What the evidence supports in a weekly behavior experiment.
ChangeLoop 7 uses a week to review one selected action. This page explains which studies inform that design, how their scope differs, and what they cannot establish about the product. Research on a mechanism is separate from evidence that a complete software workflow works.
Updated and sources accessed: October 24, 2026. Source links lead to publisher pages, authors' university repositories, or the US National Library of Medicine's records of the original papers.
How research informs the design
The working sequence has four parts: define a specific action, observe whether it happens, keep a prepared smaller fallback available, and review the record before choosing the next cycle. Seven days is a product-design interval. It makes several real opportunities visible without requiring an indefinite commitment; no cited paper establishes seven as the best interval for every goal.
For example, imagine a fictional user preparing a meeting question after opening the calendar. On three days the calendar cue occurred and the question was prepared; on two days it occurred without the action; on two days no meeting was scheduled. This record separates an observed opportunity from an attempted action. It does not establish that preparation improved the meetings, caused confidence, or created an automatic habit. Those are different questions requiring different measurements.
The product's fixed intake, three method alternatives, one selected method, current-day task, and written review are implementation choices. Multiple private model roles challenge and consolidate the proposal, but model agreement is not independent scientific replication. The customer still chooses the next step. The workflow page explains those choices; the source map below explains the evidence limits.
Source map: study, supported scope, and limit
- Lally et al. (2010), habit automaticity. In a 12-week study, 96 volunteers chose an eating, drinking, or activity behavior in a consistent context. Models were fitted for 62 participants, with a good fit for 39. Estimated times to 95% of the modeled automaticity plateau ranged from 18 to 254 days. Those are fitted estimates, including estimates beyond the observation period, not a guaranteed range for each reader. Publisher's paper and abstract. The result supports variable timelines and repeated contexts; it cannot establish a seven-day transformation or product-specific outcome.
- Gollwitzer (1999) and Gollwitzer & Sheeran (2006), implementation intentions. An if-then plan connects a specified situation with a chosen response. The 2006 review combined 94 independent tests and found a positive average effect on goal attainment. That supports discussing concrete plans while keeping populations, tasks, and study conditions visible. It does not show that every generated plan fits its recipient. 1999 author university record; 2006 publisher record.
- Webb & Sheeran (2008), planning mechanisms. One analysis examined 66 tests and found negligible effects of implementation intentions on goal intentions and self-efficacy. A second study examined cue accessibility and cue-response links. This helps explain why specifying an opportunity is different from merely feeling motivated. It does not prove an individual plan's effect or a universal weekly schedule. Original-paper abstract and DOI.
- Harkin et al. (2016), progress monitoring. The meta-analysis included 138 studies and 19,951 participants in its progress-monitoring analysis. The goal-attainment analysis used 18,398 participants and reported an average effect of d = 0.40, with variation between studies. These denominators refer to different analyses; neither is a ChangeLoop 7 customer sample. Recording a signal can support review, while a self-reported Done mark remains a report, not independently verified improvement. Authors' university manuscript.
- Carrera et al. (2018), limits of simple planning. In a randomized field experiment involving 877 gym members, one group planned when to attend over the next two weeks and another recorded previous exercise. The planning intervention did not produce a positive behavioral effect. Keeping this null result beside favorable planning research prevents a one-sided conclusion: a plausible mechanism can fail for repeated behavior. Authors' University of California record.
- Keller et al. (2021), routine-based versus time-based cues. The trial assigned 192 adults to two cue-planning formats for a selected nutrition behavior and collected daily questionnaires for 84 days. It found no group-level difference between the formats; repeated plan enactment predicted automaticity. Among participants who successfully formed habits, the median modeled time was 59 days. This is not a deadline for all participants and does not identify a universally superior cue. Original-paper abstract and DOI.
- Michie et al. (2013), Behavior Change Technique Taxonomy v1. The framework describes 93 techniques in 16 groups. Its purpose is precise naming and reporting of intervention components, so planning, prompts, monitoring, and environmental changes do not become one vague promise. A classification framework does not demonstrate that a particular technique, an AI-generated action, or the complete product improves outcomes. Original-paper abstract and DOI.
- Michie et al. (2009), combined intervention techniques. A meta-regression included 122 evaluations involving 44,747 participants in healthy-eating and physical-activity interventions. Interventions combining self-monitoring with at least one other control-theory technique had larger effects than other interventions in that analysis. This is evidence about combinations in studied settings, not proof that a binary completion record is sufficient. Original-paper abstract and DOI.
- Wood & Rünger (2016), psychology of habit. This review describes repeated responses in recurring contexts and the relationship between habit and deliberate goal pursuit. It supports examining context as well as motivation. Repetition alone does not show that a behavior is beneficial, and repeated software tasks do not automatically become habits. Authors' University of Southern California manuscript.
- Compernolle et al. (2019), sedentary behavior monitoring. Nineteen intervention studies with 2,800 participants were included in this review. It found short-term reductions in sedentary time, with substantial heterogeneity and limits on attributing effects specifically to monitoring. Results involving objective monitoring tools should not be transferred unchanged to a self-reported text checkbox or to unrelated goals. Original-paper abstract and DOI.
- Krukowski et al. (2024), feedback with self-monitoring. This review included 19 studies with 3,261 participants. Findings varied by outcome and feedback format; study heterogeneity limited comparisons, and many studies had a high risk of bias. It therefore supports investigating feedback while leaving open which presentation or generation method is best. It does not demonstrate that AI feedback outperforms human feedback in this product. Original-paper abstract and DOI.
All source records in this map were accessed October 24, 2026. These are research references, not customer testimonials or endorsements of ChangeLoop 7.
Public claim register
| ID | Claim | Class | Source | What the source supports | Limit |
|---|---|---|---|---|---|
| CL7-R01 | Seven days is used as an observation and review interval, not a habit-formation promise. | Research + design | Lally et al. (2010); weekly-cycle specification | Habit development varies by person and behavior; a fixed weekly checkpoint is a product-design choice. | Does not prove that seven days forms a habit or produces permanent change. |
| CL7-R02 | A small action may be expressed as a specific cue-response plan. | Research mechanism | Gollwitzer (1999); Gollwitzer and Sheeran (2006) | Implementation intentions can help connect a defined situation with an intended response. | Does not establish that every generated plan is suitable or effective for every customer. |
| CL7-R03 | Recording an observable signal can support review of goal-directed behavior. | Meta-analysis | Harkin et al. (2016) | Progress monitoring can support goal attainment under studied conditions. | A self-reported Done or Not done mark is a signal, not independent proof of an outcome. |
| CL7-R04 | Behavior-change techniques should be named precisely instead of treated as one vague intervention. | Research framework | Michie et al. (2013) | A structured taxonomy can describe intervention components consistently. | A taxonomy does not validate the complete ChangeLoop 7 workflow. |
| CL7-P01 | The customer answers a fixed intake, receives three method options, selects one, and follows one current-day action. | Product fact | Public workflow specification | Describes the intended and testable product workflow. | An implementation fact is not scientific evidence that the workflow changes behavior. |
| CL7-P02 | ChangeLoop 7 is not an open-ended AI chatbot. | Product boundary | Workflow comparison; FAQ | AI is limited to defined analysis checkpoints; ordinary daily status does not open unlimited chat. | This distinction does not establish lower cost, better outcomes, or superiority for every user. |
| CL7-C01 | One introductory seven-day cycle is offered without charge; published subscription prices apply before cycle two. | Commercial fact | Current pricing page; machine-readable pricing | Describes the currently published offer and billing boundary. | Pricing can change and is not evidence of effectiveness, payment, revenue, or customer demand. |
| CL7-O01 | Product-specific completion, retention, sustained change, and comparative-effectiveness outcomes are not established. | Evidence gap | Product-specific evidence gaps | Makes the current absence of approved outcome evidence explicit. | No testimonial, success rate, clinical claim, or guaranteed result should be inferred. |
Read a claim in three steps
- Identify the kind of statement. A research mechanism, a software rule, a price, and a customer outcome require different evidence. The register's R, P, C, and O identifiers preserve those categories.
- Check the match. Compare the actual population, behavior, measurement, duration, and result with the sentence beside the citation. A paper about gym attendance does not automatically establish workplace productivity or mental-health benefit.
- Read the limitation before applying it. A null result, a self-report measure, or a short follow-up can change what is reasonable to conclude. A relevant source can inform a small personal test without promising that the test will succeed.
Traceability is not proof. A stable source identifier helps a reader locate the work. It does not settle whether an interpretation is accurate or whether the findings transfer to a particular person. If a statement cannot be supported, the appropriate response is to narrow it, remove it, or mark the gap rather than substitute model confidence.
What still needs product-specific evidence
The complete ChangeLoop 7 workflow has not been established as a clinical intervention. This page supplies no verified product completion rate, retention rate, sustained-change rate, or comparative-effectiveness estimate. Those would need a defined population, denominator, observation period, outcome measure, and appropriate comparison. A fictional example, deployment, registered account, or completed week cannot supply that evidence by itself.
Authentic customer reports could describe individual experiences, but would still require context and limitations. Research mechanisms, implemented product facts, commercial facts, customer reports, and unverified outcomes remain separate evidence classes. Current prices belong on pricing; source and correction rules belong in the editorial policy.
ChangeLoop 7 does not diagnose, treat, prevent, or cure a medical or mental-health condition. It is not a substitute for qualified professional support or an emergency service. It does not promise permanent behavior change or a result within seven days.
Use the structure without overstating the evidence
The free browser-local experiment worksheet records one cue, one minimum action, a fallback, an observable signal, and a day-seven review question. It requires no account or AI call. Use the record to decide what is worth attempting next; do not treat it as a controlled study, diagnosis, or general recommendation for other people.