Conversion rate is the most reported number in digital marketing and one of the least actionable. It tells you something went wrong somewhere between arriving and buying, which you knew, and nothing about where.
Four things in particular disappear into it.
1. Which step failed
A site converting at 2 percent could be losing almost everyone at the landing page and converting brilliantly thereafter, or holding attention beautifully and dying at checkout. Same headline number, opposite problems, opposite fixes.
The fix is a funnel rather than a rate, and the diagnostic that matters is which step loses the most people relative to how many reached it. Absolute losses point at the top of the funnel by construction, since that is where the people are.
2. What each step cost
Percentages measure the leak in people. Paid traffic leaks money, and the two rankings disagree more often than they agree.
- Step 1 to 2: people lost3,100
- Step 1 to 2: spend lost$5,580
- Step 3 to 4: people lost240
- Step 3 to 4: spend lost$432
Worked example from a 5,000 click, $9,000 campaign. Every entrant cost the same to acquire, so the biggest count of losses is also the biggest financial loss. That equivalence is exactly why the checkout is rarely the emergency it feels like.
The same funnel, in people and in money. Step 1 to 2: people lost: 3,100. Step 1 to 2: spend lost: $5,580. Step 3 to 4: people lost: 240. Step 3 to 4: spend lost: $432.
That worked example is unpacked properly in put a price on every funnel step. The short version is that the step everyone wants to redesign is usually the cheapest one to lose people at.
3. Who the traffic was
A blended conversion rate mixes a channel converting at 6 percent with one converting at 0.5 percent. When the blended figure moves, the most common cause is a change in traffic mix rather than any change in how well the site works.
This produces an entire genre of false result. A campaign scales up, the blended conversion rate falls, and someone concludes the site got worse. The site did nothing. The denominator changed. Your traffic sources do not convert equally covers the spread that makes this possible.
4. Who was never counted
The last one is the least discussed and the one we care most about, because it is a measurement failure rather than an analysis failure.
- 15–30%Blocked by ad blockers, at the top end for technical audiences
- variesConsent declined, which in the EU can be a large share of a page's visits
- unknownCross-device journeys counted as two unrelated people
- countedBot traffic, unless it is filtered before it reaches the numbers
Bots are the one that cuts in the wrong direction. Blocked and unconsented visitors shrink your numbers, so your conversion rate looks better than reality. Unfiltered bot traffic inflates your visit count with sessions that never convert, so it drags conversion rate down and makes traffic look better and quality look worse at the same time.
None of these are fully solvable. What is achievable is knowing the size of each and saying so on the report, so that a 2.1 percent conversion rate arrives with the honest caveat that it is 2.1 percent of the visitors you could see.
What to watch instead
- 01Step to step conversion, so you know which step failed
- 02Cost per completion at each step, so you know which failure was expensive
- 03The same funnel per traffic source, because they are different funnels
- 04The count of unattributable visitors, printed next to every cost figure
- 05Time to convert, which tells you whether your attribution window is even plausible
The last one catches a surprising number of B2B teams. If your median time to convert is 40 days and your ad platform closes its attribution window at 7, the platform is structurally incapable of seeing most of your conversions, and no amount of arguing with the dashboard will change that.