Your AI film got views — what happened next?
Start with the job you gave the film. Views can show that it was watched, but they do not tell you who became interested or contacted you. Keep platform views, website sessions and verified enquiries separate. Compare the same reporting window and audience, and record missing data as unknown rather than zero.
Choose the question before opening the dashboard
A film made to introduce a gallery has a different job from a film inviting people to book a visit. When I plan a project with SOLONSKY Films, I want that job written in one sentence. Otherwise a large view count can make an unclear result look convincing, and a small view count can hide a useful conversation.
Write the action you hope someone will take after watching. It might be recognising your name, opening the programme or sending a specific question. Then decide where you can observe that action. A social platform, a website and an inbox see different parts of the journey. They should not be treated as one continuous counter.
- One purpose for this release
- One observable next step
- One reporting period shared by the versions
- A separate note for what cannot be measured
Put the same window around each version
Imagine releasing one cut on Monday and another on Friday. A Sunday report gives the first one more time to find viewers. Comparing their lifetime totals would reward the earlier release before you even examine the films. Compare equal age windows, and keep the publication times beside the numbers.
The audience also matters. A post sent to existing followers is not directly comparable with an ad shown to strangers. Language, placement and distribution can change the result alongside the edit. I would write those differences into the report rather than silently crediting the film for everything that happened.
Follow a fictional gallery invitation
For a teaching example, picture a gallery film that ends with an invitation to its weekend programme. The social post links to a page with the dates and a contact button. A viewer may enjoy the film without opening the link; someone who opens the page may still decide that the dates do not suit them.
There is no client result hidden in this example. It shows why each number answers a different question. A click towards the contact page records interest in that next step. A successfully received message records an enquiry. Whether the message concerns a relevant project requires a separate human check.
| Observation | What it can support | What it does not prove |
|---|---|---|
| Platform views | The platform counted viewing under its own definition | A unique person or a future order |
| Website sessions | Recorded visits under the website measurement rules | All people who saw the film |
| Verified enquiries | Messages actually received and checked | That this film alone caused the enquiry |
Find the part that needs a closer look
If people follow the link but rarely take the next step, open the landing page on the device they actually use. Is the date easy to find? Does the page continue the promise in the film? Can a visitor understand what to ask? A weak handoff between the film and the page can deserve attention before another production round.
If few viewers follow the link, inspect the invitation and where the link is placed. A platform may put the link in a different part of the interface, or the audience may be enjoying a story without needing the service. This is a question to investigate, not a diagnosis obtained by dividing two small numbers.
Leave a visible place for unknowns
Consent choices, tracking blockers, a message sent through another channel and delayed reporting can all leave gaps. Do not fill an empty cell with zero unless the relevant source actually reports zero. Nor should you add views from different platforms and call the sum unique reach; the same person may appear more than once.
A gallery visitor might remember the film and later search the name. A referral might arrive through a friend. Asking how someone found you can add context, but it still does not establish a controlled experiment. I keep that context beside the measured result instead of converting it into a precise attribution claim.
Change one decision and keep the record
Use the worksheet to compare two sets of website sessions and sessions containing a defined next step. These are same-source counts, not platform views mixed with website clicks. Blank inputs stay unknown. A rate is descriptive: it cannot prove a winning edit or tell you whether a small sample is enough.
After the first report, choose one practical change you can inspect: a clearer final invitation, a landing-page correction or an opening that identifies the audience sooner. Keep a copy of the previous cut and the reason for the change. For the next conversation with SOLONSKY Films, bring that evidence and the remaining question; it is more useful than asking for a vaguely more viral film.
Watch an authored AI film
An authored film from SOLONSKY Films. It illustrates the storytelling approach.
The video provider connects only when you press play. Watch on YouTube.
Questions
Does a view count tell me how many people watched?
Only if the platform explicitly reports unique viewers under its current definition. A view count may include repeat viewing; record the exact metric name.
Can I call every contact click a lead?
Keep a click as a click. Verify that a message was received, then review whether it is relevant. A payment or confirmed order is a further stage.
Should I change the film after one quiet day?
Check whether the planned reporting window has finished, whether the link works and whether data is delayed. One quiet day is not enough to assign a cause.
Technical sources
- Google Analytics — automatically collected eventsOfficial definitions. A form event is not proof of a qualified enquiry; the examples and decision worksheet here are my own.