Level 3 · Recommend: where should the mobile work go?

From numbers to a recommendation

Jordan does not need the query. He needs a paragraph he can say out loud in a planning meeting, and it has three parts, sometimes taught as what, how much, so what:

• What is the main finding? Mobile loses people at checkout, not while browsing.
• How much supports it? 44.1% of sessions reach three pages on both mobile and desktop; of engaged, signed-in sessions, 12.9% buy on mobile against 21.4% on desktop. Mobile’s share of traffic rose from 47.8% to 58.0% in two years.
• So what does it mean for the decision? Fund the checkout steps, not a full rebuild.

The decision is sized to the finding. The data points at one part of the journey, so the recommendation spends money there first.

Caveats belong in the paragraph

Jordan asked what the data cannot see. Say it in plain words, in the same paragraph:

• Anonymous sessions are unmeasured. 23,002 sessions, 57.5% of the total, have no customer id, and none converted. That is how purchases are recorded, not proof that those visitors never buy.
• Cross-device journeys are invisible. Someone who browses on a phone at lunch and pays on a laptop at night counts as a mobile session without a purchase. Part of the mobile gap may be this, and the data cannot separate it.
• Where, not why. The funnel shows the step; a test of a new checkout shows the cause.

Caveats are not hedging. They tell the reader what the result can carry.

The traps in the other options

• One number instead of a funnel. 2.41% against 3.87% is true, but it hides that browsing is level. A rebuild aimed at the pages would fix what is not broken.
• A blended rate read as a trend. The overall rate went from 2.99% to 3.08% while every device improved. The traffic moved towards the device with the lowest rate. Keeping 2023’s device mix, 2025 would have converted about 3.24%. This is the mild form of Simpson’s paradox: always split a rate by its segments before you call it flat.
• A tracking gap read as behaviour. “Anonymous visitors never buy” confuses what the data records with what people do.

The report structure behind this paragraph

In a written report this paragraph is the executive summary: what was analysed (three years of web sessions), what was found and what is recommended, in five or six lines. Below it come the objective (the SMART question from the first level), the data and its limits (the step 1 counts), the method (the funnel stages), the findings with the chart, the recommendation and the next step: a test of the new mobile checkout against the current one.