estimated $13 to $15 billion a year to customers who leave because of what happened after checkout (Source: Keeyu analysis). The reason it compounds is that there is no second strike: 78% of customers will not shop with a brand again after one bad delivery experience (Source: Convey). And 60% of the reasons customers do not return happen after checkout (Source: Narvar) - not the product, not the price, the delivery experience.

Late is not an edge case here either. When the ACCC took Mosaic Brands to the Federal Court in 2024, the finding was that over 26% of items ordered were dispatched at least 20 days after purchase, and some more than 40 days (Source: ACCC v Mosaic Brands Limited). That is a regulator describing a fulfillment failure as a consumer law problem, which is where a broken promise ends up when nobody catches it upstream.

Then there is the work itself. Agents spend only 19% of their week actually resolving anything. The other 81% goes to administrative friction: jumping between Shopify, the ERP, the WMS and carrier portals to assemble an answer. You are turning empathetic people into human search engines.

Nicola Clement of Top 50 in eComm describes where that ends: "Service levels drop, customers stop returning, and you have to offset that with acquisition when costs are highest. Now you are in a death spiral."

We watched this play out publicly during peak 2023, when White Fox Boutique and Frank Green both buckled under warehouse demand. The pattern was identical each time: no system to detect issues proactively, warehouses overwhelmed, customers finding the delays before the brands did, and a social media firestorm that followed. As Mal Chia of Ecom Nation puts it, "the black hole of communication loses you the customer, and they tell their network. Suddenly you are burning community trust at scale."

The five levels of maturity

We analyzed more than 100 Australian e-commerce brands and found a clear maturity curve. Most sit at the bottom of it. This is not a helpdesk ladder. It is a description of how much of the promise your systems can keep without a human noticing first.

• Level 1, Chaotic reactive. "Customers tell us about problems." No systematic monitoring. WISMO is 50 to 60% of tickets and past 80% in peak, and support costs 4 to 5% of revenue. White Fox at Christmas 2023 is the textbook case.
• Level 2, Organized reactive. "We respond fast, but we are still reactive." Centralized helpdesk, real SOPs, faster replies. WISMO drops to 40 to 50%, costs to 3 to 4%. The ceiling is structural: you cannot respond faster than the customer discovers the problem.
• Level 3, Manual proactive. "We check for problems daily." Someone audits orders each morning across five or more systems. It works, which is the point, and WISMO falls to 30 to 40%. It also breaks at around 1,000 orders a month, because it is a person doing a machine's job.
• Level 4, Predictive. "Systems alert us, but humans must fix it." Algorithms watch carrier feeds and inventory and raise the flag. Leading Australian brands catch 80% of issues before the customer makes contact. The resolution labor is still fully human.
• Level 5, Orchestrated. "Systems prevent and resolve problems automatically." A backorder is detected, the supplier ETA is checked, the customer is notified with alternatives, the CRM updates and the resolution is logged. Twenty-six seconds, no human (Source: Keeyu platform data). WISMO drops below 10% and support costs land at 1 to 1.5% of revenue.

The distribution is the opportunity. Roughly 35% of Australian brands sit at Level 1, 44% at Level 2, 15% at Level 3, 5% at Level 4 and about 1% at Level 5. Four in five brands are being beaten on the promise by a competitor who simply looked first.

The gap this opens on retention is the number worth taking to a board. Reactive brands run repeat purchase rates of 30 to 40%. Proactive ones run 70 to 80% (Source: Shopify, Narvar). Same products, same ads, same prices. The difference is whether the customer found out about the problem from you or from the tracking page.

This is what we mean by proactive e-commerce operations, and why we describe it as a system of action rather than a system of record. A record tells you what happened. A system of action changes what happens next.

Nicola Clement asks the question that reframes the whole category: "What if your customer service team was your loyalty program?" At LSKD, Jade Cameron describes a team where "every single individual in the team operates autonomously", running a consistent CSAT of 4.8 out of 5 and an NPS above 80. That is only possible because the operational floor is automated. Free the team from data entry and they become a concierge service. Leave them in the inbox and they stay a call center.

The truth test: score yourself

Most operations directors believe they are at Level 3. Our assessment of 100+ brands puts around 80% at Level 1. The gap between those two numbers is the whole problem, because you cannot fix what you will not measure honestly.

Twelve questions, four categories, 60 points. Answer them as your worst week would answer them, not your best.

Answer all twelve honestly. Each answer scores 1 to 5 points, for a maximum of 60. Keep a running total as you go, then read your band against the guide below.

Visibility

1. How do you currently monitor order status across systems?

• 1 point - Manual checks in multiple systems
• 2 points - Some integration, but still checking 6-8 platforms
• 4 points - Mostly unified, occasional manual checks
• 5 points - Unified real-time dashboard

2. How quickly can you identify which orders are at risk of delay?

• 1 point - Only when a customer contacts us
• 2 points - Daily manual audit
• 4 points - Automated alerts, checked regularly
• 5 points - Real-time exception monitoring

3. Can you see carrier performance issues before they become customer complaints?

• 1 point - No
• 2 points - Sometimes, if we manually check
• 4 points - Yes, with automated reports
• 5 points - Yes, real-time carrier performance tracking

Communication

4. How do customers find out about order delays?

• 1 point - They discover it themselves and contact us
• 2 points - We tell them after we discover it
• 4 points - Automated alerts for some issues
• 5 points - Proactive communication before they know there is a problem

5. What percentage of your support tickets are WISMO?

• 1 point - 50 to 60%
• 2 points - 40 to 50%
• 3 points - 30 to 40%
• 4 points - 20 to 30%
• 5 points - Under 20%

6. How do you handle carrier delays?

• 1 point - Wait for the customer to ask
• 2 points - Respond when the customer contacts us
• 4 points - Proactive email after we manually detect the delay
• 5 points - Automated proactive communication within hours

Workload

7. How much of your support team's time is spent firefighting versus strategic work?

• 1 point - 85%+ firefighting
• 2 points - 70 to 85% firefighting
• 3 points - 50 to 70% firefighting
• 4 points - 30 to 50% firefighting
• 5 points - Under 30% firefighting

8. How does your team handle peak season?

• 1 point - Chaos, emergency hiring, team burnout
• 2 points - Stressful, everyone working overtime
• 3 points - Manageable, some overtime but controlled
• 5 points - Seamless, automated systems scale effortlessly

9. How long does it take to resolve a typical backorder ticket?

• 1 point - 15+ minutes, manually checking multiple systems
• 2 points - 10 to 15 minutes, some automation but still manual steps
• 4 points - 5 to 10 minutes, mostly automated with human verification
• 5 points - Under 5 minutes, AI resolves end to end without agent involvement

Customer impact

10. What is your repeat purchase rate?

• 1 point - Under 50%
• 2 points - 50 to 60%
• 3 points - 60 to 70%
• 4 points - 70 to 80%
• 5 points - Over 80%

11. How often do customers post negative reviews or social complaints about delivery?

• 1 point - Frequently, multiple times per month
• 2 points - Occasionally, a few times per month
• 4 points - Rarely, maybe once per month
• 5 points - Almost never, less than once per quarter

12. What percentage of revenue do you spend on customer support?

• 1 point - 3.5 to 4%
• 2 points - 2.5 to 3.5%
• 3 points - 2 to 2.5%
• 4 points - 1.5 to 2%
• 5 points - Under 1.5%

0 of 12 answered. Score: 0 / 60

Whatever you scored, Paul Greenberg's rule applies at every level: "Where possible, treat cause, not symptoms." A better reply to a WISMO ticket is a symptom fix. Knowing the parcel was going to be late before the customer did is a cause fix.

What your score means

Total your twelve answers, then find your band.

• 12 to 24 points - Level 1 and 2: Reactive. Customers discover problems before you do, or you are efficient at being reactive but still reactive. Start with Phase 1 and 2 of the roadmap: audit your true WISMO rate, then run one manual proactive check daily.
• 25 to 36 points - Level 2: Organized reactive. You respond fast, but you cannot respond faster than the customer discovers the problem. Start proactive communication on one issue type and measure the ticket drop.
• 37 to 48 points - Level 3: Manual proactive. Someone checks for problems daily and it is working, but it breaks around 1,000 orders a month. Your next move is automating exception detection.
• 49 to 54 points - Level 4: Predictive. Systems find the problems, humans still fix them. Add automated resolution for your highest-volume routine issue and measure the labor you get back.
• 55 to 60 points - Level 5: Orchestrated. Unified visibility and automated resolution. Scale to more issue types and keep optimizing. Most brands who score here have one or two issue types still running manually.

Three brands that made the shift

These are not hypotheticals. Each of these operators ran the reactive version first.

EHP Labs: from five minutes to 26 seconds

A global supplements brand running thousands of orders a month across multiple warehouses and systems. Jeanette, EHP's CX operations lead, described the old reality: "We were never on the front foot. We would not recognize a problem until it was pointed out down the line. That meant long, drawn-out resolutions and a lot of stress." Backorders meant hopping Shopify to NetSuite to ShipBob to email, five minutes each. A hundred backorders in a sale was 8.3 hours of manual labor.

After connecting the stack into one source of truth with automated workflows, backorder handling went from five minutes to 26 seconds, a 92% reduction in resolution time. In Jeanette's words: "What used to take us days or weeks to catch, we catch in minutes. It is happening in real time."

Clutch Glue: a three-person job, automated

An FMCG brand needing delivery visibility across DTC and B2B. Annabel, founder and CEO: "There is absolutely nothing worse than receiving an email from a disgruntled customer saying their order is delayed. Responding to those emails was honestly a two to three person job." A unified view of every order's health absorbed that workload entirely. Her assessment: "I would have to hire at least five people on the spot to just absorb the capacity this provides." They now diagnose whether a problem is a postage problem or a warehouse problem instantly, and tell customers before customers tell them.

Budgy Smuggler: one person, 12,000 orders a week

An iconic Australian swimwear brand where a single person, Emma, owns the entire post-purchase experience through a peak that reaches 12,000 orders a week. The reactive model broke at that volume, because one person cannot manually check 12,000 orders. Proactively nudging customers whose parcels were awaiting collection cut return-to-sender parcels by 40 to 70% and saved $1,000 to $1,750 a month in wasted shipping. Emma calls the platform "the Bible. It is my wingman. If it disappears, I would miss having that second eye."

The pattern across all three is the same, and it is worth being precise about it, because this is not a story about replacing people. It is about taking the robotic work off humans so they can do the part that actually builds loyalty.

The six-week roadmap

Do not try to boil the ocean. Four phases, in order.

• Phase 1, week 1. Audit your true WISMO rate. Not ticket counts. Total shipping tickets divided by total orders shipped. Above 20%, or spiking past 50% in peak, means the promise is breaking systematically rather than occasionally.
• Phase 2, weeks 2 and 3. Triage manually. Pick your single highest-volume failure, usually the stuck or delayed parcel, and check for it every morning by hand. Contact those customers first. This is deliberately unscalable. The point is to prove the ticket volume falls before you automate anything.
• Phase 3, weeks 4 and 5. Automate the detection. Replace the human doing the morning hunt with a rule that watches every order continuously. Detection is where the leverage is, and it is the step most brands skip straight past on their way to buying a chatbot.
• Phase 4, week 6 onward. Automate the resolution. Once detection is trustworthy, let the system act: check the ETA, notify the customer, offer the alternative, update the record. Start with one issue type, measure the labor returned, then add the next.

Emily Elvey's warning is the reason for the sequence: "I have seen brands suffer month-long ticket backlogs because they did not optimize until it was too late." Peak season does not create these problems. It reveals them.

Building the business case

Operations directors struggle to get budget because support reads as a cost center on the P&L. Orchestration is not an efficiency play, it is a profitability play, and the arithmetic is not subtle.

Take a growth-tier brand doing 15,000 orders a month. At Level 1, support runs 3.5% of revenue, or $611,100 a year. At Level 5 it runs 1.2%, or $209,520. That is $371,580 in net annual savings, more than $1.1 million over three years, against a platform cost measured in single-digit thousands per month.

Three objections come up in every boardroom, and each one inverts on inspection:

• "Adding another platform increases costs." Reactive operations already tax you 4 to 5% of revenue in labor and churn. Orchestration costs roughly $2k a month to remove a $30k a month problem. The question is not whether you can afford it.
• "Our systems do not integrate." That is the reason to do it, not the reason to wait. Manually checking five systems is a tax paid on every single order. The platform is the integration layer.
• "We do not have time to add proactive checking." Proactive audits take about 15 minutes. Reactive spikes take the week. You reclaim 40 to 60% of the team's time for work that actually drives revenue.

Three protocols to deploy now

Proactive communication works because it closes the anxiety gap: the time between a problem occurring and the customer finding out. In reactive operations that gap is days, and the customer fills it with the worst available explanation. Mareile Osthus of humii puts the principle well: "Every time a shopper has to go find out something, you have already added friction. The retailers who win are the ones who remove guesswork before it even enters the customer's mind."

• The proactive apology. When an order sits unfulfilled beyond three days, tell them before they ask. Be specific about the backlog or supplier delay. Acknowledging it early buys roughly 48 hours of goodwill and protects the brand through a warehouse failure.
• The pre-dispatch correction. When a courier API rejects an address, usually a missing unit number, do not wait for the failed delivery. Send an automated SMS asking the customer to verify while the parcel is still in the building. This is the cheapest fix in the entire chain.
• The collection nudge. People miss the card in the mailbox. A digital reminder that a parcel is waiting cuts return-to-sender rates by 40 to 70% and saves the double freight on every parcel it rescues.

All three are the same move: detect the issue before the customer does, communicate transparently, and give them a clear next step.

The operators who contributed

This playbook is not a vendor white paper with a survey stapled to the front. The operators below gave their time and their words to it, and several of them have been living this problem longer than Keeyu has existed. Where a quote appears above, it belongs to one of them.

• Mareile Osthus, co-founder of humii, on friction and removing guesswork.
• Carla Penn-Khan, co-founder of Profit Peak, on social media fallout and proactive communication.
• Mal Chia, founder of Ecom Nation and host of This Week in Ecommerce, on the black hole of communication.
• Jade Cameron, Head of Customer Service at LSKD, on autonomy and service as a loyalty program.
• Paul Waddy, founder of Learn Ecommerce, on MER and the post-purchase experience.
• Emily Elvey, founder of Emily Elvey Consulting, on optimization and ticket backlogs.
• Paul Greenberg, founder of NORA and independent e-commerce consultant, on customer attrition and self-service friction.
• Nicola Clement, founder of Top 50 in eComm, on team silos, profitability and the death spiral.
• Nathan Bush, host of the Add to Cart podcast, on customer ghosting, grudge purchases and staffing.

Related reading

• Post-purchase operations, explained
• The reactive tax
• The five levels of post-purchase operations maturity
• Proactive operations case studies
• WISMO: where is my order

• 1. Convey, State of the Consumer 2023: Delivery Experience Report. Global delivery satisfaction study, n=2,500 online shoppers across the US, UK and Australia.
• 2. Keeyu analysis, based on the Australia Post eCommerce Industry Report 2024 ($67.5B Australian e-commerce GMV), Forrester Research benchmark data on churn from poor delivery and fulfilment, and lifetime value calculations across Australian DTC brands.
• 3. Narvar, Post-Purchase Experience Report 2023. Survey of 3,000+ online shoppers on the primary drivers of non-repeat purchases.
• 4. Zendesk, Customer Experience Trends Report 2024: E-commerce Edition (10M+ support tickets across 5,000+ brands); Gorgias, E-commerce Support Benchmark Report 2024 (WISMO classification across 3,000+ Shopify brands).
• 5. Keeyu analysis of 100+ Australian e-commerce brands' financial data, 2023 to 2024, cross-referenced with Gartner, Customer Service Cost Benchmarking Report 2024.
• 6. Keeyu analysis. Operational maturity assessment of 100+ Australian e-commerce brands, 2023 to 2024.
• 7. Shopify, E-commerce Benchmarks Report 2024 (repeat purchase behaviour across 50,000+ merchants by CX quartile); Narvar, Consumer Report: Expectations vs Reality in E-commerce 2024.
• 8. ACCC v Mosaic Brands Limited, Federal Court proceedings, 4 March 2024. Finding: over 26 per cent of items ordered were dispatched at least 20 days, and in some cases more than 40 days, after the purchase date.
• 9. Keeyu platform data, 2024. WISMO rates, ticket resolution times and operational benchmarks.
• 10. Verified results from EHP Labs, Clutch Glue and Budgy Smuggler, collected through direct operator interviews during 2023 to 2024 implementation periods.

Statistics are current as of the playbook's publication (2024 to 2025) and some may have been updated at source since collection. Where a reference has no link, the source is a subscription or PDF report rather than a public web page.