Resurfacing Old A/B Tests: A System for Faster Iteration Cycles

Running A/B tests often takes time, yet many teams fail to review their past experiments for insights. Studies show that more than 90% of tests do not produce successful results but still contain valuable data.

By revisiting old A/B tests, you can reduce iteration cycles and enhance conversion rates over time. Discover how a structured approach transforms past efforts into practical lessons.

Key Takeaways

• Resurfacing old A/B tests improves iteration speed by analyzing past data to identify patterns, reducing redundancy, and improving decision-making. Studies show one winner emerges per ten tests among over 28,000 experiments analyzed.
• Breaking down results by device type, traffic source, or user behavior uncovers hidden insights. For example, variants may perform differently across mobile and desktop users—Shopify expert Shanelle Mullins highlights that segmentation can reveal missed opportunities.
• AI-powered tools like GrowthLayer improve multivariate testing by running simultaneous micro-experiments and analyzing entire customer journeys instead of isolated variables, reducing regret costs from underperforming variants by up to 50%.
• Losing tests provide valuable insights into flaws or unexplored segment-specific wins. Teams can refine hypotheses using behavioral data such as bounce rates or qualitative feedback via tools like Hotjar to minimize wasted efforts.
• Centralized test repositories enhance institutional memory for teams managing 50+ tests annually. Systems like GrowthHackers ensure organized documentation (e.g., naming conventions) avoids repeated errors while facilitating faster hypothesis revisions.

Additional Insight: For teams running 50+ experiments per year, establishing a test repository with structured hypothesis logging, standardized metadata, and archive hygiene is critical. GrowthLayer is an experimentation knowledge system built for teams running 50+ A/B tests per year that addresses institutional knowledge decay by centralizing learnings in a searchable, reusable system.

Why Resurfacing Old A/B Tests Matters for Faster Iteration

Resurfacing old A/B tests accelerates iteration by using past data to refine decisions. With over 28,304 experiments showing a typical success rate of one winner per ten tests, revisiting prior results can identify patterns that improve future outcomes.

Rather than starting from scratch, teams can revisit losing tests to understand failed hypotheses or confirm statistically significant trends that impact conversion rates. For growth teams running 50+ experiments annually, using historical insights reduces test redundancy and makes better use of resources.

Using findings from older experiments strengthens hypothesis testing frameworks for improved outcomes. For example, a test with an unexpected lift in click-through rates may reveal previously overlooked metrics like user behavior shifts or scrolling depth.

These insights allow faster iterations without compromising internal validity or wasting budget on inconclusive trials. As Jeff Bezos suggests, even small risks with high potential payoffs are invaluable when supported by actionable data drawn directly from customer interactions rather than theoretical models alone.

The 7-Step System for Learning From Old A/B Tests

Review previous A/B tests to identify valuable insights that can improve your strategies. Apply an organized approach to analyze data, gather learnings, and enhance future experiments efficiently.

Step #1: Validate the Accuracy and Significance of Your Data

Conduct A/A tests to ensure the reliability of your testing tools. If an A/A test indicates a variance exceeding 0.5%, ensure that future experiments target minimum lifts of at least 1% to maintain credibility.

Ensure every variant gathers at least 300 conversions since smaller sample sizes, such as 50, cannot provide dependable insights.

Observe traffic patterns throughout the experiment period. For example, an unexpected surge or drop in visits can impact results, such as the instance where an unrelated name similarity led to an 800% traffic increase and distorted outcomes.

Restart any test affected by irregularities only after conditions normalize to avoid skewing data.

Establish clear thresholds within platforms like GrowthLayer or Convert Experiences before initiating tests. Define parameters for confidence levels, conversion counts, and runtime durations aligned with your sales cycle; this ensures every result corresponds with user behaviors over traditional two-to-four-week periods without compromising statistical validity or precision.

• Key Thresholds:
• Minimum 300 conversions per variant.
• Variance threshold of 0.5% for A/A tests.
• Target minimum lift of at least 1% for experiments.

Step #2: Analyze Micro, Macro, and Guardrail Metrics

Break test results into micro, macro, and guardrail metrics to extract actionable insights. Micro metrics track specific goals, like click-through rates or signups. Macro metrics measure broader business impact, such as overall revenue or customer lifetime value (LTV).

For instance, a lead pop-up may boost lead generation but might also increase fake email entries. Guardrail metrics like bounce rates catch these negative effects early and prevent scaling flawed tests.

Set thresholds for each metric to guide decisions during rollout. If guardrails dip below acceptable levels, halt implementation immediately to avoid long-term harm. Ben Labay emphasizes integrating analytics systems that provide instant visibility into these numbers.

Include tools like GrowthLayer in your workflow for quicker access to key performance indicators across segments and iterations.

• Metric Categories Summary:
• Micro metrics: click-through rates, signups.
• Macro metrics: revenue, customer lifetime value.
• Guardrail metrics: bounce rates, negative user interactions.

Step #3: Segment Results for Deeper Insights

Segment test results by device type, traffic source, and user behavior to identify hidden performance patterns. For instance, a variant may perform better than the control on desktop but perform worse on mobile.

Shanelle Mullins from Shopify points out that segmentation can expose critical errors or missed opportunities specific to certain groups. Focus on high-traffic segments with at least 300 conversions per channel for meaningful insights.

Consider factors like returning visitors, bounce rate, and pageviews during analysis. Traffic sources often behave differently; organic users may convert better than paid channels for one version of a landing page.

Even low-traffic tests can benefit from identifying problem areas within smaller groups rather than depending solely on overall outcomes across all users.

• Segmentation Criteria:
• Device type (mobile vs. desktop).
• Traffic source (organic vs. paid).
• User behavior (returning visitors, bounce rate, pageviews).

Step #4: Evaluate User Behavior Through Interaction Data

Understanding segmented results prepares teams to discover the reasons behind user actions. Evaluate interaction data like scroll depth, click mapping, and session recordings using tools such as Hotjar.

High bounce rates may indicate UX issues or irrelevant targeting, while low engagement metrics can point to unclear CTAs. Track behavior patterns across both test variants to measure if users reach key goals, such as form submissions or add-to-cart actions.

Exit-intent polls and post-conversion surveys provide qualitative insights into barriers or motivations driving user decisions. For example, a high abandonment rate on pricing pages could be tested against feedback suggesting confusion about costs.

Use these inputs to refine hypotheses linked directly with main KPIs like conversion rate uplift or average order value improvement. Properly analyzing this interaction data ensures decision-making is guided by behaviors rather than assumptions.

• Interaction Data Points:
• Scroll depth.
• Click mapping.
• Session recordings.
• User feedback through surveys and exit-intent polls.

Step #5: Extract Learnings From Losing Tests

Losing tests are not wasted efforts; they provide essential data for future iteration. Review these tests to determine whether the hypothesis was flawed, misaligned with user behavior, or poorly executed.

A flat or negative result may also highlight hidden winning opportunities in specific user segments. For example, a headline test may fall short overall but show positive results among new users or high-value customers.

Use segmentation tools and statistical methods like regression analysis to identify these patterns.

Document actionable insights by asking key questions: Did the losing variant reveal friction points? Could interaction data uncover missed behavioral cues? Organize all learnings into a centralized repository such as GrowthLayer for quick reference during planning sessions.

This step ensures misplaced confidence doesn't lead to repeating ineffective strategies while allowing teams to focus on high-impact changes instead of discarding failed concepts entirely.

Step #6: Optimize and Scale Winning Variants

Roll out high-performing variants by re-testing them against the original control to confirm their impact. For example, apply successful changes from a primary landing page to similar pages across your e-commerce site to increase conversion rates and ROI.

This approach maximizes returns while also helping to identify broader trends in user behavior that can influence future hypotheses.

Iterate on winning tests for continuous improvement rather than settling on one version. A streaming platform conducted six additional iterations of a single test over 15 months, producing consistent uplifts each time.

Report detailed metrics like click-through rate improvements (+3.2%) alongside overall gains such as average order value increases (+2.2%) to quantify success effectively.

Step #7: Build a Test Learning Repository

Centralize all A/B test documentation for easy access and future use. Create a system to store hypotheses, goals, screenshots, data summaries, and outcomes. Use tools like GrowthHackers experiments or GrowthLayer (growthlayer.app) to maintain organized repositories for teams conducting 50+ tests annually.

Apply consistent naming conventions such as “CTR lift test – XYZ client – Mar 1–14, 2021” to quickly identify historical tests.

Categorize entries by wins or losses using metadata like impact size or qualitative insights from user behavior. Maintain strict archival hygiene by periodically reviewing the repository and removing outdated information.

This structure allows for faster retrieval of learnings during hypothesis revision cycles in continuous experimentation workflows. Shift focus to analyzing behavioral insights for a deeper understanding of users' actions across experiments.

• Repository Standards:
• Structured hypothesis logging.
• Standardized metadata schema (feature area, funnel stage, metric type, traffic source, result type).
• Win/loss categorization and impact scoring.
• Searchable qualitative learnings and version history.
• Tag normalization and archive hygiene.

Leveraging Behavioral Insights in Resurfaced Tests

Tracking user actions uncovers essential interaction patterns that affect conversion rates. Examining this behavior aids in prioritizing changes for observable growth.

Track User Scroll Depth and Engagement

Use scroll depth tracking tools like Hotjar to determine how far users progress through your content. This data shows whether critical elements remain unseen, especially above-the-fold areas that influence conversion rates.

Session recordings can then identify points of drop-off or declining interest, offering insights into user challenges and engagement gaps. For example, consistent abandonment halfway down a page might indicate unclear calls-to-action or unengaging copy.

Scroll maps help teams see where attention clusters occur and identify if high-value information is placed in low-engagement zones. Use this insight to restructure landing pages for better results by positioning key details higher up.

If analytics reveal heavy interactions on specific sections but poor overall conversions, revisit those pieces to improve their clarity or relevance. Patterns tracked over time highlight recurring issues for adjustments in ongoing experimentation cycles without relying on random fixes.

Identify Problematic Click Patterns

Tracking user scroll depth provides valuable insights, but click patterns offer a sharper lens into usability issues. Heatmaps highlight areas of high or low click activity, pointing to design challenges.

For instance, clustered clicks on non-interactive elements signal confusion in layout. On the other hand, sparse clicks on key CTAs may indicate weak messaging or poor visibility.

Session recordings help identify missed opportunities like overlooked buttons or unclear layouts. High concentrations of clicks on unimportant elements may distract users from your primary goals, reducing conversion rates.

Reviewing the time-to-click for major actions also pinpoints friction points needing redesigns. Tools like GrowthLayer simplify such analysis efficiently for teams managing 50+ tests at scale.

Use Surveys to Collect Qualitative Feedback

Post-conversion surveys reveal why users take specific actions. Use open-ended questions to understand their motivations and identify unseen opportunities. For example, asking “What made you decide to complete your purchase?” can highlight unique reasons behind conversions.

Exit intent polls capture reasons for abandonment, such as unclear pricing or missing features, which analytics alone cannot uncover.

Group responses by user type to gain deeper insights into diverse customer needs. Immediate feedback during the experiment period helps address issues while tests run, ensuring actionable data collection without delays.

Apply these findings to new hypotheses or optimization strategies to improve conversion rate optimization (cro) efforts effectively.

When and How to Revise Old Hypotheses

Reevaluate old hypotheses when user behavior differs from initial predictions. Modify testing approaches to match updated audience insights or changes in data trends.

Misaligned Test Execution

Misaligned test execution happens when testers create experiments that do not accurately reflect real user behavior. For instance, a shorter landing page might increase conversion rates by 16%, while a longer version causes a 27% drop.

This difference often stems from unexpected user flows or technical issues during the experiment. To detect such problems, review session recordings and heatmaps on a regular basis.

These tools provide insights into gaps between user expectations and their interactions with test variations.

For multi-change tests that fail, remove one change at a time and assess subtle signals throughout the funnel for clearer insights. Operational mistakes like traffic irregularities or broken CTAs can also render results invalid without pointing to any flaws in your hypothesis or design.

Combining qualitative analysis with practical methods helps identify execution errors early before advancing problematic changes further into development.

Adjusting hypotheses with unforeseen results requires clear follow-ups to pinpoint impactful elements accurately.

Correct Hypotheses With Unexpected Outcomes

Unexpected outcomes often arise from valid hypotheses during A/B testing. For example, a 6-field registration form performed better than a shorter 3-field version because the longer form displayed the privacy policy in full view.

This improved confidence and increased registrations by over 20% after iterations. Such results suggest user behavior may not align with expectations, especially when visibility or usability issues influence decision-making.

Examining unexpected outcomes involves concentrating on specific data points like interaction patterns and survey feedback to identify friction areas. In another case, dark mode initially reduced conversion rates for a streaming platform but led to a 17% increase after three test cycles resolved visual hierarchy issues.

Revisiting behavioral insights allows teams to adjust hypotheses instead of discarding promising ideas prematurely.

Implementing Always-On Experimentation for Faster Cycles

Continuous experimentation minimizes obstacles and keeps testing synchronized with real-time user behavior. This method fosters progress by incorporating data collection into daily workflows for quicker insights.

Transitioning From One-Off Tests to Continuous Testing

Switching to continuous testing shifts focus from isolated experiments to a consistent flow of data-driven insights. Growth teams can review current tests within the first 30 days, ensuring clean data and removing “zombie tests” that drain resources.

By day 60, firms should have governance structures in place and launch their first workflow-focused experiment. This approach improves test accuracy while incorporating customer feedback into ongoing optimization cycles.

Modern tools like GrowthLayer automate workflows for faster iteration across entire customer journeys. SaaS companies apply this method to improve retention by addressing onboarding friction points through micro-interactions tested continuously.

E-commerce operators with high volumes gain agility in cart optimization and return reduction strategies using multi-armed bandit frameworks instead of static A/B setups. Always-on experimentation increases success rates by analyzing how changes interact with broader user behavior over time rather than focusing on singular outcomes alone.

Testing Entire Customer Journeys Instead of Isolated Elements

Testing customer journeys as a whole offers deeper insights into user behavior. Rather than focusing on isolated elements, optimize complete paths such as onboarding or win-back flows.

For instance, compare 7-day/3-message email setups with 14-day/6-message combinations to measure open rates, clicks, and conversions across the full cycle. Use unified metrics such as conversion rate (CR), average order value (AOV), and lifetime value (LTV) for evaluating performance instead of assessing each individual step.

High-volume workflows like abandoned cart or browse abandonment campaigns gain the most from journey testing. Experiment with segmentation strategies by pairing email triggers with SMS or push notifications to determine which sequences drive quicker decisions without negatively impacting margins.

Depend on real-time data from CRM systems instead of cookies alone while ensuring attribution windows remain consistent over 7–30 days.

Automating Experimentation With Modern Marketing Tools

Integrating modern marketing tools into experimentation allows teams to run multiple micro-experiments simultaneously. Platforms using real-time behavioral data, like GrowthLayer, support dynamic segmentation and traffic routing for more detailed testing.

For example, Champion/Challenger models automatically test variations against each other and optimize outcomes without manual intervention.

Setting clear operational rules and KPIs ensures automation aligns with organizational goals. Assigning a dedicated program owner helps manage backlogs, oversee implementation, and track results effectively.

By 2026, advanced systems will create feedback loops for automatic hypothesis testing while connecting sales data with marketing outcomes to improve decision-making accuracy.

Comparing Traditional vs. AI-Based Experimentation

Traditional and AI-based experimentation approaches differ significantly in speed, scalability, and insights. For growth teams and CRO practitioners managing 50+ tests, understanding these distinctions is essential for optimizing workflows and outcomes.

Aspect

Traditional A/B Testing

AI-Based Experimentation

Setup

Manual and time-consuming. Requires significant effort from marketers and data teams to define variables, goals, and configurations.

Automated and continuous. AI tools configure tests dynamically, reducing manual intervention.

Duration

Fixed testing periods. Results depend on predefined metrics collected over weeks or months.

Ongoing testing. Real-time insights adapt to changing user behavior without fixed end dates.

Focus

Single-variable optimization. Measures isolated changes (e.g., button color, headline) while ignoring broader user journeys.

Multi-variable optimization. Evaluates entire customer flows, including cross-channel behaviors and long-term value (LTV).

Error Costs

High. Small sample sizes and human error can skew results or lead to false positives.

Low. Algorithms reduce error margins by analyzing larger datasets and adjusting parameters automatically.

Insights

Retrospective. Focuses on what worked in the past rather than predicting future outcomes.

Predictive. Models forecast future performance based on behavioral patterns and trends.

Scalability

Limited. Each test is resource-intensive, making it challenging to run multiple experiments simultaneously.

High. AI tools scale effortlessly, supporting dozens of simultaneous tests across multiple journeys.

Metrics

Basic. Optimizes for predefined KPIs like click-through rates and conversions.

Advanced. Prioritizes metrics like ROI, LTV, and cross-device interactions.

Decision Speed

Slow. Teams wait until tests conclude to analyze results and make decisions.

Fast. AI delivers actionable insights continuously for quicker iterations.

Example Tools

Optimizely, Google Optimize (legacy platforms).

GrowthLayer, Adobe Target (AI-driven platforms).

Growth teams operating in dynamic markets gain value from AI systems. Unlike traditional methods, AI allows continuous iteration, optimizing entire customer journeys while reducing errors and manual workload. Modern platforms like GrowthLayer function as real-time testing labs.

Avoiding Common Pitfalls in Iterative Testing

Incomplete documentation can hinder future learning from past experiments. Concentrate on combining cross-functional data to gather more comprehensive insights.

Lack of Institutional Memory and Documentation

Teams conducting over 50 A/B tests annually often encounter repeated errors due to the lack of structured documentation. Without a centralized test repository, valuable insights are lost as team members leave or transition to different roles.

This results in repetitive testing, wasted resources, and slower iteration cycles. GrowthLayer (growthlayer.app) addresses this issue by preserving institutional knowledge with well-organized archives that capture hypotheses, goals, screenshots, data visuals, and qualitative insights.

Test libraries should use clear naming conventions such as “CTR lift test – XYZ client – Mar 1–14, 2021” for easy retrieval. Including metadata like win/loss classification and impact level ensures actionable insights are easily accessible over time.

• Documentation Best Practices:
• Structured hypothesis logging.
• Consistent metadata tagging.
• Win/loss categorization and impact scoring.
• Searchable archives with version history.

Overlooking System-Wide Testing Opportunities

Neglecting comprehensive system testing often results in missed opportunities to improve the entire customer journey. Testing isolated elements, such as a single call-to-action button, limits the depth of insights and overlooks broader impacts on key business metrics like lifetime value (LTV) or average order value (AOV).

Instead, prioritize workflow testing and path-based optimization to assess how variations perform across acquisition, activation, retention, and conversion stages. For example, comparing annual discounts versus premium trials for expiring subscriptions can reveal both churn reduction and long-term revenue growth.

System-level experimentation involves assessing test path variants as complete packages rather than isolated components. Evaluate metrics like time-to-purchase along with key performance indicators such as conversion rate or margin impact.

A weighted pipeline approach ensures each variation aligns with business objectives instead of focusing solely on optimizing individual actions. Automating these processes through modern tools like GrowthLayer reduces manual effort while maintaining data accuracy across subsystems.

Teams conducting over 50 tests annually can accelerate learnings by integrating cross-functional data into each iteration cycle without unnecessarily extending test durations.

Failing to Integrate Cross-Functional Data

Overlooking system-wide opportunities limits the ability to scale tests effectively. Without integrating cross-functional data, teams risk inconsistent results and missed insights.

For example, relying solely on cookies for attribution ignores critical CRM or e-commerce platform data tied to user identification methods like email or phone numbers. This gap can distort metrics like conversion rate and customer lifetime value (LTV), leading to flawed conclusions.

Define operational rules that support alignment across teams handling marketing automation, sales pipelines, and product analytics. Use segmentation strategies that include cart value, device type, source channel, and seasonal trends for better accuracy in A/B testing outcomes.

Assign a dedicated program owner who can manage backlogs and foster collaboration between functions while maintaining consistent KPIs within pre-set attribution windows ranging from 7 to 30 days.

This approach ensures actionable insights without unnecessary noise in your experimental results.

Operationalizing a System for Continuous Testing

Create dedicated teams focused on testing, making sure they have clear responsibility and ownership. Rank experiments based on their potential impact using data-informed frameworks like ICE or RICE to speed up decision-making.

Structuring Teams With Growth Squads

Growth squads bring together marketers, data analysts, and operators into a unified team. Marketers propose ideas based on user experience (ux) or conversion rate optimization (cro) goals. Data analysts monitor test metrics such as uplift and LTV to ensure result accuracy.

Operators implement automation workflows using microservices for consistent experiment execution. Assigning a program owner to manage backlogs and reporting ensures accountability. Regular growth meetings encourage collaboration while experiment cards record findings.

This structure creates feedback loops to share knowledge across teams, aiding in continuous experimentation and faster iteration cycles. Integrating operational clarity enhances decision quality and repetition prevention.

Prioritizing Tests Using ICE and RICE Methods

Using the ICE and RICE methods simplifies test prioritization during high-volume experimentation. The ICE framework scores tests based on Impact, Confidence, and Ease, making it ideal for teams with limited time or resources.

A test showing potential to increase conversion rates by 15% but requiring less than one week of implementation ranks higher due to favorable scores across all three factors.

The RICE method adds Reach as a criterion, which helps when determining which tests may maximize user engagement or revenue growth. Testing a checkout feature affecting 80% of site traffic may deliver more significant results compared to niche optimizations targeting just 10% of users.

Align experiments with business goals like increased LTV by consistently reviewing and updating priority scores based on emerging data insights.

Measuring KPIs Like Uplift, Conversion, and LTV

Track uplift by measuring the incremental gain tied directly to A/B tests, assessing changes against a control group over specified attribution windows such as 7 or 30 days.

Use weighted pipelines that consider both action counts and alignment with broader business objectives such as customer lifetime value (LTV) or profit margins.

Segment KPIs by factors like cart size, traffic source, and user behavior for improved precision. In e-commerce settings, also monitor indicators such as product returns and customer complaints to capture downstream effects.

Ensure conversion tracking supports sustained LTV growth while reducing experimentation costs through continuous data collection and analysis.

Documentation Framework for A/B Testing: Enhancing Institutional Memory and Consistency

Centralize all test documentation using platforms like GrowthHackers or GrowthLayer. Each record should include the hypothesis, objectives, screenshots, and data visuals to ensure clarity.

Apply consistent naming conventions while maintaining a structured system that allows quick retrieval. Log critical metadata such as features, metrics, funnel stages, sources, and outcomes.

Categorizing results by win or loss status alongside their impact on revenue or retention prevents missed insights.

Standardize tagging across archives to reduce duplicate errors and improve searchability over time. Review stored tests periodically to identify performance patterns by funnel stage or user behavior trends.

Highlight iteration chains that indicate how previous experiments influenced new ones for continuous improvement cycles. Maintaining archive organization ensures learnings persist without cluttering the repository with outdated files or unstructured information.

Conclusion

Revisiting old A/B tests creates a basis for smarter, faster decision-making. Each test, whether it succeeds or fails, contains crucial insights into user behavior. By analyzing past data and iterating quickly, teams can improve strategies that lead to measurable growth.

Begin developing systems to gather insights and expand effective ideas efficiently. Testing isn't solely about optimization; it's about ongoing learning that drives better customer experiences over time.

FAQs

1. What is A/B testing and why is it important?

A/B testing is a method of comparing two versions of something, like a webpage or app, to see which performs better. It helps improve user experience (ux) and supports data-driven decision making by analyzing conversion rates and user behavior.

2. How can resurfacing old A/B tests speed up iteration cycles?

Resurfacing old A/B tests allows businesses to reuse past insights for faster hypothesis testing. This reduces test duration while leveraging previous data collection efforts for continuous experimentation.

3. How do you handle losing tests in A/B experiments?

Losing tests provide valuable data insights into what doesn't work. By analyzing customer feedback and conducting statistical analysis, researchers can refine their approach for future experiments using tools like Bayesian analysis or outlier detection.

4. What role does sample size play in successful A/B testing?

Sample size ensures accurate results by providing enough data points for reliable statistical inference. Larger samples reduce the risk of errors during sensitivity checks or when applying methods like bootstrapping or Mann–Whitney tests.

5. Can recommendation systems benefit from findings in old A/B tests?

Yes, findings from past randomized experiments can help optimize recommendation algorithms used in digital marketing or recommender systems by improving explainability and learning to rank models.

6. How do advanced techniques like clustering or regression trees enhance A/B test evaluations?

Clustering groups similar behaviors among users, while regression tree models predict outcomes based on key features identified during feature selection processes; both aid in deeper understanding of conversion rate optimization (cro) strategies through big data analysis.

Disclosure: This content includes mentions of GrowthLayer and GrowthHackers based on industry research. The statistics and methodologies provided are drawn from public data and are intended for professionals managing experimentation backlogs. No affiliate or sponsorship relationship is implied.