Test screenings are one of the least-discussed forces in mainstream filmmaking. Audiences queue up in shopping-centre car parks, sign non-disclosure agreements, and watch unfinished films with temp music and incomplete visual effects. Then they fill out a card. Those cards, and the data that flows from them, can redirect a film's ending, cut a character entirely, or push a studio to order weeks of reshoots. The process is neither as crude as critics assume nor as precise as studios pretend.
What a test screening actually involves
A preview screening, in the industry sense, is a structured research exercise. Studios partner with market-research firms, most prominently NRG (National Research Group), to recruit a sample audience that mirrors the film's target demographic. Recruiters intercept people at shopping centres and cinemas, screening them with short demographic surveys before handing out tickets. The goal is a room of real filmgoers rather than industry insiders.
The audience watches the film, then completes a written questionnaire. Scores are tallied across two key metrics: overall rating (typically "excellent, very good, good, fair, poor") and definite recommend ("would you definitely recommend this film to a friend?"). A film is considered to be tracking well when its "excellent or very good" score exceeds 80 and its definite recommend exceeds 70. Those numbers aren't guaranteed targets. They're benchmarks studios use to assess whether they have a problem.
After the written portion, a moderator pulls 20 to 30 audience members into a focus group. These sessions run for up to 90 minutes. The moderator asks about specific scenes, characters, and the ending. The director and producer typically sit behind a one-way mirror, or in an adjacent room watching a live feed.
When the cards say something unexpected
The most famous test-screening rescue story belongs to Fatal Attraction (1987). Preview audiences hated the original ending, in which Glenn Close's character drowns herself and Michael Douglas is arrested for her murder. Cards were devastating. The studio ordered reshoots, and the film got its now-iconic bathtub resurrection sequence. Box office followed: the film grossed over $320 million worldwide against a $14 million budget. That figure made the reshoot mythology stick.
Less flattering examples exist. Blade Runner (1982) famously screened poorly in early previews, prompting Warner Bros. to add Harrison Ford's voiceover narration, which Ridley Scott considered a creative violation. Scott spent years restoring his vision through subsequent director's cuts. The tension between preview data and directorial intent is where test screenings earn their contested reputation. For a deeper look at how those restoration decisions unfold, the craft involved in the director's cut process offers useful context.
What the data can and cannot measure
A written survey card captures immediate reaction. It cannot capture whether that reaction will age well. The Shining (1980) screened to poor audience response on its first preview. It's a Wonderful Life (1946) underperformed at preview and at the box office. Both are now considered canonical. Audience research captures the average viewer's gut response to an unfinished film in an artificial setting. The methodology isn't broken, but it has a known ceiling.
The data also flattens tonal complexity. Horror comedies, slow-burn dramas, and formally experimental films tend to score below genre average at previews because audiences calibrate expectations mid-screening and rate the film against what they thought they were watching. Studios that rely heavily on test-screening scores for these genres often end up with safer cuts. It's a structural bias built into the format itself.
That said, test screenings catch genuine problems that internal teams miss. When a film has been in edit for six months, every collaborator is blind to its pacing. A fresh audience will tell you, with remarkable consistency, exactly where the second act loses them. The feedback on pacing and structural drag is often the most actionable output of the entire process. This is especially relevant to what makes the second act so difficult to execute, a problem preview scores frequently expose first.
The reshoot pipeline that preview scores trigger
When preview scores fall below studio thresholds, the calculus shifts quickly. The producer calls the director. The director calls the editor. If the problem is structural, reshoots are scheduled. Reshoots ordered from test-screening feedback are distinct from planned additional photography, though the distinction collapses in public perception. Both get labelled as creative failure. Neither usually is.
The logistics are considerable. Cast members have moved on to other projects. Locations have been released. A reshoot triggered six months after principal photography might require new sets, wigs to match actors' changed haircuts, and digital doubles for secondary characters whose actors are unavailable. One week of reshoots can cost between $1 million and $5 million depending on the scale of the production. Studios factor this risk into their previewing schedule, front-loading screenings early enough that reshoots remain financially viable.
Streaming's effect on the preview process
Netflix, Amazon Prime Video, and Apple TV+ have complicated the traditional test-screening model in two ways. First, they have access to real-time viewing data at a scale no preview card system can replicate. Second, their films don't always need to perform in cinemas, which removes one of the primary financial pressures that made preview scores so consequential for traditional studios.
Netflix has largely moved away from extensive traditional preview screenings in favour of internal data modelling and limited press screenings. The trade-off is visible. Some Netflix originals have released with structural problems that a live preview audience would likely have flagged early. The feedback loop is faster in data terms and slower in human terms. Whether that produces better films is genuinely unclear.
Traditional studios, meanwhile, have held onto the preview system precisely because theatrical box office remains their primary revenue model. A film that opens to $20 million instead of $40 million costs the studio more than six weeks of post-production reshoots. The economics keep the research process alive.
What filmmakers actually do with the feedback
Experienced directors treat preview feedback as data, not instruction. They read every card, cross-reference comments with their own instincts about what the film needs, and ignore the noise. The noise is considerable: preview cards frequently contain contradictory instructions from the same screening, with some audience members demanding a faster pace while others request more time with the same scenes.
The useful signal tends to be consistent: two-thirds of the audience losing interest at the same scene, or near-universal confusion about a character's motivation. That kind of convergence is hard to argue with. Ignoring it requires either great confidence or a track record that justifies it. Most directors, even those with significant creative control, acknowledge the value of the data while reserving the right to disregard the proposed solutions.
The preview screening doesn't make a film. It tells the people who made it what a room of strangers saw. What they do next is still craft.