Designing an Explainable Golf Performance Advisor

GHIN gave me detailed statistics about my golf performance, organized primarily by individual rounds and selected subsets of my history. That helped me review what happened. It did not help me interpret it — I could see my numbers, but I could not tell whether they were good, how they compared with stronger golfers, or what I should work on next.

I began with one target user: myself. That gave me direct access to the problem and allowed me to iterate quickly, while limiting the range of perspectives represented in the first version. The first release placed my results within visual benchmarks across club, age, and handicap. Once I could see where I stood, I expanded the product to answer a more actionable question: where should I focus my limited practice time? The resulting Game Improvement Finder compares several areas of performance, identifies relative gaps, and ranks potential improvement opportunities.

Outcome — A functioning product prototype that places personal performance within relevant benchmarks, shows what stronger golfers achieve, and translates the gaps into a ranked practice priority.

ShotSense evolved from showing where my performance stood to identifying where my practice time could have the greatest value.

The Gap in GHIN

Specifically, GHIN data left three things unanswered:

• 1

  No benchmarks

  I could not compare my results with similar or lower-handicap golfers.

• 2

  No progression path

  I could not see the performance required to reach the next level.

• 3

  No prioritization

  I could not tell which gap deserved my limited practice time.

I could see my numbers, but I could not tell whether they were good, what better golfers were achieving, or what I should work on next.

Current GHIN Round Data

GHIN showed what happened in the round, but not how the results compared, what better golfers achieved, or which gap mattered most.

Making a Number Meaningful

My hypothesis: golf-performance data becomes more useful when it’s placed within relevant visual benchmarks. A single round score didn’t tell me whether it was a strong result for my handicap or an average one, so I built benchmark views that plot a result against club, age, and handicap cohorts — including what stronger golfers in the same cohort achieve.

Benchmark View

1 2 3

The benchmark view answered two of the three original questions: is this good for a golfer like me, and what are better golfers achieving.

Benchmark View

• 1

  Percentile range

  10th, 50th, and 90th percentile lines show a realistic range, not one blended average.

• 2

  Cohort comparison

  Filter by handicap group — 0–5, 6–15, 16–25, 26+ — instead of one blended average.

• 3

  Stronger golfers

  The upper band shows what better golfers in the same cohort are achieving.

From Context to Action

Benchmarking showed where my performance stood, but it did not tell me where to focus. Knowing the size of a gap was useful; deciding which gap deserved my next hour of practice was more valuable.

I addressed that second problem with the Game Improvement Finder. It compares a focused set of performance inputs against relevant benchmarks, identifies the largest relative gaps, and ranks the resulting opportunities into one clearer practice priority.

Recommendation Inputs

A focused set of inputs is enough to place a golfer in a cohort and compare each stat against that cohort’s average.

Recommendation Inputs

User Inputs

↓

Benchmark Comparison

↓

Relative Gaps

↓

Ranked Recommendation

A percentile alone did not indicate whether a performance gap was worth prioritizing. I paired each comparison with a plain-language interpretation and ranked the relative gaps, reducing several performance measures to one recommended focus area.

Ranked Recommendation Result

1 2 3 4

Raw statistics → cohort comparison → prioritized action. Several separate measurements become one recommended place to spend practice time.

Ranked Recommendation Result

• 1

  Rank badge

  “Your #1 Opportunity” surfaces the single highest-priority gap first, not a full list to sort through.

• 2

  Plain-language interpretation

  What the percentile means and what to actually practice.

• 3

  You vs. Your Cohort

  The personal result placed directly next to the figure it’s being judged against.

• 4

  Relative gap badge

  The distance between the two, e.g. “-5% vs cohort.”

Outcome, Trust, and Next Steps

How I Made Modeled Data Trustworthy

Public golf-performance data is incomplete across age, handicap, club, and performance category. I combined published benchmarks with modeled estimates, then made the source, assumptions, and limitations visible so the output could be useful without appearing more certain than it was.

GHIN vs. ShotSense