The Playbook.
This is the operating side: how I decide where to look before I start changing things.
Not a list of universal best practices. A set of checks designed to stop one plausible explanation becoming the only explanation.
The expensive mistake is often not choosing the wrong fix. It is correctly fixing the wrong constraint.
1. Find the constraint first
I do not automatically work traffic → CTR → conversion. I look for the first stage that is genuinely limiting the commercial outcome.
Demand
Is there enough qualified demand, or are we optimising a page nobody useful is reaching?
Click
Does the promise in the ad, result or email earn the next step from the right person?
Decision
Can the customer identify the product, understand the price, calculate quantity and remove the expensive doubts?
Checkout
Is there actual checkout friction, or did the weakness begin earlier?
Fulfilment
Can stock, warehouse, delivery and service keep the promise being sold?
Repeat
Did we acquire a useful customer, or merely create an expensive first order?
2. Follow the money properly
Useful, but only the top line.
What the order contributes before acquisition and fulfilment costs.
Judge against the type of customer acquired, not just the first basket.
Picking, packing, delivery, returns and service can make two identical revenue orders economically different.
Repeat behaviour, full-price behaviour and category movement decide how much acquisition you can rationally afford.
3. Replatforming / ERP migration
Changing the system underneath ecommerce is not an IT side project. It can affect orders, stock, SEO, warehouse work and customer promises at the same time.
Map the data
- Products and variants
- Customers and orders
- Stock locations
- Pricing and tax rules
Protect continuity
- Order capture
- Payment / refunds
- Stock accuracy
- Dispatch and exceptions
Protect discovery
- URLs and redirects
- Feeds
- Tracking
- Search / category structure
Protect the people
- Picking workflow
- Training
- Fallback process
- Who owns failures on day one
4. Warehouse throughput
Before adding people, I want to know how much work each order creates.
How many times is the order read, keyed, printed, moved or checked manually?
Are people walking because of the order, or because the system makes them?
How much capacity disappears into missing stock, substitutions, split orders and bad data?
What promise can the operation repeatedly support without heroics?
5. New product / procurement check
Demand proof
Do customers already want the product, the job it solves, or a close substitute?
Landed economics
Margin after buying, freight, storage and fulfilment — not just supplier price.
Stock risk
MOQ, lead time, seasonality, damage risk and how quickly cash comes back.
Commercial leverage
Exclusivity, private label, range fit, wholesale potential or a reason the customer should buy it from you.
6. Retention without noise
Checkout beats cart. Cart beats browse. A purchase should stop recovery.
Send reassurance, compatibility, stock, delivery or project help — not another reminder because the flow has a spare slot.
Test incentives against incremental gross profit. Do not teach people to abandon for a code.
High intent can justify urgency. Low intent cannot justify harassment.