Earn attention.
Build the evidence.
Distinctive cars are the creative idea. Careful testing is how we learn what they actually deliver. Here is the plan, the measurement approach, and what we will not claim.
A hypothesis.
Not a headline stat.
We are preparing founder-operated pilots in the Bay Area. The campaign images on this site are fictional creative concepts. We have not yet published Vectonic-specific audience, sales-lift or return-on-investment results.
Our working question: can a distinctive, well-branded car create useful local attention and trackable customer interest at a price that makes sense for a small business?
Every trip.
On record.
We plan to use a GPS tracker or logging app on the campaign vehicle to record each approved trip. A trip ID and aligned timestamps would connect the route log, footage and review notes.
Where we went
A timestamped route within the approved service area, with the campaign portion separated from personal travel.
Miles delivered
Logged distance and eligible campaign miles, with travel outside the agreed scope identified separately.
Time delivered
Start and end times, driving time, approved stationary appearances and pauses. Elapsed trip time and active campaign time are reported separately.
Record quality
Missing GPS points, interruptions and any corrections remain visible in the report. We would check the log against trip notes and odometer readings where practical.
Same route.
Measured carefully.
Start with an affordable, approved promotional design on one car. Compare it with the same car without the promotion before investing in a larger wrap or expanding to other drivers.
- Plan five matched pairs. Each pair has one baseline run and one promotional run: 10 valid runs total. Match the route, time of day and day type as closely as practical. Record weather, traffic and interruptions.
- Keep the comparison fair. Use the same car, observation rules and creative treatment. Alternate or randomize the baseline/promotion order within pairs where practical, and document deviations.
- Record three checkpoints in each two-hour run. Capture 30-, 60- and 120-minute totals and interval counts. Those checkpoints belong to one run; they are not three independent experiments.
- Record the test for manual review. After the camera setup and recording process are approved for the pilot, use a vehicle-mounted camera, such as a suitable 360-degree camera, with timestamps aligned to the GPS log. Test visibility and image quality first. The driver concentrates on driving.
- Apply one written counting rule. Review footage after each run. Define assessable encounters and the visible behavior that qualifies as a look toward the car before counting. Mark obstructed or ambiguous cases as uncertain, record the reviewed time window, and have a second reviewer check 20% of the footage. Record agreement and resolve disagreements. If footage is unusable, a rehearsed passenger or stationary observer method may be used and clearly labelled.
- Track actions separately. Use a dedicated campaign link and, where a merchant supports it, an offer code. Reconcile delivery logs, observations and tagged visits after each run, then compare the matched pairs.
Record the first observation window, delivered miles and operating conditions. This is a checkpoint in the same run, not an independent experiment.
Delivery. Attention.
Action.
These are different measurements. We will keep them separate and explain the collection method, coverage and limitations with each result.
| Measure | What it can show | What it cannot establish |
|---|---|---|
| GPS campaign delivery | Trip IDs, timestamped routes, eligible miles, start/end times, active campaign time, approved stops, pauses and gaps, supported by delivery photos. | How many people looked at, read or remembered the advertisement. |
| Manually observed looks | Assessable encounters, observable look events, uncertain cases and footage coverage for the reviewed windows, with reviewer agreement. | Verified eye contact with the ad, a unique audience, intent, recall or every person who may have seen the car. |
| Tagged visits | Visits attributed to a campaign-specific link or QR destination in the chosen analytics system. | Every person exposed to the car, or proof that the car caused every visit. |
| Enquiries & redemptions | Submitted enquiries and offer-code redemptions confirmed by the participating merchant. | Total incremental sales or campaign ROI without additional evidence and an appropriate comparison. |
Teach the process.
Then test the software.
Future development · Not yet built
We want to explore AI-assisted footage review that flags possible looks toward the car, links them to a trip and time window, and helps a reviewer prepare counts. It would support a defined observation method; it would not establish that someone read, remembered or liked an advertisement.
- Build a manual reference set. Review pilot clips using the same definitions for assessable encounters, observable looks and uncertain cases. Have a second reviewer check a sample and resolve disagreements.
- Evaluate on separate footage. Compare the proposed system with human review on clips not used to develop it. Check false detections and missed looks across camera angles, distance, lighting and obstructions.
- Keep a human in the reporting process. Review flagged and uncertain events, report the portion of footage that could be assessed, and disclose the method and measured error before using AI-assisted counts in client reports.
Small study.
Clear boundaries.
Five matched pairs can help improve operations and inform the next test. This remains a small, local observational pilot. Traffic, weather, route conditions, repeat passersby, missed observations and differences between days can affect the results.
We will show the individual runs, explain missing or excluded data and distinguish estimates from directly recorded activity. We will not turn a traffic count into a verified impression count or apply one route’s result to every car, city or campaign.
Planned data handling
Before camera testing begins, define the recording purpose, applicable permissions, authorized reviewers and a retention period for that pilot. Keep raw footage and detailed trip logs access-restricted. The initial visual study is planned without audio recording or facial recognition. Shared reports should use aggregate counts and campaign-only routes; any shared clips should obscure identifiable faces and licence plates.
Recording coverage is part of the result. Camera angle, darkness, distance, obstructions and repeat encounters can affect a count. We will label uncertain or unreviewable periods and will not present a model’s output as a verified “view.” See planned campaign data handling.
Read the foundations.
A standards reference for distinguishing audience and exposure measures and documenting how they are produced. Used here to guide terminology, not to claim accreditation.
How campaign parameters identify referral sources in analytics. Tagged traffic is one part of a report, not a count of everyone who saw a car.
Industry background on planning and interpreting out-of-home measurement. It does not validate Vectonic’s proposed campaigns.
Help shape
the first pilot.
Bring a local offer and a clear goal. We will discuss a manageable campaign, a realistic scope and the evidence worth collecting.
Discuss a pilot