Benchmarking: Size the Prize Before You Pilot a Change

Benchmarking compares current process performance against internal benchmarks, competitors, or best-in-class to set improvement targets and uncover proven practices. That is the whole job. It does not diagnose. It does not redesign. It tells you how far you are from where good looks like, and it tells you which gaps are worth the effort of closing.

In RDMAICS, this sits in the Improve phase for a reason. By the time you open it, you have already recognised a problem, defined its boundary, measured a baseline and analysed the cause. You know what is going wrong and roughly why. What you do not yet know is how much better the process could reasonably get, or which of several candidate changes deserves a pilot slot first.

That is the gap Benchmarking fills. A diagnosed cause is not yet a change worth piloting. It becomes one when you can say, in a sentence a sponsor will accept, how big the prize is and where the evidence for that number came from. Benchmarking gives you that sentence.

Reach for it when you have more improvement ideas than capacity, when a sponsor asks what good looks like, or when a target has been set by assertion rather than by evidence and you want to test whether it is achievable.

Everything below uses the sample data built into the tool. The case is fictional. It describes an accounts payable function that processes supplier invoices, and it tracks five metrics that together describe how well that function performs.

The five metrics are invoice cycle time in hours, first-pass match rate as a percentage, error rate as a percentage, cost per invoice in dollars, and on-time payment as a percentage. Those five were not picked at random. They cover speed, quality, cost and service, which is the minimum spread you need before a benchmark comparison tells you anything useful about trade-offs.

Imagine the Analyse phase has already landed. The cause looks like weak three-way matching at intake, which forces manual intervention downstream. That is a diagnosis. It is not yet a business case. The question now is whether fixing intake matching is worth a pilot, and what result you should expect if it works.

The worksheet asks for four things per row. The metric name, the current value, an internal benchmark, and an external best-in-class value. You also record a source for the comparison. The tool calculates the gap column for you.

In the sample case, invoice cycle time is 46.2 hours today, against an internal benchmark of 30 and a best-in-class figure of 18. First-pass match rate is 65 percent, against 85 internally and 95 at best-in-class. Error rate is 12 percent, against 5 and 2. Cost per invoice is 8.40 dollars, against 5.00 and 2.50. On-time payment is 78 percent, against 92 and 98.

The two-benchmark structure matters more than it first appears. The internal benchmark is what another team, site or period inside your own organisation already achieves. It is proof that the result is reachable with your systems, your people and your constraints. The external best-in-class figure is what the wider field achieves. It is proof that the internal benchmark is not the ceiling.

The source column is the discipline. Each row in the sample names where its comparison came from: a published accounts payable benchmark for cycle time and cost, an industry survey for first-pass match rate, a research firm's data for error rate, and industry data for on-time payment. Record the source while you enter the number. If you cannot name one, you are entering an opinion, and opinions do not survive a steering committee.

The tool reports five metrics compared and an average absolute gap of 18.82. It then draws each metric as three bars, current, internal benchmark and best-in-class, on a shared scale, and prints a gap in both absolute and percentage terms beneath each one.

The percentage view is where the ranking changes. On raw size, invoice cycle time looks like the headline, with a gap of 28.2 hours. But as a percentage the picture reorders. Error rate shows a gap of minus 10 and minus 83.3 percent. Cost per invoice shows minus 5.9 and minus 70.2 percent. Invoice cycle time shows minus 28.2 and minus 61 percent. First-pass match rate shows 30 and 46.2 percent. On-time payment shows 20 and 25.6 percent.

Read the signs carefully. A negative gap means the current value is above best-in-class and needs to come down, which is correct for cycle time, error rate and cost. A positive gap means the current value is below best-in-class and needs to rise, which is correct for match rate and on-time payment.

The average absolute gap of 18.82 is a single summary number across mixed units. Treat it as a tracking figure. If you rerun this worksheet after a pilot and that number falls, the overall distance to best-in-class has narrowed. Do not treat it as a performance score in its own right.

The tool also offers an AI review that explains the gap analysis in plain language. Use it to sanity check your own reading, not to replace it.

The first decision is target setting. You now have two candidate targets for every metric, and a defensible reason to pick between them. For a first pilot, the internal benchmark is usually the right target. Moving first-pass match rate from 65 to 85 is a known-achievable step. Moving it to 95 is a programme, not a pilot.

The second decision is sequencing. First-pass match rate carries a 30 percentage point gap, the largest of the three percentage-point gaps in the set. It is also the metric most directly tied to the diagnosed cause at intake. That coincidence is the argument for piloting there first. If matching improves, you would expect error rate, cycle time and cost to move as consequences rather than as separate projects.

The third decision is scope control. The gap table makes it obvious what you are not doing. On-time payment has the smallest percentage gap at 25.6 percent. It stays as a monitored measure, not a workstream.

The fourth decision is how you will know the pilot worked. You have the before values, the targets and the units already written down. The post-pilot review is a rerun of the same worksheet.

The most frequent error is benchmarking against a figure nobody can trace. A number without a source gets challenged in the first review and the whole analysis stalls. The sample data names a source on every row. Copy that habit.

The second error is comparing processes that are not comparable. A cost per invoice of 2.50 at best-in-class may reflect a different invoice mix, a different level of automation or a different scope of work. Before you adopt it as a target, ask what is included in the measure on both sides.

The third error is chasing the largest absolute gap without checking the percentage. The 28.2 hour cycle time gap looks dominant until you see that error rate carries an 83.3 percent gap. Units are not comparable across rows. Percentages are closer to it.

The fourth error is setting every target at best-in-class at once. You would be committing to move cost per invoice from 8.40 to 2.50 and error rate from 12 to 2 in the same breath. That is five simultaneous transformations, and pilots do not survive that load.

The fifth error is treating the average absolute gap of 18.82 as a key performance indicator. It mixes hours, dollars and percentage points. It is useful for tracking direction over time and nothing more.

Benchmarking tells you the size of a gap. It does not tell you the cause of that gap, and it will not tell you how to close it. The sample shows a first-pass match rate 30 percentage points below best-in-class. It is silent on whether that is supplier data quality, purchase order discipline, tolerance settings or something else. That answer comes from the Analyse phase, before you open this tool.

It will not validate your comparison data. If the external figure is from a different industry, a different invoice volume or a different definition of cycle time, the tool will still calculate a tidy gap. The quality of the output depends entirely on the quality of the source you typed in.

It will not price the change. There is no effort estimate, no implementation cost and no timeline in the output. A gap of 5.90 dollars per invoice is a prize, not a net benefit. You still need to work out what capturing it costs.

It will not tell you whether the gaps interact. Cycle time, error rate and cost in this sample are almost certainly connected, but the tool treats each row independently. It cannot tell you that closing one gap will partly close another, which means the five gaps added together will overstate the total available gain.

It will also not detect a missing metric. If something important is absent from your worksheet, the result looks complete anyway. Five rows compared is five rows compared, not a full picture.

Use it for what it does well. Turn a diagnosed cause into a sized, sourced, defensible target, pick the one worth piloting, and move on to the pilot with the before values already recorded.