Badges, Points and Safety Incidents: Reading the Gamification Evidence

How CodoraTech is funded: CodoraTech is supported by advertising and, in some articles, by affiliate links. Where an article contains affiliate links, we say so at the top of that article.

Two corporate gamification programmes are cited more than all the others combined: Salesforce Trailhead, and Walmart Logistics’ safety training with Axonify. Both are real, both are large, and the numbers attached to them are older and softer than their reputation suggests. Set against them is a body of peer-reviewed work that finds a moderate effect on what people learn, a smaller and less stable effect on what they do, and no measurable effect at all on intrinsic motivation.

Salesforce last published a badge milestone in February 2018

Trailhead is Salesforce’s free gamified learning platform: learners earn badges and points and climb through named ranks, with Ranger requiring 100 badges and 50,000 points. Salesforce’s corporate blog, published on 6 June 2023 and updated on 17 July 2023, states that more than 3 million people have skilled up with Trailhead and that the library holds more than 1,000 badges.

The only hard cumulative milestone Salesforce has ever published is older than that. On 14 February 2018 its newsroom announced 5 million badges earned alongside more than 22 million challenges completed. No updated badge total has been published since. Checked on 1 September 2026, the Trailhead homepage carries no badge count and no learner count, only a reference to more than 60 Salesforce certifications and a community numbering in the millions. Any current badge figure quoted elsewhere is not Salesforce’s, and the accurate statement is the narrow one: the last official milestone is 5,000,000 badges, dated 2018.

The career outcomes are a vendor survey of its own learners

The claims that make Trailhead sound transformative come from a 2020 Salesforce study summarised on the company’s own blog: more than 50 percent of learners said they gained skills leading to a promotion or a raise, one in five reported a salary increase above 20 percent, and one in three found new employment using Trailhead skills. This is a company surveying users of its own product, with no control group. No third-party assessment of Trailhead’s learning effectiveness could be located.

One outside datapoint exists and needs its own caveat. The Mason Frank Career Survey of more than 2,500 Salesforce professionals, reported by Salesforce Ben, found that the average respondent holds roughly 200 Trailhead badges and that 93 percent use Trailhead. That is a recruitment firm surveying its own candidate pool, so it measures the engaged end of the distribution, not the average employee.

Walmart’s 54 percent comes from a 2012 pilot, and both accounts trace to the same people

Walmart Logistics began deploying Axonify’s gamified microlearning platform for safety training in 2012, eventually reaching more than 75,000 associates across more than 150 US distribution centres, with employees answering safety questions for three to five minutes per shift. Axonify’s customer story is the source for every result quoted below: the vendor reporting on its own deployment.

Ken Woodlin, Walmart Logistics’ vice president of compliance, safety and asset protection, is quoted on that page with a hedge that is easy to skim past: “Feedback about the Axonify platform has been phenomenal, and we believe the process has been a significant contributing factor to our improved performance and engaged associate base.” A significant contributing factor is not a causal claim.

The same numbers reached the trade press. An unbylined article in Industrial Safety and Hygiene News of 26 February 2014 adds that the 2012 pilot covered 5,000 employees. It is sourced entirely to Walmart and Axonify representatives, with no independent verification, so it is not corroboration but the same claim in a second venue. What Axonify publishes remains the most widely cited set of hard numbers for corporate gamification at scale, and none of it has been independently replicated or audited in the twelve years since:

• A 54 percent decrease in recordable incidents at eight pilot distribution centres.
• Lost-time incidents down by more than 50 percent.
• 91 percent voluntary participation.
• 96 percent positive behaviour observations.
• Knowledge levels up to 15 percent higher on safety topics.

The meta-analysis: 0.49 cognitive, 0.36 motivational, 0.25 behavioural

The strongest independent evidence is Michael Sailer and Lisa Homner’s meta-analysis, “The Gamification of Learning”, published in Educational Psychology Review in 2020. It pools 44 studies with a combined sample of 4,883 participants.

The effect sizes separate by outcome type: cognitive outcomes g of 0.49, 95 percent confidence interval 0.30 to 0.69, across 19 studies; motivational outcomes 0.36, interval 0.18 to 0.54, across 16; behavioural outcomes 0.25, interval 0.04 to 0.46, across 9. The authors note that the motivational and behavioural effects were less stable than the cognitive ones.

That behavioural interval is the number a training buyer should sit with. Its lower bound of 0.04 almost touches zero, meaning the effect on what people actually do differently at work is small, fragile and built on the fewest studies. Gamification’s best-evidenced result is that people learn the material somewhat better, not that they behave differently afterwards.

Design moderates the effect, points do not

Sailer and Homner also tested what changed the size of the effect, and the answer was not the reward mechanics. Game fiction and social interaction significantly moderated behavioural outcomes, and competition augmented with collaboration was particularly effective. What distinguishes a programme that works is design, not the presence of a scoreboard.

That finding lines up with the field’s oldest criticism. Ian Bogost, writing in Game Developer on 3 May 2011, argued that gamification “proposes to replace real incentives with fictional ones” and proposed renaming the practice “exploitationware”, grouping it with malware and spyware, because organisations use it to extract loyalty through “counterfeit incentives that neither provide value nor require investment.” The term “pointsification” was coined by Margaret Robertson in 2010 to separate the misuse of isolated game elements from the real thing, her argument being that points are the least crucial aspect of games.

Anna-Sara Hellberg and Jonas Moll formalised the distinction in Frontiers in Education on 10 July 2023, defining pointsification as a focus solely on points, badges and rewards, without the game thinking that defines gamification. The same paper carries the most inconvenient finding in the literature, from Mekler and colleagues in 2013: points, levels and leaderboards “significantly increased performance” but “did not affect perceived autonomy, competence, or intrinsic motivation.” They work as progress indicators. They do not appear to work as motivators.

The standard citation for gamification eroding motivation over time is Hanus and Fox’s 2015 longitudinal study in Computers and Education. Its effect sizes, sample and duration are not reproduced here: the paper sits behind a publisher paywall, and the CodoraTech editorial team has verified the citation without reading the results.

What the evidence supports doing

Hellberg and Moll are not opponents of points. Their argument is that pointsification implemented with care can raise achievement and engagement and can correct equity problems in group work. The distinction they draw is between a reward layer bolted onto unchanged content and a design that uses game thinking to change how the work is structured. On the published evidence, only the second reliably earns its budget. Four things follow from the research above:

1. Treat points and badges as progress indicators, which is what Mekler and colleagues measured, not as a motivation system.
2. Build in the moderators Sailer and Homner identified: game fiction, social interaction, and competition paired with collaboration rather than competition alone.
3. Apply Hellberg and Moll’s conditions: clear criteria, bidirectional points that can be lost as well as earned, and individual accountability inside group tasks.
4. Set expectations by outcome type: the evidence for improved knowledge is considerably stronger than the evidence for changed behaviour.

Sources: Educational Psychology Review, via ERIC · Frontiers in Education · Game Developer · Salesforce · Axonify