All six human skills also moved up at a smaller customer

A 53-person Thai professional services firm has been on Happily long enough to accumulate sustained rating data for 22 of its employees. Across that cohort, mean ratings on the six human skills (also called soft skills, or Power Skills in Happily) rose from 3.14 to 4.13 over a median 12 months. 93% of person-skill cells moved in the positive direction. This is one customer's slice of the cross-customer longitudinal study, corroborating the larger Customer A case.

The companion longitudinal study reported a cross-customer average rating gain of +0.29 points over a similar window, with heavier users gaining three to six times more than lighter users. That is a useful aggregate, but it averages across 80 organisations with very different cultures, sizes, and engagement levels. To see what happens at one organisation, this article zooms in on a smaller customer (Customer B).

The customer is a 53-person Thai professional services firm. Twenty-two of those employees, or 42% of the workforce, were active on Happily long enough to clear the cohort bar for the cross-customer study: at least six skill ratings on the same skill, over at least 180 days, with at least two ratings in each half of their window. That is high coverage for a small organisation, and it is what makes the case study possible at all.

What the cohort showed, over a median 12 months: mean ratings on every one of the six human skills rose. The cohort-wide average climbed from 3.14 to 4.13 on a 0.5 to 5 scale, a gain of +1.00. 93% of person-skill cells moved in the positive direction, and 78% gained at least half a point. The size of the move is striking, but the more important fact is the consistency: every skill, the same direction.

Empathy ratings at Customer B rose from 2.73 to 4.03 (+1.30). All ten Empathy cells were positive. It is the skill with the largest gain in the cohort.

Why this matters

The cross-customer study is the broader prior: across many organisations, ratings rise with use, and rise more with engagement. A single-customer case study cannot reproduce the statistical power of the full cohort, but it can show what the pattern looks like when an organisation actually leans into the practice. Customer B is one example. The shape of the result, all six skills moving up, applies whether the magnitudes generalise or not.

Methodology

Organisation

A smaller Thai professional services firm, 53 employees. Anonymized at the customer's request; referred to as Customer B throughout.

Cohort

22 individuals (42% of the workforce) who accumulated ≥6 ratings on the same skill over ≥180 days, with ≥2 ratings in each half of their window.

Time window

Per-person window spans first to last rating. Median 364 days (12.1 months). Minimum 189 days. Maximum 464 days.

Outcome

Per-skill mean rating in the first half of each person's window vs the second half. 86 person-skill cells total across the six skills.

Comparison

Cross-customer baseline from the companion longitudinal study: mean Δ = +0.29 across 2,630 individuals at 80 organisations.

What this case study does not do

Run the within-company regression-to-mean correction (n is too small to stratify by both baseline quartile and engagement quartile inside Customer B). The RTM caveats from the cross-customer study apply here.

The shape of the result

The cleanest view is per-skill. Every one of the six human skills shows a positive average shift between the first half and the second half of the cohort's analysis window.

Figure 1 Each pair of dots represents one skill's journey at Customer B. Light dot = first-half mean rating, dark dot = second-half. All six pairs move right, in the same direction, with sizes ranging from +0.76 (Optimism, which started the highest) to +1.30 (Empathy, which started the lowest).

Two patterns worth noting. The first is that Optimism shows the smallest gain (+0.76) and started the highest (3.61). The second is that Empathy shows the largest gain (+1.30) and started the lowest (2.73). That is exactly the pattern the cross-customer analysis predicts: skills with more room to grow gain more. Regression to the mean is partly responsible, and we cannot disentangle it with n=22 at the company level. The cross-customer stratified result remains the cleaner test of the dose effect.

How Customer B compares to the cross-customer baseline

Three numbers put Customer B's result against the broader cohort.

Customer B vs cross-customer baseline (2,630 individuals across 80 organisations)
Metric	Customer B	Cross-customer	Difference
Mean rating change (Δ)	+1.00	+0.29	+0.71
% of person-skill cells positive	93%	67%	+26 pp
% cells with strong positive change (≥0.5)	78%	41%	+37 pp
Mean span of analysis (days)	364	556	−192
Median response-active months	17	19	−2

Customer B's headline numbers run substantially above the cross-customer averages on every measure of rating change, despite a shorter mean analysis span. Engagement (measured by response-active months) is comparable, sitting just below the cross-customer median. The picture is one organisation where the rating-shift signal is bigger than the broader prior would suggest.

What this case study shows

This case study is a single-organisation slice of an observational study. It does three things:

Show what the rating trajectory looks like in concrete terms, for one customer, in one window.
Confirm that the pattern we see across customers (all six skills moving up, with room-to-grow effects) also holds within an organisation that engaged sustainedly.
Provide one credible reference point for what is achievable at a small organisation with strong engagement.

Read it for what it is: one customer's engaged cohort, 22 of 53 employees, in one window. Some of the +1.00 gain is the room-to-grow effect the cross-customer analysis quantifies; applying that correction implies a coaching-related gain of roughly +0.60 to +0.70, still a substantial move. Other customers moved less, so this magnitude is an example of what strong engagement can produce, not a number every small team will hit. The dose-response pattern from the cross-customer study is the strongest evidence that the loop contributes to the movement, and a randomised comparison would measure that contribution directly.

Limitations

Small cohort (n=22 people, 86 cells). Confidence intervals around per-skill estimates are wider than they look. The smallest skill cell (Empathy, n=10) carries the largest reported gain, which deserves caution even though all ten cells moved positively.
The within-company regression-to-mean correction we ran in the cross-customer study cannot be replicated cleanly with n=22 inside Customer B. The cross-customer correction is the best evidence that the dose-response signal survives RTM; we extrapolate that finding to Customer B rather than re-derive it.
Selection. The 22 individuals here were engaged enough to clear the cohort bar. 31 other employees at Customer B were not. This article describes the engaged cohort, not the company average.
Observational design. As with the cross-customer study, the dose-response comparison is the strongest lever available short of a randomised trial.
The AI rating model that produces these ratings has evolved over the analysis window. Some of the observed gain may reflect model drift rather than skill change.
One time window. Customer B's future trajectory may differ. The result is a snapshot, not a forecast.

References

Happily Research (2026). Human skills rise with practice: a 2,630-person study. Cross-customer longitudinal analysis, May 2026. The broader cohort context for this case study.
Happily Research (2026). The six skills that show up where engagement does. Framework defense for the six Power Skills.
Campbell, D. T., & Kenny, D. A. (1999). A Primer on Regression Artifacts. Guilford Press.
Happily Research (2026). Customer B cohort dataset (anonymized smaller Thai professional services firm). Internal analysis, 22 individuals, 86 person-skill cells, May 2026.