First, read the rate on ended trials
The trial-to-paid conversion rate entry is precise about the count: take every trial that started in a window, wait until the last of them has expired, and divide the ones that paid by the ones that started. Dividing this month's payments by this month's starts mixes groups and moves whenever starts or trial length change. So rebuild the figure by trial-start CohortA group of customers who share a starting event and period, tracked together over time., show only cohorts whose trials have all ended, and split by plan and by acquisition channel, because a channel that buys many cheap starts can pull the blended rate down without anything in the trial changing.
Next, make the inside of the trial visible. Google's help centre page about events describes how analytics records user interactions as events, and the sign-up flows guide sets out which lifecycle events belong to which system: the sign-up and activation in product data, the trial start and first payment in billing, each sent on to analytics. Record the milestones that matter for this product, such as the first project created or a colleague invited, with the account they belong to. Then compare paying and non-paying trials step by step: the first milestone the two groups stop sharing is where the diagnosis starts.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Many trials never complete the first session or never return | Step rates from the trial button to the first screen, and what is asked on each | A trial start that asks for more than the first step needs | Ask only what the first session uses; move the rest inside the product |
| Users who stalled and users who progressed receive the same emails | Whether each onboarding email is sent on a day number or on a missing step | Onboarding emails sent on fixed days rather than triggered by the step not taken | Trigger each email from the step the user has not taken |
| Teams adding colleagues or reaching limits lapse without a conversation | Which trials show the signals that preceded past purchases, and who contacted them | No sales contact for trials that show buying signals | Route signalling trials to an owner for a helpful first conversation |
| Active users visit pricing near the end and leave without choosing | Pricing page exits from trial users, and whether the trial's plan is obvious there | A pricing page that makes the plan choice hard at the end of the trial | Show which plan fits what the trial user did and keep the choice short |
| The trial's length or card rule was set once and never questioned | Whether any test compared lengths or card rules on paying customers per sign-up | Trial length or card requirement chosen without a test | Run the length or card test before changing either |
| The rate fell after a change, or differs between dashboards | Whether each report divides by ended trials from one start cohort | Trial-to-paid read on the wrong cohort, mixing trials still running with ended ones | Report by start cohort only once every trial in it has ended |
Fix, verify and prevent each cause
A start that asks too much
Every question before the first screen is a reason to stop, and a trial that never reaches its first session cannot convert. The sign-up flows guide treats the trial start as a funnel of its own and gives the rule: ask only what the first session needs, and ask the rest later, once the person has seen enough to want to answer, or infer it from what they do. Questions added for sales routing or attribution are the usual culprits.
Fix: list every field and step between the trial button and the first screen, keep only those the first session uses, and move the rest into the product.
Verify: the share of trial starts that complete the first session rises in the cohorts after the change, and their trial-to-paid rate, once those trials end, holds or rises.
Prevent: require any new sign-up question to name the first-session feature that uses it.
Emails on a calendar, not on behaviour
A sequence that sends the same message on the same day to everyone tells finished users to do what they have done and leaves stuck users waiting. The software onboarding emails tutorial builds the alternative: activation defined from product data, a map of the steps between sign-up and activation, and each email triggered when a user has not completed its step after a fair wait. It also times trial-end messages from the trial's end attribute rather than a day count.
Fix: rebuild the sequence around the steps that separate paying from non-paying trials, one email per stalled step, with exits on activation.
Verify: activation within the trial rises against a holdout, and emails per user fall.
Prevent: whenever the product adds or renames a step, update the step map and the event that triggers its email together.
Buying signals nobody answers
Some trials behave like buyers: several people from one organisation, heavy use of the core feature, a plan limit reached, a visit to the security page. Left to self-serve, some of them stall on a question only a person could answer. The guide to product-qualified leads chooses those signals from the business's own history, comparing accounts that bought with those that did not, and routes each signalling account to one owner with an agreed time to first contact.
Fix: define two or three signals from past cohorts, route matching trials to an owner, and open with help rather than a pitch.
Verify: trials that met the signals and were contacted convert above comparable signalling trials from before the change.
Prevent: review the signal definitions with sales each quarter, retiring any that stopped predicting purchase.
A pricing page that reopens the decision
A trial user who arrives at the pricing page ready to pay should find a short choice: the plan that fits what they did in the trial, its complete cost, and a button that keeps their work. The pricing pages guide treats existing users as a different journey from new buyers, asks for plan cards that name the buyer they suit and a call to action that matches how each plan is bought, and warns against a grid of ticks that makes the buyer do the packaging.
Fix: route trial users to a version of the plan choice that shows the plan matching their usage first, with the cost and what carries over.
Verify: the share of trial users who reach the pricing page and then pay rises, without a rise in early downgrades or refunds.
Prevent: report trial users' pricing visits apart from new visitors, so a change aimed at one group is never judged on the other.
Length and card rules never tested
The trial's length and whether it takes a card at sign-up shape the rate more than most copy changes, and both are often chosen once by imitation. The free trial length test randomises sign-ups to lengths and judges them on paying customers per sign-up at a fixed horizon, with churn as a guardrail; the card-required trial test does the same for the card step. This page does not restate their designs.
Fix: if the length or card rule has never been tested, run the relevant test rather than switching on a hunch.
Verify: the decision follows the test's own rule on paying customers per sign-up, not on the trial-to-paid rate alone.
Prevent: treat any change of length or card rule as the start of a new series, so later cohorts are never compared with earlier ones as if nothing changed.
A rate read on unfinished trials
Sometimes the trials are fine and the arithmetic is not. A rate calculated before every trial in the window has ended counts running trials as failures, and a rate that divides one month's payments by the same month's starts moves whenever volume or length changes.
Fix: rebuild the dashboard by start cohort, show only finished cohorts, and keep the conversion rate (CVRShare of visits or clicks that convert.) per visitor beside it.
Verify: every report gives the same figure for the same cohort.
Prevent: store the definition with the dashboard and annotate any change of trial design on the chart.
A prevention routine for the trial
Treat the trial as a product with its own funnel review. Each week, read the latest finished trial-start cohorts: starts, first-session completion, activation, signalling accounts and who contacted them, pricing page visits from trial users, and payments, by plan and by channel. Keep the activation definition fixed and written down; the sign-up flows guide warns that an event chosen because it is easy to count defeats the purpose, so choose the step that separates paying cohorts from the rest and change it only with evidence.
Before any release that adds a sign-up question, renames a product step or changes the plans, check the three places that depend on it: the events recorded for the trial, the onboarding sequence that listens for them, and the pricing page shown to trial users. Before any change of trial length or card rule, plan the test and annotate the start of a new series. Each quarter, sit with sales and compare the accounts they spoke to with those that paid unaided, and retire signals that no longer predict buying.
Finally, keep a short register of every change to the trial: what changed, the date, and the first cohort affected. Most false alarms about trial conversion trace back to a change nobody wrote down, and most real improvements are only provable when the cohort before and after can be named. The conversion troubleshooting section lists the neighbouring symptoms, such as abandoned forms, that can sit at the very start of the trial.
Quibbix Timesheets finds the step that predicts payment
Quibbix Timesheets, an invented timesheet app with a fourteen-day trial and no card, reports trial-to-paid at 6% and plans a discount. Rebuilt by start cohort and limited to finished trials, the figure for its last four cohorts is 7.8%: 312 paid out of 4,000 trials. Milestones recorded as events show the split. Of the 1,040 trials that invited at least one colleague, 251 paid, 24.1%; of the 2,960 that never invited anyone, 61 paid, 2.1%. Onboarding emails go out on days one, three, seven and thirteen to everyone, and none asks the user to invite a colleague. Nobody contacts teams that invite three or more people, even though 118 such trials lapsed. Quibbix triggers an invitation email when a user has logged time twice without inviting anyone, and routes trials with three or more users to an account manager. Over the next four finished cohorts, invitations reach 37% of trials and trial-to-paid rises to 10.9%, while early cancellations hold steady. The discount is dropped.
Common mistakes
Reading trial-to-paid before the trials have ended. Offering a discount to users who never reached the product's value. Comparing a card trial's rate with a no-card trial's. Sending every trial user the same emails on the same days. Judging a sign-up change on starts instead of payments.
