What LTV is and why fraud distorts the estimate

The abbreviation LTV does not appear on the Retention reports page. The help describes long-term value differently: the number of visits, page views, conversions, or revenue of a group over time. The page does not give a formula for “revenue per person over a lifetime” or an adjustment for invalid clicks. If by LTV you mean that long-term value, you read it under “Reports → Retention,” not in a field of that name.

The help suggests judging the quality of ad traffic by comparing sources and campaigns on who returns more often, makes more visits, and brings more revenue. It also warns that a campaign with a good near-term conversion can bring people who never come back, while a campaign with a weaker first conversion can bring a more active audience later. The price of the nearest lead therefore does not replace a long slice. How the lead price itself is built is covered in the article on CPA.

A cohort accumulates; Retention cuts by period

Cohort analysis on the same page answers how value or activity accumulates over time for people who share a first-visit trait. Retention answers what share comes back and what value they create in a specific period after entering the cohort. Those are two questions about one group of visits, not two LTV formulas.

The windows are days, weeks, or months, calendar-based or from the first visit. You can take new users or everyone. The cohort grouping is a ready-made one or your own.

For the Retention and Rolling Retention table you can choose users, visits, view depth, time on site, goal completions, and revenue per user. The “Users” metric is not available in cohort analysis. A cell holds two numbers: a percent or an average on top, and the number of users or the metric sum below.

“Revenue per user” is calculated from the goal or commerce revenue you passed. The page does not say Direct fraud has already been subtracted from it. The top number is a percent or an average for the period. The bottom number is the volume that average stands on. Comparing sources by the top percent alone, without the bottom number, means comparing shares of different sizes.

A campaign with a high near-term conversion and a campaign that brings people back later diverge on this table by period, not by one sum “for all time.” You set the window: a day, a week, or a month, calendar-based or from the first visit.

Robots in the denominator you use to compare sources

By default, Metrica statistics include robot visits identified by behavior. The help gives a direct example: robots can lower view depth and raise bounces. The example is qualitative: the page does not give a bounce share that robots “usually” add.

You can compare a slice without robots and with robots in the report setting “Data: without robots,” using “Compare with data that includes robots.” You can remove robots from a single report with a segment. Until the comparison is on, a source’s depth and bounces are counted together with those visits. A source whose robots inflated bounces and shortened depth looks worse on the same metrics Retention offers for traffic quality: visits, depth, and returns.

What the robot-share percent means on its own, and what not to expect from it, is covered in the article on robot share. That percent is not subtracted automatically from the “revenue per user” cell: revenue comes from the goal or commerce revenue you passed, and the page does not say Direct fraud has already been taken out of it.

CTR is not in the cohort. An ad’s click-through rate does not answer whether the person came back a week later or what revenue the group brought in that window. Budget lost to clicks with no buyer is also counted outside Retention: the page gives no adjustment for invalid clicks. How not to add clicks already refunded into the estimate is in budget loss.

Fraud distorts this estimate where it sits inside the slice you use to compare sources. Robot visits are in the statistics by default and can raise bounces and lower depth. Revenue per user can include whatever was passed as goal or commerce revenue, with no promise that Direct clicks were subtracted. A near-term conversion, even a clean one, does not guarantee that people will return, by the help’s own warning. Long value in Metrica is accumulation and the share that returns by period, not a field named LTV.