How to measure B2B video beyond views

Measure the job you hired the video to do

Choose a primary outcome before production: a buyer understands the offer, a prospect takes the next step, a user completes a task or a sales team can explain a workflow consistently. Then measure playback as context, not as the outcome itself. A view says the asset was encountered under a platform’s counting rules. It does not tell you what the viewer understood.

For a small B2B team, a useful scorecard combines three things: where the video was used, what viewers did next and what the team learned from actual conversations. You do not need to attribute every sale to a film to make a better production decision.

Different videos need different scorecards

A homepage explainer and a support tutorial should not compete on the same completion-rate target. A buyer may get the point early and click through. A learner may replay one difficult step. A high completion rate can coexist with confusion, while a useful short visit can look like a low-engagement session.

Choose one primary question and a supporting signal. For a product demo: did qualified prospects move to the relevant next step? For a tutorial: could users complete the task, and what still required help? For a sales asset: was it used at the right point in conversations, and did it resolve a recurring explanation problem? Treat these as questions to investigate, not outcomes guaranteed by production.

Check what your analytics actually records

Do not assume installing analytics tracks all video playback. Google Analytics enhanced measurement documents video events for supported embedded YouTube players with the JavaScript API enabled. A self-hosted website video requires its own appropriate player-event implementation; verify the events in the real player.

Separate an intentional play from a silent autoplay preview. Otherwise a looping homepage film can inflate a number that looks like active interest. Use stable asset IDs and record the page context and video version so different edits do not silently share one result.

Test one play, pause, progress and completion sequence where those events are supported. Check for duplicate listeners and repeated loop completions. Exclude internal QA where practical. Do not send names, email addresses or form-message text into analytics event fields. Measurement should not turn private enquiries into tracking data.

Sources: Google Analytics Help: Enhanced measurement events

Use evidence at the strength it actually supports

In the public Krista.ai case study, John Michelsen’s feedback calls for the work to reach partners and be used by sales representatives. That is useful evidence of intended sales use and client response. It is not a revenue attribution study, and we should not present it as one.

The same distinction helps an internal report. “Sales used the video in six conversations” is adoption evidence. “Three prospects said it clarified the workflow” is qualitative comprehension evidence, if recorded accurately. Neither statement should become “the video generated three customers” without evidence of that different claim. These numbers are illustrative examples, not FDM results.

Sources: FDM: Krista.ai product explainers and client response

Do not mistake an association for an experiment

Google Analytics describes attribution as assigning credit to touchpoints under a model. That can help organise a conversion journey, but a simple comparison of people who watched with people who did not watch is not a controlled test of your video. More interested prospects may be more likely to watch in the first place.

If traffic and circumstances allow, a properly designed test can answer a narrower question, such as whether a specific placement improves a defined next step. Keep other changes stable and agree the outcome before looking at the result. With small samples, report the counts and uncertainty rather than declaring a winner from a dramatic-looking percentage.

When a controlled test is impractical, combine a consistent before-and-after observation with sales feedback and known context such as a campaign or pricing change. Call it indicative evidence. Honest limits make the report more useful than a confident claim the data cannot support.

Sources: Google Analytics Help: About attribution and attribution modelling

End the report with a production decision

Review a small set of signals together. Where are viewers encountering the video? Are the right people taking the intended next step? Which question still comes up? What part is being replayed or abandoned? Use the answers to decide whether to change the opening, clarify a scene, move the placement or create a missing follow-up asset.

Do not automatically shorten a film because completion is low. First check whether the audience, placement or opening promise is wrong. Do not make a second film merely because the first accumulated views. A useful measurement system tells you what to improve or stop, not only what to celebrate.