Internet-based training: formats and metrics

Most advice on internet-based training starts with the platform. Pick the LMS, upload the videos, chase completions. I'd start at the other end. Write down what you'll measure two to four weeks after the course ends, then choose a format that can produce that result. If you can't name the changed behaviour, you don't have a training problem yet. You have a content pile.

What internet-based training covers today

Internet-based training is learning delivered through a browser or app, live or on demand, instead of in a room.

That covers far more than a self-paced course. A typical onboarding program might combine a short module on policy, a live session for questions, a scenario where the new hire handles a difficult customer, and a document pack they can reopen before their first call. Learners reach those pieces through an LMS login, an emailed link, or an embed inside a tool they already use.

Formats do different jobs

Format	How learners use it	Best for	Main weakness
Self-paced e-learning module	Learner works through screens, quizzes and video at their own speed	Compliance, product knowledge, onboarding	Low completion without reminders
Live virtual classroom	Instructor teaches a scheduled session with chat, polls and breakout rooms	Skills that need questions and live feedback	Fixed times and harder to scale
Webinar	One presenter speaks to a large audience, often recorded for replay	Announcements, awareness and product updates	Little practice and mostly passive viewers
Blended program	Online modules before or after in-person sessions	Complex skills with a hands-on component	Coordination overhead across two formats
Micro-lesson	Short video or single screen covering one task	Just-in-time support and refreshers	Too narrow to stand alone as a full course
Scenario simulation	Learner makes choices and sees the consequences	Decision-making, sales conversations, safety calls	Expensive to build well
Document with video walkthrough	Learner reads a PDF or deck while narration explains each page	Proposals, policies, contracts, onboarding packs	Limited assessment unless paired with a quiz

The format matters, but it matters less than what learners do inside it. A polished webinar where nobody answers a question, makes a decision, or tries the task is still a lecture. It's just a lecture with a chat box.

If learners need to practice a skill, a single video won't do the job. Pick a format that gives them room to practice and get corrected.

Say you're training support reps on refunds. A module can teach the policy. A scenario can make them choose what to say when a customer pushes back. A live session can then deal with the calls they got wrong. That's one program using three formats for three different jobs, and that's normal. The mistake is forcing every job into whichever format you bought first.

What the research says about online versus classroom results

Online training is not weaker than classroom instruction by nature. The relevant question is what learners can demonstrate afterward, not which channel delivered the material.

The best-known test is the Sitzmann, Kraiger, Stewart and Wisher meta-analysis of web-based and classroom instruction, which covers 71 effect sizes and 10,910 learners. It found web-based training about 6% more effective than classroom instruction for declarative knowledge, but that aggregate is shaped by study design and self-selection. In the random-assignment subset, classroom instruction did better, so don't read it as a general edge for online delivery. Standalone web-based and classroom instruction came out near-equivalent for procedural knowledge, while blended formats beat classroom instruction. Learners were equally satisfied with both.

Declarative knowledge means knowing a refund policy. Procedural knowledge means handling a refund call when a customer is angry and the system is slow. Online learning suits facts, concepts and policies that people can review at their own pace. Handling a difficult call takes practice, and standalone online delivery gives you no edge there; blending it with live practice does. Measure both what learners can recall and whether they can apply it.

The method explains the gap

When both formats used the same teaching methods, the same paper found the difference in declarative knowledge disappeared. Interactivity, feedback, practice and assessment design mattered more than the channel. A format can support learning, but it can't make up for weak practice or assessment.

Online does not automatically save time

Moving a course online is no reason to assume learners will finish sooner. Track completion time and assessment performance together to see whether a shorter course still prepares people to do the work.

A course that takes twenty minutes but transfers poorly is short, not efficient.

Instructional design rules that move results

The most common mistake I see in internet-based training is simple. A team takes a classroom deck, turns each slide into a web page, adds a next button and calls it a course. Learners click through, pass a quiz they could've passed beforehand, and nothing changes at work.

A course that's only watched is like a gym where people only watch others lift. Attendance looks great. Nobody gets stronger.

Five rules that work together

Break lessons into short clips: give each clip one concept, one procedure or one decision, so a learner can return to the exact explanation they need.
Put a retrieval question straight after each clip: make learners recall the point before moving on, which is the core of retrieval practice.
Add a scenario decision with consequences: a customer objection, a compliance call, a pricing choice where the wrong answer costs something visible.
Give immediate corrective feedback: explain why the answer was wrong, not just that it was wrong, while the decision is still fresh.
End with a practice task on actual work material: the tool, form or document the learner will use on the job, not a screenshot of it.

On length, an MIT study of 6.9 million video-watching sessions, summarised by Quality Matters, recommended keeping instructional videos shorter than six minutes.

Short is not the same as effective

Don't treat that as a magic number. A randomised study comparing one 14-minute video with the same content split into three segments of four to five minutes found no significant difference in learning, interaction or perceptions of the video. Participants who multitasked, other than texting, performed worse on the learning measure.

Segmenting alone doesn't guarantee learning. That's why each clip needs the question, the scenario and the practice task underneath it. Short clips without questions just give you passive viewing in smaller pieces.

Run one more check before you ship. Captions, transcripts, mobile playback and a language the learner actually reads. Putting a course online doesn't solve reach by itself. If a warehouse worker can't play audio on the floor or a new hire is reading in a second language, your completion data will tell you a story about access, not ability.

Choosing delivery tools without buying a stack you do not use

Most buying mistakes happen before anyone opens a pricing page. Teams pick products first, then work out what they need to track, then discover the tool doesn't track it. Start from the measurement and the stack gets much smaller.

Tool category	Job it does	Example tools	Watch for
Learning management system	Hosts courses, assigns them and records completions	Moodle, TalentLMS, Docebo	Admin overhead, overkill for a handful of learners
Authoring tool	Builds interactive modules with branching and quizzes	Articulate Storyline, Adobe Captivate	Steep learning curve and per-seat licensing
Video recording and hosting	Records and serves lesson clips	Loom, Vimeo, unlisted YouTube videos	Little built-in tracking beyond views
Virtual classroom	Runs live sessions with chat, polls and breakout rooms	Zoom, Microsoft Teams	Recordings need separate management
Document and video sharing	Pairs a PDF or deck with narrated walkthroughs and tracks page-level views	LiveDocument	Suited to reference and onboarding content, not graded courses

Be honest about the trade-offs

An LMS earns its keep when you need certificates, expiry dates and compliance records that an auditor can check. A shared link won't do that job. But a team training twelve people on a new process may never need an LMS at all, and buying one means someone now administers enrolments, passwords and reports for a course that runs twice a year.

Authoring tools are the other quiet cost. Storyline and Captivate can build proper branching scenarios, but they need a person who knows them, and per-seat licensing means that person becomes a bottleneck. If your content changes every quarter, that bottleneck matters more than the feature list.

Practical rule: name the one thing you need to track, then buy the tool that tracks it. Stop there.

Metrics that show real engagement beyond completion

A completion rate tells you someone reached the last screen. It tells you nothing about whether they learned anything, yet it is the number many organisations report to leadership anyway. That gap is where bad training survives for years.

Metric	What it tells you	Where it misleads
Completion rate	Learner reached the last screen	Says nothing about what they absorbed
Time on module	How long the learner stayed on the lesson	Long time can mean confusion, not engagement
Video watch-through	Share of a clip the learner actually played	Passive playback can look like attention
First-attempt quiz score	Recall before any retry	Easy questions inflate the score
Replays of a specific clip	Where learners got stuck	Replays also happen for reference, so read them alongside the quiz
Delayed retest two to four weeks later	Whether the knowledge stuck	Takes effort to run, so teams skip it
Workplace measure such as error rate or call quality	Whether behaviour changed on the job	Needs a baseline and a manager to collect it

Knowledge gains overstate impact

The meta-analysis of internet-based public health training found a pooled standardised mean difference of 1.92 (95% CI 1.05 to 2.78) against no intervention. It's an exploratory analysis with substantial heterogeneity, so don't generalise that figure to all internet-based training. Against face-to-face formats it was 0.39 (95% CI -0.06 to 0.83; p = 0.09), which wasn't statistically conclusive (eight studies, I² = 88%).

Read that as a warning about your comparison. Beating no training is easy. Beating a decent instructor-led session with a thin online course isn't guaranteed at all. So evaluate with pre- and post-tests plus delayed transfer or workplace-performance measures, because knowledge gains alone may overstate operational impact.

Document-based training gives you a different signal. Page-level analytics show time on each page and whether someone came back. If a rep reopens the pricing page the morning of a call, that may indicate they're preparing, and it's a good question for a coaching chat. It isn't proof they understood it. There's more on reading those signals in PDF analytics and what page-level data actually shows.

The delayed measure is the one most teams skip. Send a five-question retest three weeks later as one retention signal. It shows what stuck, not whether behaviour changed, so keep the workplace measure as your test of operational impact. It's cheap, it's mildly annoying, and it will embarrass at least one course you thought was fine.

Micro-training works best when the video sits next to the page

A new sales rep opens a six-page onboarding PDF about pricing, discount rules, and the ideal customer profile. Between calls, she skims it, reaches page four without the context that makes the information useful, then closes the tab. The document is available. The reading may not happen, and opening a file is not evidence of understanding.

Build the pack page by page

Outline the pages first: identify the decision each page supports, such as when a rep can discount without approval.
Record one clip per page: keep each narrated walkthrough to 60 to 90 seconds. Re-record a clip when the pricing changes.
Turn on email capture: use it when you need to identify who viewed the pack, rather than counting anonymous opens.
Check page-level time after week one: follow up on pages people skipped instead of resending the entire pack.

A short video beside its relevant page gives learners context as they read. I build LiveDocument, so I'm biased, but this is exactly the job it does: a clip sits beside the page it explains, so a rep can hear why a discount rule exists while looking at the rule. Here's how to go about adding video to PDF pages.

The manager can see which pages were opened and how long each one held attention. Those signals show how the pack was used, not whether the learner understood it. A return visit may mean the document works as a useful reference, but it doesn't prove the rep can apply the rule.

A document pack doesn't test understanding. Pair it with a short check, such as three scenario questions in a form, or you are measuring reading and calling it learning.

Pick your format and measure transfer before you build

Write the one behaviour you want changed: one sentence a manager could observe, such as reps quoting the discount limit without checking.
Choose the format by skill type: knowledge and policy suit online modules and documents, while skills and judgment need practice, simulation and feedback.
Pick the smallest tool set that tracks your measure: the behaviour from step one can be measured in a manager review, CRM or QA system rather than the training tool itself, but if the tool can't hand that data off, it isn't earning its place.
Pilot with learners who match the actual audience: not the friendly team down the hall who'll be polite about it.
Record a baseline before the course and a follow-up two to four weeks after: the baseline is the step that teams often skip, and skipping it makes every result impossible to read.
Roll out only if the follow-up shows the behaviour changed: if it didn't, fix the practice and feedback before you scale the problem.

Build one module or one document pack this month and track it against that baseline.