Start
Starting audit and score
A score for the measurement setup and the funnel: what is not measured and what is measured wrong.
- Sales compared with the analytics tool
- Missing and double-counted events
Growth
First we fix the measurement: which channel brings people in, at which step they drop off. Then we set up experiments on the biggest loss, one at a time; each experiment’s hypothesis, metric and result stay on record.
Experiments that do not win are logged too, so the same idea is not tried twice.
- Measurement
- Event list, channel tags, sales matching what the analytics tool reports.
- Funnel
- What stays and what drops off at each step from visit to purchase.
- Experiment
- One variable, a hypothesis and metric written in advance, enough visits.
- Channels
- Organic search, AI answers, social, email, ads; the share and cost of each.
04.1The work in three parts
Most growth work starts with fixing something that was measured wrong.
aWhat we do
- Measurement auditDo sales in the analytics tool match real sales; if not, where do they go missing.
- FunnelLoss step by step, split by device, channel and new versus returning visitors.
- ExperimentsHypothesis, primary and guardrail metrics, decision rule; written before the experiment starts.
- Channel mixWhich channel brings new customers and which brings returning ones.
bWhat we deliver
- Event listWhat is measured and why; set up and verified.
- Funnel dashboardA single screen, updated weekly.
- Experiment logEach experiment’s hypothesis, duration, result and decision.
- The next three experimentsRanked by expected impact and difficulty.
cHow we measure
- ConversionRate at each step; weekly and next to the previous period.
- Customer valueRepeat purchase rate after the first order, and average order value.
- Channel costCost per new customer by channel, as far as it can be measured.
- ExperimentsThe difference in the primary metric and its confidence interval; no damage to the guardrail metric.
04.2Funnel and experiment log
A store’s funnel over four weeks. Click a step to see an experiment idea for the loss at that step.
Conversion funnel
One store, four weeks
-
Visit
48,200 · start
- What happens
- All sessions arriving at the store, split by channel.
- Experiment idea
- No experiment at this step; channel shares are watched.
- Primary metric
- Sessions per channel.
- Guardrail metric
- Bounce rate.
-
Product view
21,700 · 45% move on from the previous step, 55% drop off
- What happens
- Fewer than half of visitors open a product page.
- Experiment idea
- A best-seller list on the home page instead of a sliding banner.
- Primary metric
- Product views per session.
- Guardrail metric
- Page load time.
-
Add to cart
3,900 · 18% move on from the previous step, 82% drop off
- What happens
- The biggest loss is here: 18% of people who see a product add it to the cart.
- Experiment idea
- Shipping cost and delivery time shown under the price, before adding to cart.
- Primary metric
- Rate from product view to add to cart.
- Guardrail metric
- Return rate, checkout completion.
-
Checkout start
1,640 · 42% move on from the previous step, 58% drop off
- What happens
- Fewer than half of those with a cart move on to checkout.
- Experiment idea
- Order total and instalment options shown in the cart from the start.
- Primary metric
- Cart to checkout rate.
- Guardrail metric
- Average order value.
-
Purchase
1,050 · 64% move on from the previous step, 36% drop off
- What happens
- A third of those who start checkout do not finish.
- Experiment idea
- Guest checkout and a clear card error message.
- Primary metric
- Checkout completion rate.
- Guardrail metric
- Payment error types, support requests.
| Hypothesis | Primary metric | Result |
|---|---|---|
| If delivery time shows under the price, add-to-cart goes up. | Add to cart | +9% · kept |
| Guest checkout increases checkout completion. | Checkout completion | +6% · kept |
| Removing the sliding home page banner increases product views. | Product views | No difference · rolled back |
Bar: share that moved on from the previous step; the dashed part dropped off.A made-up example; it does not belong to any client.
04.3Related packages
The three packages that most often go with growth work.
Visibility
Client dashboard
The funnel, channel shares and running experiments on one screen, updated weekly.
- Shared within the team
- Linked to the experiment log
Continuity
Regular reporting
Monthly growth report: what was tried, what won, what was removed, what is next.
Write where you think you lose the most.
Tell me which analytics tool you use and how many sales you had in a week, and I will first check whether the measurement adds up.