Learning path · Level 6 · Tutorial 3

Prioritize Conversion Improvements from Evidence

Combine funnel counts and structured usability observations to identify plausible conversion barriers and prioritize improvements. Work from report tables and interface descriptions presented in the tutorial, with no analytics account or live store required.

  • Level: Intermediate
  • Duration: 20 min
  • 8 steps
Start tutorial Back to the path
Prioritize Conversion Improvements from Evidence

What you will go through

  1. Define the Funnel Before Reading the Losses Define the ordered session funnel and its counting contract before interpreting losses. 2 min
  2. Calculate Completion and Drop-Off Calculate step completion rates, lost-session counts, and drop-off rates from adjacent milestones in the Cedar Trail simulated funnel. 3 min
  3. Turn Funnel Losses into Investigation Questions Use funnel losses to choose an investigation focus while keeping explanations provisional. 2 min
  4. Write a Neutral Usability Task Create a neutral usability task for the Cedar Trail simulated case and define what an observer should record. 2 min
  5. Separate Observations from Interpretations Classify simulated usability-case records without turning observations or self-review into unsupported causal or frequency claims. 3 min
  6. Build an Evidence-Linked Backlog Record Connect a plausible conversion barrier to evidence, a calibrated confidence rating, and a next check that reduces uncertainty. 3 min
  7. Rank Improvements by Impact and Effort Make an evidence-led ranking of improvement candidates while keeping likely impact, effort, and uncertainty distinct. 3 min
  8. Finalize and Defend Your Priorities 3 min

What you will learn

  • Calculate step completion and drop-off rates from a consistently defined funnel table.
  • Explain what a large funnel loss suggests and what it cannot establish about its cause.
  • Draft a neutral usability task and classify case observations separately from assumptions.
  • Rank proposed improvements by evidence, likely impact, and effort, identifying what needs further validation.

Before you start

  • Draft a Landing Page That Matches Visitor Intent
  • Review Product Pages, Carts, and Checkout
  • Set Marketing Objectives and Calculate Core KPIs
  • Plan GA4 Events and Key Events
  • Read GA4 Acquisition, Engagement, and Key-Event Reports

Step 1 of 8

Define the Funnel Before Reading the Losses

Define the ordered session funnel and its counting contract before interpreting losses.

See the funnel as an ordered path

Start with a comparable journey

Conversion rate optimization is an evidence-led cycle: identify a possible barrier, propose a focused change, then check what happens. Before interpreting any loss, make sure the funnel itself measures a consistent journey.

An ordered funnel is not a list of unrelated totals. Each later milestone must follow the earlier one for the same counting unit. In this case, the unit is a session.

An ordered session funnel

A four-stage funnel narrows from a product bottle to a cart, checkout form, and confirmation checkmark, showing that each later stage is a subset of the earlier stage.

Simulated case: the ordered milestones are Product view, Cart addition, Checkout start, and Order confirmation. A session can reach a later milestone only after reaching the earlier milestones in that same session.

Read the report with its rules

Simulated case funnel report

This report is a closed, ordered funnel for Cedar Trail's Trail Bottle. Read each number as sessions, not raw button clicks or event totals.

Funnel counts

Simulated case: eligible sessions entered through Product view during the same seven-day reporting period. They were observed through the end of their session.

Each session counts at most once per milestone.
Ordered milestoneSessions
Product view1,000
Cart addition400
Checkout start250
Order confirmation75

The counting contract

The counting contract makes the rows comparable. Keep it intact before drawing conclusions from the counts.

Simulated case: rules used for every row in this funnel.
Part of the contractRule
Counting unitSession
Entry requirementA session must reach Product view
Milestone orderProduct view → Cart addition → Checkout start → Order confirmation
Time windowSession starts in the same seven-day reporting period
Completion ruleMilestones occur in order within that same session
Repeat countingAt most once per milestone per session

Practice: preserve the contract

Check the comparison

Which statement is a valid interpretation of the Simulated case funnel report?

Step 2 of 8

Calculate Completion and Drop-Off

Calculate step completion rates, lost-session counts, and drop-off rates from adjacent milestones in the Cedar Trail simulated funnel.

Read One Funnel Transition

Choose the denominator

For one adjacent funnel transition, the current-step sessions are the earlier milestone and the next-step sessions are the milestone immediately after it. Use the current-step sessions as the denominator for both rates.

In the Cedar Trail funnel, moving from Product view to Cart addition means 1,000 current-step sessions and 400 next-step sessions. Do not use a later or earlier milestone as the denominator.

One transition, two roles

Annotated funnel transition showing Product view with 1,000 current-step sessions as the denominator and Cart addition with 400 next-step sessions as the numerator for step completion.

Simulated case: the earlier milestone supplies the denominator; the following milestone supplies the numerator for step completion.

Calculate a Complete Transition

Apply the three measures

Use these three calculations for every adjacent transition:

step completion rate (%) = next-step sessions / current-step sessions × 100

lost sessions = current-step sessions - next-step sessions

drop-off rate (%) = lost sessions / current-step sessions × 100

The two rates describe complementary outcomes for the same defined transition, so they add to 100.0%. A lost-sessions value is a count, not a percentage.

Example

Worked example

Simulated case calculation for Cart addition → Checkout start:

Current-step sessions = 400
Next-step sessions = 250

Step completion rate = 250 / 400 × 100 = 62.5%
Lost sessions = 400 - 250 = 150
Drop-off rate = 150 / 400 × 100 = 37.5%

Both percentages use 400 as their denominator: 62.5% + 37.5% = 100.0%.

Calculation summary

Simulated case: calculations for the Cart addition to Checkout start transition.

The count and the two rates answer different, compatible questions about one transition.
MeasureCalculationResult
Step completion rate250 / 400 × 10062.5%
Lost sessions400 - 250150
Drop-off rate150 / 400 × 10037.5%

Recognize an Undefined Rate

Handle a zero denominator

A rate needs a nonzero denominator. If current-step sessions are 0, neither step completion rate nor drop-off rate can be calculated: the transition rate is undefined, not 0%.

Report the zero current-step count and check the report definition or data collection before drawing conclusions. A 0% rate would claim a calculated result; here, division by zero prevents one.

Warning

Avoid a false 0%

Keep the labels precise: a zero count at the current step does not mean “zero drop-off.” It means there is no valid current-step denominator for that transition.

Practice: Complete a Transition

Find the lost-session count

Simulated case: Checkout start has 250 sessions and Order confirmation has 75 sessions. Enter the number of lost sessions for this transition.

Calculate the drop-off rate

Simulated case: Checkout start has 250 sessions and Order confirmation has 75 sessions. Enter the drop-off rate for this transition.

Step 3 of 8

Turn Funnel Losses into Investigation Questions

Use funnel losses to choose an investigation focus while keeping explanations provisional.

Read the loss in two ways

Two useful views of a loss

Simulated case — ordered funnel report

For each adjacent transition, read two different signals:

  • Lost sessions: how many sessions did not record the next milestone.
  • Drop-off rate: what share of the current-step sessions did not record the next milestone.

The largest count can point to the largest volume of unrecorded progression. The highest rate can point to the transition where progression is weakest relative to the sessions that reached it. They do not have to be the same transition.

Compare counts and rates

Simulated case: Funnel transitions and calculated losses.

Both measures use the current-step sessions for their transition.
TransitionLost sessionsDrop-off rate
Product view → Cart addition60060.0%
Cart addition → Checkout start15037.5%
Checkout start → Order confirmation17570.0%

Count versus proportion

A funnel diagram highlights the largest missing session count at the first transition and the highest proportional loss at the final transition.

The first transition has the greatest lost-session count; the final transition has the highest drop-off rate.

Turn a pattern into a question

What a loss does—and does not—show

A funnel loss is an investigation signal, not a diagnosis.

Under this funnel definition, a loss means a session did not record the next milestone in order before that session ended. It does not mean the person was permanently lost. They may return in another session, purchase elsewhere, or decide the product is not a fit.

Use the loss to choose where to look next. Do not treat it as proof of why the progression was not recorded.

Keep alternatives open

The same pattern can have different explanations:

  • Visitor intent: Some product viewers may be browsing, comparing options, or not ready to buy.
  • Usability friction: A page, control, cost, or form interaction may make progression difficult.
  • Measurement problem: A milestone may be missing, defined inconsistently, or fail to record.

These explanations can also coexist. Funnel counts alone cannot tell you which explanation caused a loss.

Example

Finding versus claim

Start with a descriptive finding, then state the uncertainty.

Descriptive finding: “In the Simulated case, 175 sessions did not progress from checkout start to order confirmation; the drop-off rate was 70.0%.”

Unsupported causal claim: “The checkout has a broken form.”

The finding accurately describes recorded progression. The causal claim adds an explanation that the funnel table has not established.

Choose a focus without overclaiming

Ask a question, not a verdict

Suppose you choose the checkout-start-to-order-confirmation transition as an investigation focus because its drop-off rate is highest.

A useful question is: What prevents some mobile checkout sessions from reaching order confirmation, if anything?

To narrow the possibilities, seek evidence that can distinguish them. For example, inspect whether order confirmation is recorded correctly, review the checkout flow for obstacles, and gather relevant observations from people attempting the flow. Each source has limits; none should be stretched beyond what it shows.

Tip

Use case details carefully

The Simulated case provides a possible clue: shipping first appears at final order review, and mobile postal-code errors say “Check your details” without identifying the field.

Those interface details make shipping disclosure and error guidance reasonable topics to investigate. They still do not prove that either detail caused the 70.0% drop-off rate. A high rate may reflect a mix of intent, friction, and measurement factors.

Phrase the next claim proportionately: “These details are plausible barriers worth checking.”

Example

Evidence should match the claim

What the case supports: “The product-view-to-cart-addition transition has the largest lost-session count, so it is a reasonable place to investigate high-volume exposure.”

What it does not support: “A brighter Add to cart button will recover those 600 sessions.”

The second statement assumes both a cause and an outcome. More evidence is needed before making either conclusion.

Check your interpretation

Interpret the two loss signals

Which statement correctly interprets the Simulated case funnel?

Step 4 of 8

Write a Neutral Usability Task

Create a neutral usability task for the Cedar Trail simulated case and define what an observer should record.

Give the task clear boundaries

Case interface

Simulated case — Cedar Trail mobile text wireframe

A shopper sees the Trail Bottle product page, priced at $30.00, with an Add to cart button. The cart shows the selected bottle and a checkout action. Checkout includes contact details and a postal-code field. At final order review, shipping is shown as $5.00 and the total is $35.00. An order confirmation appears after a completed order.

You can use this description to write and review a task; no working store is needed.

Three parts of a task

A structured usability task gives a shopper a realistic scenario while leaving their route open. It contains:

  • Goal: what the shopper wants to accomplish.
  • Starting point: the screen or situation where they begin.
  • Observable completion condition: a visible point that tells the observer when the task is complete.

The task asks the participant to act naturally. It is not a test of whether they can guess what the observer suspects.

Task structure

A simplified mobile shopping flow with visual callouts for a goal, a starting screen, and a visible finish point.

A task defines the destination and the boundaries of observation, not the path the shopper must take.

Use neutral wording and prompts

Do not steer the shopper

Use ordinary shopper language. State the goal, where to begin, and when to stop. Do not name a suspected defect, predict what the participant will feel, or tell them which button or field to use.

After the task begins, use a neutral follow-up only when useful. Good prompts invite description without suggesting an answer:

  • “What are you looking for here?”
  • “What are you thinking at this point?”
  • “Please tell me more about that.”

Avoid prompts that imply a correct reaction, such as “Was that confusing?”

Example

Compare the wording

Leading version
“Find the checkout problem and explain why the price changes.”

This tells the participant what to look for and frames the expected reaction.

Neutral version
“You are considering one Trail Bottle for your own use. Starting on the product page, proceed as you normally would to buy it. Stop when you reach the order confirmation.”

Neutral follow-up
“What are you thinking as you go through this step?”

Observe first; interpret later

Record observable events

Record what happens before explaining what it might mean. A compact observation note can contain four factual fields:

  1. Actions: pages visited, taps, changes, or returns.
  2. Errors: messages shown and the step where they appeared.
  3. Hesitation: a pause, repeated action, or stated uncertainty.
  4. Task completion: whether the participant reached the defined endpoint.

Capture the sequence and exact words or messages when possible. Keep a separate space for possible interpretations so an explanation does not get written as if it were an observed fact.

Observation note layout

A visual note template separates actions, errors, hesitation, and completion from a separate interpretation note.

First record the sequence of observable events; keep possible explanations in a separate note.

Practice: write the task

Draft a neutral task

Using the Cedar Trail mobile text-wireframe description, write a short usability task and one neutral follow-up prompt.

Your task must include a shopper goal, a clear starting point, and an observable completion condition. Do not name a suspected problem or direct the shopper to a particular control.

Write at least 160 characters (0/160).

Step 5 of 8

Separate Observations from Interpretations

Classify simulated usability-case records without turning observations or self-review into unsupported causal or frequency claims.

Name what the record actually contains

Source is not an evidence category

A source label tells you where a record came from. An evidence-category label tells you what the record says.

For this tutorial, label the supplied material Simulated case because it describes fictional Cedar Trail sessions. Then classify each record as one of these categories:

  • Behavior: an observable action, sequence, error, hesitation, or outcome. State what happened, without inventing a reason.
  • Participant opinion: what a participant says or reports feeling. It is useful perspective, not proof of a cause.
  • Analyst assumption: an explanation added by the analyst. It can be plausible, but it remains unverified.

Keep these two dimensions separate: Simulated case is the source; Behavior, Participant opinion, and Analyst assumption are evidence categories.

Separate provenance from interpretation

A source card branches to three visually distinct evidence-category cards: observed action, spoken opinion, and an inferred explanation.

Use “Simulated case” as the source label, then apply one evidence-category label to the record.

Classify wording, not plausibility

Read the case records precisely

The wording determines the category. Notice how the Behavior records describe sequences, while the Analyst assumption adds a cause that was not directly observed.

Case-record categories

All records below have the source label Simulated case.

A category describes the claim in the record, not whether the source is simulated.
RecordEvidence categoryWhat it supports
Three of four participants returned to the cart after the $5.00 shipping charge first appeared at final order review.BehaviorA visible sequence; not a stated motive.
A participant says, “I could not tell which detail needed fixing.”Participant opinionA reported perception; not proof of a design cause.
The shipping charge caused participants to return to the cart.Analyst assumptionA possible explanation that needs checking.
Two participants encountered the generic postal-code message, “Check your details.”BehaviorAn observed error state; not evidence that it caused stopping.

State the limits of the case

Use self-review for clues, not proof

A text-wireframe walkthrough or your own self-review can reveal possible issues: unclear copy, missing information, an awkward sequence, or an error message that does not identify a field. Treat those discoveries as questions to investigate.

It does not reproduce another person’s attention, device conditions, prior expectations, typing behavior, or interpretation. A walkthrough is therefore not a substitute for observing real interaction on a functioning interface.

Warning

Do not turn four sessions into a population claim

Simulated case contains four fictional mobile sessions. The behavior groups overlap: three participants returned to the cart after shipping appeared, two encountered the generic postal-code error, and one stopped without order confirmation while the other three reached confirmation.

These records can point to plausible barriers. They cannot establish how frequent a behavior would be among real Cedar Trail sessions, and a sequence of events does not establish that one event caused the next.

Apply the evidence labels

Classify records and state the limits

For each record, write its evidence category: Behavior, Participant opinion, or Analyst assumption.

  1. “A participant tapped Continue, saw ‘Check your details,’ and entered a different postal code.”
  2. “A participant says, ‘I could not tell what to fix.’”
  3. “The generic message made the participant abandon checkout.”

Then explain why four fictional sessions and your own text-wireframe walkthrough cannot establish real-world frequency or causality.

Write at least 250 characters (0/250).

Step 6 of 8

Build an Evidence-Linked Backlog Record

Connect a plausible conversion barrier to evidence, a calibrated confidence rating, and a next check that reduces uncertainty.

Build a traceable record

From evidence to a backlog record

An improvement backlog turns a clue into a decision-ready record. It does not declare a cause or promise more purchases.

Start with a specific plausible Barrier. Link it to a Proposed change, then state the Evidence and a Confidence rating. Add a Next check that can reduce a named uncertainty.

Use the full field names: Barrier, Proposed change, Evidence, Confidence, Likely impact, Effort, Next check, and Priority. This step focuses on the evidence-linked portion of a backlog record; impact, effort, and final priority come next.

Evidence-to-change chain

A visual chain from funnel and mobile-session evidence through a plausible barrier to a proposed interface change and a follow-up check.

A backlog record makes each link visible: evidence supports a plausible barrier, which motivates a proposed change and a focused next check.

Tip

Be specific and provisional

Write the barrier as a condition that may hinder progress, not as a claim about a visitor's hidden motive. “A generic postal-code error may make correction difficult on mobile” is more useful than “Visitors dislike checkout.”

Keep sources and categories distinct

Use sources without merging their scope

Keep the scope of each source visible. The funnel is an ordered session report; the usability case contains four fictional mobile sessions. They can point toward the same investigation, but they are not one combined sample.

The behavior groups in the case overlap. Do not add three of four and two of four together as though they describe separate participants.

Relevant case records

Both sources are labeled Simulated case. The evidence-category label describes what the record is; the source label describes where it came from.

The records suggest places to investigate, not a proven cause of the funnel loss.
SourceEvidence categoryRecordScope
Simulated caseBehavior250 sessions started checkout; 75 reached order confirmationOrdered eligible sessions in the seven-day report
Simulated caseBehaviorthree of four participants returned to the cart after shipping appeared at reviewFour fictional mobile sessions
Simulated caseBehaviortwo participants encountered a generic postal-code errorFour fictional mobile sessions
Simulated caseBehaviorone of those two stopped without confirmationFour fictional mobile sessions

Tip

Preserve the evidence category

A Participant opinion can report a perception, while an Analyst assumption proposes an explanation still needing support. Neither should be rewritten as observed Behavior.

Set confidence and name the uncertainty

Rate confidence in the barrier, not the outcome

Confidence answers a narrow question: how plausible and relevant is this barrier, given the available evidence? It is not certainty that the Proposed change will increase purchases.

Use Low when the claim is mainly an unsupported Analyst assumption. Use Medium when limited, relevant evidence supports it but important alternatives remain. Use High only when relevant evidence converges and fewer alternatives remain unresolved.

For example, the shipping pattern has Medium confidence: three of four mobile participants returned to the cart after the $5.00 shipping charge appeared at final order review, but this small simulated case cannot establish why each participant returned.

Example

Choose a check that closes a gap

A useful Next check names the uncertainty it addresses.

  • Unsure whether the mobile error state is rendered as described? Verify the interface state.
  • Unsure whether people can identify and correct the field? Observe a neutral mobile purchase task and record actions, errors, hesitation, and completion.
  • Unsure whether the funnel transition is comparable? Check the measurement definition.

“Launch the change and expect more orders” is not a next check; it leaves the uncertainty unnamed.

Warning

Do not let a plausible idea become evidence

A brighter Add to cart button is not automatically a strong backlog record. Without relevant Behavior or Participant opinion, “it will get more clicks” is an Analyst assumption and begins with Low confidence. A next check should seek relevant evidence rather than predict an uplift.

Practice: link a barrier to evidence

Draft a backlog record

Draft the evidence-linked portion of one backlog record for the mobile postal-code experience. Include all five requested fields. Use only the Simulated case records from the previous screen, keep the Barrier provisional, and make the Next check address a real uncertainty rather than promise an uplift.

Write at least 250 characters (0/250).

Step 7 of 8

Rank Improvements by Impact and Effort

Make an evidence-led ranking of improvement candidates while keeping likely impact, effort, and uncertainty distinct.

Estimate likely impact carefully

Judge likely impact without forecasting

Likely impact is a judgment about what could matter if a plausible barrier is real. Consider three questions:

  • Reach: How many sessions could encounter it?
  • Severity: How seriously could it interrupt progress?
  • Completion relevance: How close is it to a successful order?

The Simulated case has 250 Checkout start sessions and 75 Order confirmation sessions. That makes the final checkout transition an important investigation area. But 175 sessions without recorded progression are not automatically 175 recoverable orders, and they do not identify a cause.

For example, three of four mobile participants returned to the cart after the $5.00 shipping charge appeared at final review. That makes earlier shipping disclosure a plausible candidate with possible reach. It does not establish why they returned or prove that disclosure would change their outcome.

Reach is not recoverability

A funnel and three lenses illustrate that likely impact considers reach, obstacle severity, and closeness to completion, while the cause of a loss remains uncertain.

Simulated case funnel volume can indicate possible exposure. It cannot show that every lost session was caused by one barrier or can be recovered.

Tip

Keep exposure separate from uplift

Do not let a large count become an outcome promise. The 600 sessions that did not progress from Product view to Cart addition show possible exposure at that transition; they do not show that a brighter Add to cart button would recover 600 sessions.

Compare trade-offs without a score

Keep judgments visible

Compare backlog records side by side, but do not hide judgment inside a precise-looking composite score. Use the existing fields—Barrier, Proposed change, Evidence, Confidence, Likely impact, Effort, Next check, and Priority—and explain the trade-off.

Both evidence-supported candidates have Medium Confidence in this Simulated case. The shipping pattern is relevant but does not establish why participants returned. The postal-code pattern shows a real generic error state, but the small case cannot establish how often it prevents completion.

Likely impact can reasonably be judged differently. A team may emphasize shipping's broader possible checkout reach and Small effort, or the error guidance's direct relevance to correcting a blocking-looking state and its closer link to one observed noncompletion. State which consideration you give more weight.

Candidate comparison inputs

These are comparison inputs, not a hidden scoring formula. Behavior groups overlap, so do not add their participant counts together.

Simulated case: evidence and relative effort for three backlog records
Proposed changeEvidenceImpact considerationEffort
Earlier shipping disclosureBehavior: three of four returned to the cart after the $5.00 shipping charge appeared at final order review.Possible exposure across checkout; repeated returns, but the reason for returning is unknown.Small
Field-specific error guidanceBehavior: two of four encountered the generic postal-code error; one of those two stopped without Order confirmation.Potentially narrower reach; directly relevant to correction and close to successful completion.Medium
Brighter Add to cart button1,000 Product view sessions and 400 Cart addition sessions, but no Behavior record identifies the button as a barrier.Possible high-volume exposure only; no case evidence connects the button to the loss.Small

Tip

Act, investigate, or verify

A worthwhile investigation is not automatically a worthwhile improvement. Keep the brighter button behind candidates with relevant evidence until a Next check investigates whether button visibility is actually a barrier.

Conversely, if direct inspection verifies a functional or accessibility problem, fixing it can deserve priority on its own merits even when conversion uplift is unknown. The Simulated case's generic message is a clue to inspect; do not label it a verified accessibility failure without checking the interaction.

Make the decision rule explicit

Choose a defensible decision principle

Which decision principle best supports a transparent ranking of the three Simulated case candidates?

Defend a ranked priority

Rank the improvement backlog

Rank all three Simulated case candidates with Priority 1, 2, and 3, where 1 is highest:

  • Earlier shipping disclosure
  • Field-specific error guidance
  • Brighter Add to cart button

Give one concise rationale for your ranking. Name the main trade-off between the two evidence-supported candidates, using evidence, likely impact, and effort. Also explain why the brighter button does not earn an early priority merely because its Effort is Small. End with one specific Next check for your Priority 1 candidate.

Write at least 220 characters (0/220).

Step 8 of 8

Finalize and Defend Your Priorities

Use the workflow, not a shortcut

From evidence to a decision

You now have the full decision path: define a consistent funnel, locate a loss, examine structured observations, describe a plausible barrier, and prioritize the next action.

The result is an improvement backlog—not a verdict. It helps a team decide what to address or investigate first. A funnel loss can point to a stage worth investigating, while a behavior record can make one explanation more plausible. Neither source alone proves a cause or guarantees more purchases.

Your final rationale should separate:

  • what the Simulated case shows,
  • what remains an assumption, and
  • what to check next.

The evidence-to-priority workflow

A diagram shows funnel stages flowing into observation notes and then into three prioritized backlog cards.

Use the funnel and observations as inputs to a decision, while keeping the decision provisional.

Read the case before you rank it

Simulated case: funnel recap

Use this recap as your source material. The report is a closed, ordered funnel: eligible sessions started during the same seven-day period, completed milestones in order within one session, and counted at most once per milestone.

For the final transition, use:
step completion rate (%) = next-step sessions / current-step sessions × 100

The current step is Checkout start; the next step is Order confirmation. Show the resulting percentage to one decimal place in the evidence for your top-ranked backlog record.

Adjacent funnel transitions

Simulated case — funnel report

Counts are sessions in the consistently defined seven-day funnel.
TransitionCurrent-step sessionsNext-step sessions
Product view → Cart addition1,000400
Cart addition → Checkout start400250
Checkout start → Order confirmation25075

Evidence and effort recap

Simulated case — candidate changes, observation records, and relative implementation estimates. The behavior groups overlap: they do not describe separate groups of participants.

Treat an observation as a clue about a barrier, not proof that the proposed change will work.
Proposed changeCase clueEffort
Earlier shipping disclosureBehavior: three of four participants return to the cart after the $5.00 shipping charge appears at final order review.Small
Field-specific error guidanceBehavior: two of four encounter the generic postal-code error; one of those two stops without Order confirmation.Medium
Brighter Add to cart buttonThe funnel shows fewer Cart additions than Product views, but the case contains no behavior record about the button.Small

Build and defend the backlog

Final application: defend three priorities

Create a concise improvement backlog from the Simulated case recap.

Write exactly three backlog records: one for each proposed change. Use these field names in every record: Barrier, Proposed change, Evidence, Confidence, Likely impact, Effort, Next check, and Priority.

Requirements:

  • Assign priorities 1, 2, and 3, where 1 is highest.
  • In the Priority 1 record, show the Checkout start → Order confirmation step completion calculation to one decimal place and cite a relevant Behavior record.
  • Use Low, Medium, or High for Confidence and Likely impact; use Small, Medium, or Large for Effort.
  • Identify evidence categories where relevant and label the source as Simulated case.
  • State one limitation after the records. Do not say the evidence proves a cause or guarantees more purchases.

Your Next check should target a specific unresolved uncertainty, not promise a test result.

Write at least 700 characters (0/700).

Carry the decision forward

Summary

Reusable evidence-to-priority checklist

  • A consistently defined funnel identifies where recorded progression falls, not why it falls.
  • Behavior records can strengthen a plausible barrier explanation, but a small Simulated case does not establish prevalence or causation.
  • Each backlog record makes its evidence, confidence, likely impact, effort, uncertainty, and next check visible.
  • A priority rank guides the next action; it is not proof that a change will increase purchases.
  • The next tutorial examines how to evaluate proposed changes through a structured comparison.

Step complete

Congratulations! You completed: Prioritize Conversion Improvements from Evidence

You finalized a ranked improvement backlog by connecting funnel evidence, usability observations, uncertainty, and next checks.

Download the App now to have access to + 5000 free courses, exercises, certificates and lots of content without paying anything!

  • 100% free online courses from start to finish

    Thousands of online courses in video, ebooks and audiobooks.

  • More than 60 thousand free exercises

    To test your knowledge during online courses

  • Valid free Digital Certificate with QR Code

    Generated directly from your cell phone's photo gallery and sent to your email

Cursa app on the ebook screen, the video course screen and the course exercises screen, plus the course completion certificate