Executive Summary
An AI recommendation is an exposure opportunity, not a retained client. The path between them may include a website visit, a telephone conversation, a referral from another person, a conflict check, a consultation, and a decision about whether the firm can accept the matter. A marketing report that labels all of these events “conversions” obscures the decisions that partners actually need to make.
This paper proposes a measurement architecture that connects discovery evidence with intake and matter outcomes while preserving uncertainty. It is an analytical framework, not original research into consumer behavior and not a report of 360AI Legal customer results. No client records were analyzed. The examples and process exhibits are hypothetical; no conversion rates or revenue benchmarks are asserted.
The practical recommendation is to keep three views: observed channel evidence, client-reported discovery, and confirmed business outcomes. Link them only where the firm's approved systems and permissions support the connection. Do not force every client into a single source category when the evidence is mixed or missing. Use the resulting record to identify weak handoffs, evaluate acquisition economics, and decide what further testing is justified.
01. Define the acquisition stages before buying a dashboard
Agree on a lifecycle with marketing, intake, and the professionals responsible for accepting engagements. A useful starting sequence is identifiable visit, inquiry received, qualified inquiry, consultation completed, engagement accepted, and fee outcome recorded. Each stage needs an owner and an operational definition. Some practices will need different stages; the important feature is that the definitions remain stable enough to support comparison.
A qualified inquiry should reflect the firm's actual acceptance criteria, not simply a completed form. Geography, practice area, conflicts, timing, capacity, and client needs can all matter. Record disqualification reasons at a safe level of detail so that the marketing team can distinguish irrelevant demand from an intake backlog. Do not place confidential explanations into a shared marketing spreadsheet to achieve that distinction.
Define “signed client” with the firm's authorized engagement process. A scheduled meeting, an uploaded document, or an electronic signature event may not independently establish an accepted engagement. The marketing system should consume the approved status from the responsible operational system rather than invent a proxy. When dates change, retain the relevant stage timestamps so a report can distinguish the month of first inquiry from the month of acceptance.
| Stage | Evidence owner | Keep separate from |
|---|---|---|
| Discovery | Marketing / platform evidence | A confirmed inquiry |
| Qualified inquiry | Intake team | A scheduled consultation |
| Accepted engagement | Authorized engagement owner | A marketing event |
| Fee outcome | Finance | Projected matter value |
Original editorial framework. This exhibit contains no observed client results, performance statistics, or market estimates.
02. Separate observable evidence from reported influence
Start with evidence that the firm can actually observe: a recorded referring domain, an incoming campaign parameter, a landing page, an authorized call-tracking event, or an intake source field. Preserve the original value alongside any normalized channel classification. If classification rules change later, that raw field makes it possible to explain or reconstruct the difference.
Add a short, optional intake question about how the person found or evaluated the firm. Allow multiple answers and a free-text explanation in the appropriate protected system. Ask neutrally rather than prompting the desired AI answer. “Where did you first hear about us?” and “What helped you decide to contact us?” measure different memories. A person may first hear about the firm from a friend and later consult an AI assistant.
Neither source is complete. Referral evidence may be absent when a person changes devices or contacts the firm directly. Recalled discovery can be incomplete or imprecise. Treat agreement as corroboration and disagreement as mixed evidence. The correct response is not to rewrite the client's account to match the analytics platform. Report an unknown or mixed category when the available record cannot support a stronger conclusion.
03. Establish a privacy-conscious data contract
Create a written contract between website analytics and intake systems specifying allowed fields, owners, retention, access, and permitted uses. Keep free-text matter descriptions and personal identifiers out of general marketing analytics. Google Analytics policies prohibit sending information it can recognize as personally identifiable, and its guidance calls attention to leaks through URLs, page titles, and form inputs. 1
Where a lawful, approved integration needs a record identifier, review whether that identifier and the surrounding data can reveal an individual. Calling a value “anonymous” does not make a re-identifiable dataset anonymous. Use the minimum necessary fields and keep the lookup relationship within the systems authorized to hold it. Obtain appropriate technical and legal review before implementing cross-system tracking.
Law-firm confidentiality duties add a separate reason to minimize collection and access. ABA Model Rule 1.6 addresses information relating to representation, but applicable rules and obligations require jurisdiction-specific review. This article does not determine the legality of a firm's tracking setup. Its operational recommendation is to involve the firm's responsible counsel and privacy owner before sharing intake or matter information with marketing vendors. 2
04. Build a reconciliation workflow instead of a perfect-looking funnel
Create a periodic reconciliation between accepted website submissions, intake records, and lifecycle status changes. Check for duplicate inquiries, spam, internal tests, and repeated contacts about the same matter. A prospect who calls and then completes a form should not automatically become two opportunities. Document the deduplication rule and preserve exceptions for review.
Use stage totals from the operational system as the control totals. Compare the marketing export against them and explain discrepancies before reporting channel performance. If an integration loses records, label the affected period incomplete. Do not fill the gap with an estimate that looks like a measured result. Where a correction is possible, preserve the correction date and the prior published report.
For a hypothetical family-law practice, an AI-referred website visit may be followed by a phone inquiry whose source field is blank. Without an approved matching method, the records cannot confidently be connected. Keep the web visit in the traffic report and the inquiry in the operational report. A cautious “unlinked” status is more useful than an unsupported conversion that later distorts spending decisions.
05. Choose attribution rules and disclose their limits
Attribution allocates credit among recorded touchpoints. It does not, by itself, prove which marketing activity caused a person to retain the firm. Google Analytics supports different attribution settings and distinguishes event-scoped reporting from user- and session-scoped traffic dimensions. The same activity can therefore appear differently across reports without either report being an exact history of human decision-making. 3
Select a primary reporting rule that the firm can explain, then retain useful secondary views. A first-known source view may help assess discovery; a last-recorded touchpoint view may help assess the immediate contact path; a client-reported influence view may reveal channels that left no technical trace. Do not add the credited outcomes from these overlapping views and present the result as unique retained matters.
Record the attribution window, eligible channels, deduplication approach, and treatment of direct or unknown traffic. Changing those assumptions can alter historical totals. If the firm changes a rule, label the break or recompute the comparison period consistently. A dashboard should make uncertainty visible, especially when a small number of matters drives the apparent result. Precision in formatting is not the same as precision in inference.
06. Calculate acquisition economics with compatible units
Use cost and outcomes from a clearly defined scope. A proposed cost-per-qualified-inquiry calculation divides the costs included in the analysis by the number of qualifying inquiries under the agreed definition. List whether costs include agency fees, content production, software, paid distribution, or internal labor. Comparing a fully loaded program with a channel report containing only advertising spend is misleading.
Treat retained-matter economics as a cohort question when the intake cycle crosses reporting periods. An inquiry received late in one month may be accepted later. Comparing this month's spending with this month's signatures can mix unrelated cohorts. Report cohort maturity and pending matters rather than calling a still-developing cohort a failure or a success. The appropriate observation window depends on the practice and should be agreed, not borrowed from a generic marketing benchmark.
Distinguish projected matter value, contracted fees, collected fees, and contribution after relevant costs. Contingency matters and hourly engagements have different economic timing. A marketing report should not present speculative recovery or lifetime value as realized revenue. The finance and practice-management owners should define the fields before anyone calculates return on investment. When reliable financial attribution is unavailable, report the narrower operational measure honestly.
07. Test contribution rather than assuming causation
A rise in signed matters after publishing content may be encouraging, but it does not isolate the content's effect. Demand, seasonality, staffing, response time, competitor activity, and other campaigns may change at the same time. Record those conditions in the analysis. Use the report to formulate a testable explanation rather than to announce that an AI recommendation caused every subsequent engagement.
Where volume and operations permit, consider a staged rollout or a comparison between reasonably similar practice or location groups. Define the intervention, outcome, observation period, and comparison criteria before evaluating results. A control group that differs systematically in demand or intake capacity may produce an attractive but invalid comparison. An experiment can also be impractical or inappropriate; saying so is better than assigning causal language to an uncontrolled time series.
For many small firms, the most valuable early finding will be operational. A delayed callback, an unusable mobile form, or unclear eligibility language can prevent relevant demand from becoming a consultation. Correcting those defects is worthwhile even before a channel-specific causal estimate is available. Keep the engineering acceptance test separate from the later commercial evaluation, and retain a record of other changes that could affect interpretation.
08. Methodology, reporting, and accountable ownership
This paper synthesizes platform documentation and professional-responsibility references into a proposed operational framework. It does not report a survey, experiment, law-firm dataset, or original causal finding. The source list supports specific factual statements; the data model, stage definitions, and reporting recommendations are editorial analysis for review and adaptation.
A useful monthly package includes channel evidence, client-reported sources, stage conversion counts, cohort maturity, unresolved matching gaps, spending definitions, and action owners. Use a small number of views that answer distinct questions instead of a single impressive-looking total. Provide a glossary and enough detail for a partner or finance lead to challenge the assumptions.
Assign responsibility explicitly. Marketing owns campaign classification; intake owns inquiry handling and qualification; the engagement owner confirms acceptance; finance governs revenue definitions; the privacy owner approves the permitted data flow. Agencies should receive only the information needed for their authorized role. Agree a correction process so that a discovered integration or classification error updates the report without quietly rewriting the historical record.
09. Key findings and business implications
The principal analytical finding is that discovery, influence, and acquisition should be connected without being conflated. AI visibility measurement can show how a firm appears in a defined set of answers. Website analytics can show a subset of visits. Intake and engagement records establish operational outcomes. Each source has a boundary that remains important after integration.
For a CMO or marketing director, start with common lifecycle definitions and reliable intake reconciliation. For an agency, make the evidence and attribution rules inspectable before making a performance claim. For a managing partner, ask which numbers are observed, which are allocated by a model, and which are forecasts. That distinction is more valuable than a guarantee about an emerging channel.
Future AI-mediated discovery may make cross-channel measurement harder or more useful; either direction is a planning hypothesis, not an established result here. Build a reporting process that can accommodate new evidence without changing what “retained client” means. The defensible goal is better decisions about acquisition investment, supported by records the firm can verify, rather than a dashboard that assigns certainty to an incomplete customer journey.
Sources & references
Official documentation reviewed November 13, 2026. Platform features and policies can change. The linked sources support the cited facts; proposed workflows are editorial analysis.
- Google Analytics: Avoid sending personally identifiable informationhttps://support.google.com/analytics/answer/6366371
- ABA Model Rule 1.6: Confidentiality of Informationhttps://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_1_6_confidentiality_of_information/
- Google Analytics: Select attribution settingshttps://support.google.com/analytics/answer/10597962


