Developer marketing: what to put alongside a video

Of the 5,680 people in the analysis, 1,149 both named video creators among their skill-choice influences and reported using technical documentation to learn coding. A further 603 named video creators without reporting documentation learning. Together, those two groups give the denominator for the 65.6% result: 1,149 out of 1,752.^[1]

The full table matters. There were also 2,677 people who reported documentation learning without naming video creators, and 1,251 who selected neither of those particular answers. These last two groups may have watched videos or used documentation in other circumstances. A response not selected here is not evidence that someone never uses that resource.

Documentation learning was reported by 67.4% of the complete-answer group overall: 3,826 out of 5,680. Among people who did not name video creators, it was 68.2%, or 2,677 out of 3,928. The 65.6% figure therefore shows coexistence, not a higher documentation-use rate associated with naming video creators. It gives no basis for claiming that adding a video will increase documentation use.

For a developer-tool company, the useful question is whether the campaign leaves someone without a resource they may need. An audience interested in creators is not, by itself, a reason to omit a precise written reference. The study does not determine how much to spend on either format or whether a particular creator is a good fit.

The influence item asks what helps respondents decide which technical, AI or professional skills to learn next. It allows up to three sources. The learning-method item asks how they learned coding in the past year and allows multiple selections. One describes a choice about direction; the other describes reported learning activity. They need not concern the same skill, product or moment.^[3][4]

That distinction becomes visible in another result. Among 1,491 video-influenced respondents who did not also name Documentation among their up-to-three influences, 941 still reported documentation learning: 63.1%. Documentation can appear in an answer about learning without appearing in a short list about what to learn next. This is not a test of whether respondents remembered to credit a resource.

The reverse percentage answers a different question again. Of the 3,826 people reporting documentation learning, 1,149 named video creators, or 30.0%. Both 65.6% and 30.0% describe the same 1,149 people. Neither should be used without saying which group is underneath it.

The distinction can improve a brief. Ask what the demonstration should help someone understand, then ask which details should remain easy to find afterward or independently. A reference page should not depend on the reader having watched a video, and a video should not promise a working example that its linked material no longer supports. Those are practical content choices, not outcomes measured by this survey.

The survey is voluntary, and answering these items was optional. We identified 11,071 primary-job coders who reported past-year learning and supplied learning-method answers. Of them, 4,952, or 44.7%, did not answer the skill-influence question. We cannot classify those people as naming or not naming video creators. After excluding answers that included not currently deciding what to learn, 5,680 respondents remained.^[1][2]

This missing group is the main reason to keep the result attached to its denominator. It would be misleading to turn 65.6% into a claim about all developers, all survey respondents or everyone watching a technical video. The survey was promoted mainly through Stack Overflow channels, which also affects who encountered and chose to complete it.^[2]

Two predefined checks change the overlap only slightly. Keeping otherwise usable answers that combined not currently deciding with a named influence gives 65.7%. Restricting the primary group to people whose main employment status was employed or independent gives 65.9%. These checks address those specific choices; they do not remove the missing-answer or recruitment limits.

A useful first step is to open the material a developer would receive today. If an existing guide already explains the demonstrated task, the campaign may only need a clear link and a check that the example still works. Creating a separate campaign page can introduce a second set of instructions to maintain. More assets are not automatically more useful.

Look for a specific gap before adding anything. A feature overview may explain why a tool exists but omit the inputs needed to try it. A technical reference may list every parameter without showing a complete example. Those are different problems. The first may need a worked example; the second may need a short explanation of how the pieces fit together. Decide which question remains unanswered, then choose the smallest useful addition.

The campaign’s scope matters too. A video introducing a technical idea can point to an explanation or a demonstration repository. A video promising that someone can reproduce a result needs instructions sufficient for that promise. The survey does not select the right destination for either case. It gives a reason to check the written resource as part of the campaign, rather than assuming a video-oriented brief has settled the whole content task.

This approach also gives a small team a way to prioritize. Fix the missing prerequisite or outdated example first. Commission another format when it addresses an identifiable need, rather than using the overlap percentage as a target for how many assets to produce.

Consider a fictional team launching a tool that helps developers inspect webhook events. It commissions a four-minute creator demonstration: send a sample event, inspect a failed request, correct the example and retry it. The example is invented to show a planning approach. No audience response, conversion rate or campaign result is assumed.

The first draft of the brief says only: show the tool and link to the homepage. The team makes that request more useful by naming the task. The creator will explain how to distinguish a malformed payload from a delivery failure using synthetic test data. The demonstration has a concrete purpose, and the linked material can be checked against it.

A product engineer prepares a reference page with the same sample payload, setup steps, supported tool version and expected output. The page also explains the error used in the demonstration. Someone can follow it without opening the video. The creator can focus on why the diagnostic steps matter instead of reading every configuration detail aloud.

Before publication, a colleague who did not build the example tries the written instructions in a clean test environment. If a command fails or a prerequisite is missing, the team fixes the example or narrows the claim. The colleague checks the result against the expected output, rather than assuming that finishing the instructions means the example worked.

The creator then checks that the description points to the intended reference and that the visible result agrees with it. If a version changes later, the product owner has a named responsibility to update the page and flag any material difference from the demonstration. Keeping the two assets aligned is a more concrete assignment than asking either team to make the content more engaging.

Start with the developer’s task. Write one sentence describing what someone should be able to understand or attempt. For the fictional webhook campaign: “Identify why this sample request failed and reproduce the corrected request in a test environment.” A broad promise such as learning the whole product would require a different scope.

Specify what belongs in the video. Name the problem, the decision the creator will explain and the result they will show. Leave room for the creator’s own explanation while agreeing on claims that must match the product. A useful demonstration can acknowledge a prerequisite or a case the tool does not support.

Specify what belongs in the reference. List the setup, example inputs, expected output and relevant constraints. Give the exact destination, its owner and the version it covers. A homepage may be appropriate for a general introduction; a task-specific example needs somewhere that actually answers the task.

Agree on a handoff check. The creator checks the demonstration and description; the product owner checks the instructions and claims. Ask a colleague to attempt the example, then note what needs correction. The reusable deliverable is a shared brief with named responsibilities, not just two assets that happen to mention the same feature.

Finally, decide what feedback would change the material. Repeated questions about setup can prompt a clearer prerequisite section. An inaccurate example needs correction even if a video attracts attention. Views, reference visits and successful attempts describe different actions; use evidence available to your team without treating one as proof of another.

We used the complete public 2026 Stack Overflow Developer Survey file, containing 30,903 responses. The official methodology gives a response-extract window of June 23 through August 5, 2026, and reports usable responses from 169 countries. Respondents were recruited mainly through the Stack Overflow website, blog and opted-in email list. There are no survey weights for this analysis, and we report sample descriptions without population confidence intervals.^[1][2]

The eligibility steps were fixed before calculating the joint results. The code-creation question identified 23,234 people who said creating code was a primary part of their job. Of these, 11,527 said they had learned coding in the past year, 11,071 supplied learning methods, and 6,119 also supplied skill-influence answers. We excluded 344 who selected only not currently deciding and 95 who combined it with another influence, leaving 5,680. Blank answers were never recoded as nonselections.

The video variable is the selection “YouTube or video creators” in CommunitySkills. The documentation-use variable is the technical-documentation selection in LearnCodeHow. The Documentation influence selection remains a separate variable. We split the released multiple-choice answers at semicolons and checked the decoding against all 27 published categories across the two official tables; every category count and both response bases matched.^[1][3][4]

The archived question bank differs from the released results in some other learning-method options. The documentation option used here agrees across the instrument, schema and public results; the creator-influence option agrees too. We use the publisher’s public export and cannot independently reapply its unpublished response-quality flags. Both details are documented with the source instructions.

The mixed-answer check retains the 95 excluded mixed responses but still removes pure not-deciding answers. The paid-work check uses only employed or independent main employment statuses within the primary group. One otherwise eligible person naming video creators lacked a learning-method answer. Treating that one person as either using or not using documentation gives a logical range of 65.54% to 65.60%. This narrow range concerns that one missing item only. It does not address the 4,952 missing influence answers or voluntary survey selection.

Earlier research already describes people combining learning resources. Stack Overflow’s separate February 2026 pulse survey, for example, reported combinations involving AI and documentation. Our contribution is the reproducible joint between creator influence and documentation learning in this annual release. We do not treat the pulse survey as a comparison year or claim to have discovered that formats can complement one another.^[5]

The source database is made available by Stack Exchange Inc. under the Open Database License (ODbL) 1.0; individual contents are under the Database Contents License (DbCL) 1.0. The downloads provide aggregate results and the complete transformation method, with source links and license notices. They contain no respondent-level records or identifiers. Calculations and interpretation are by Influence IQ, without publisher endorsement.^[6][7]