Orbital update: a service brief the research assistant can check

In brief

The completed change is the new public measurement review. The retrospective record is a company service sheet last modified on 23 May 2025. The next proposed checkpoint is to audit 10 permitted brief fields by 5 November 2026, then compare a source-based human summary with an assistant draft using the same inputs. A client should be able to trace each proposed measurement to the decision it serves. The assistant's drafting speed would be recorded separately from its errors and the time a person needs to correct them.

What changed since the previous edition

The new review gives the portfolio a public explanation of why measurement choice matters. It also points to a current industry discussion: the IAB opened its cross-channel measurement guide for public comment on 20 October 2026, with comments due on 4 November. The guide is a draft. Its decision-first framing helps a client state the question before choosing a measurement approach.

Our proposed assistant workflow now has a concrete editorial reference to check against. A draft that turns a gaze measure into a sales claim would fail that check. Your team can now refer to that published measurement review when checking a draft brief. The inspected archive does not show a new signed client study or a completed assistant evaluation since the previous company edition.

The evidence behind a careful handover

In Gordon and colleagues' 2019 advertising study, 15 US Facebook experiments comprised 500 million user-experiment observations. That denominator counts observations across experiments, not unique people and not Caribbean customers. An assistant should preserve the unit when it extracts the number.

The same 2019 source reports 1.6 billion ad impressions. Impressions count exposures. They do not count incremental purchases. The study found that observational approaches often failed to recover the randomised estimates, even with extensive measured information.

Lewis and Rao's 2015 published abstract describes 25 field experiments with US retailers and brokerages, collectively representing $2.8 million in digital advertising spending. These are external study inputs. They establish neither an Orbital study budget nor an expected return for a Barbados business.

That 2015 abstract reports a median ROI confidence interval wider than 100 percentage points. The quantity describes uncertainty in that study set. A research assistant should keep it attached to its population and design; it cannot use the number as a forecast of a client's commercial outcome.

Eye-tracking answers another question. The 2010 Krajbich, Armel and Rangel abstract describes a human binary-choice model tested with eye-tracking and links an erratum. We use its scope to distinguish looking from later purchasing, without extracting a numerical effect from the abstract.

These inspected sources explain why a brief needs definitions and limits. This update contains no pooled analysis. The source-data file preserves the study year, population, quantity and unit for each number so your team can check the extraction without receiving a copyrighted article collection.

What this means for a Barbados retail brief

For a proposed Barbados food-retail example, start with a price label and its package quantity. The Barbados Statistical Service explains that its Retail Price Index tracks a fixed basket monthly and that prices must be associated with specific quantities. That national index supplies economic context. It does not establish what a shopper understood about a particular offer in a shop.

Your local brief could ask whether a shopper can identify the total price, package amount and conditions of a proposed offer. Retain the currency, exact label and task wording. If a later study asks about purchases, define access to permitted transaction records and an appropriate comparison before commissioning it. A Caribbean campaign cannot inherit a US advertising result just because both concern retail.

The Barbados photographs show a fruit stand and local buildings. They locate the example and do not depict an Orbital study, client or participant. The proposed task would require a separately agreed setting and consent. Your team should state which customer group a future study intends to reach and which groups its recruitment would miss.

Apply Adrian Dunkley's People First, Problem Second, AI Last approach. A shopper needs an intelligible offer; the brand manager needs evidence for a specific decision. Define that problem before choosing an AI assistant or a behavioural research method. For this Caribbean example, the first useful product output is a complete, reviewable brief.

The proposed check this week

Our checkpoint for 5 November 2026 is a documentation and drafting test. The counts below are practical audit targets. They do not determine the sample required for a consumer experiment or establish an assistant performance result.

1. Write a Barbados retail decision and its observable outcome. Use a permitted example label and preserve the quantity, currency and conditions shown.
2. Audit 10 brief fields for source, date, definition and permission. Count complete fields and unresolved questions before an assistant receives the packet.
3. Compare one human summary with one assistant draft from the same packet. Record unsupported claims, missing limits, source errors and total correction time.
4. Ask a second reviewer to trace every proposed conclusion to the permitted inputs. Keep a failed draft in the audit and correct the brief before planning a client trial.

Risks for consumers and clients

An assistant can remove a qualification while producing a fluent summary. A client may then commission a study that cannot answer the business question, while a consumer may face unnecessary data collection. Keep the decision and outcome visible, restrict the input packet and require a person to approve the final interpretation.

The main evidence gap is current delivery and evaluation evidence. The archive supports the historical portfolio and the new public review; it does not demonstrate that the proposed assistant improves research quality or work time. Caribbean applicability also needs a local study. The next audit can expose missing fields, but it cannot establish a commercial gain or validate a psychological inference about individual shoppers.