Vendor conversations go better when you arrive with numbers instead of complaints. Most teams do the opposite, describing the process as slow and letting the vendor define what slow means. Everything worth knowing before you scope workflow optimization services can be measured in about a week with no outside help.
Five numbers below, each pointing at a different intervention. The thresholds attached to them are working rules of thumb from operational practice rather than findings from any study, so treat them as starting points for judgment rather than as fixed lines.
Number One: Elapsed Time Divided by Active Time
Take five recent cases. Record total time from request to completion, then total time anyone was actively working on them. Divide the first by the second.
Under 3: the process is genuinely busy. Task automation will help.
Between 3 and 10: mixed. Some coordination waste, some real work.
Above 10: you have a waiting problem, not a work problem. Automating the active portion changes little, because the active portion is not where the time goes.
This is the most useful number on the list and the one most likely to contradict what the team believes. Processes that feel busy are frequently mostly idle.
Number Two: How Many Times One Record Gets Typed by a Human
Follow a single order, ticket, or case through its full life. Count manual entries.
One: fine.
Two: integration candidate. Every retype is an error source, and error handling usually costs more than the typing did.
Three or more: the real problem is that systems do not exchange data. Buying automation without addressing that means automating the retyping.
Number Three: Percentage of Cases Following the Standard Path
Pull thirty cases. Count how many went exactly as documented.
Above 85 percent: rules will hold. Deterministic automation is cheap and maintainable.
Between 60 and 85 percent: exception handling needs designing in from the start rather than added later.
Below 60 percent: the documented process is not the real process. Find out what the real one is first, because a rulebook written against fiction breaks continuously.
Most teams guess above 90 and measure closer to 65. The gap is usually informal workarounds that two or three people know about.
Number Four: Hours Per Person Per Week Spent Assembling Information
Count time spent gathering, reconciling, or compiling data rather than acting on it. Include the Monday morning spreadsheet, the reconciliation meeting, and anyone chasing status.
Under 2 hours per person per week: background noise. Not worth a project on its own.
Between 2 and 5: a real recurring cost, and usually the strongest business case available because it has names attached to it.
Above 5: someone’s role has quietly become data assembly. At this level the work is a data movement problem more than a workflow problem, which is why it often gets scoped alongside business intelligence and analytics services rather than as pure automation. Notionmind lists data integration and centralization as separate from dashboard work, reflecting that connecting sources and deciding what the data means are different jobs.
If a figure requires someone to prepare it, its definition lives in their head rather than in any system.
Number Five: Percentage of Cases Where Someone Bypassed the Process
The forward looking number, and the one nobody records. Count manual overrides, side channel approvals, and decisions made before the system caught up.
Under 5 percent: the process matches reality.
Between 5 and 15 percent: friction is building. Worth investigating before you automate, since you will be enforcing whatever you build.
Above 15 percent: people are routing around the process for reasons that probably make sense to them. Enforcing the current design faster creates friction rather than removing it.
What These Numbers Look Like in a Real Situation
Notionmind’s published Granger Delivery engagement is a clean illustration of several of these firing at once. The documented starting position included manual order taking, inconsistent pricing calculations, no centralized order tracking, time consuming quote generation, and missing performance analytics.
Read against the five numbers, that is a high assembly figure, a high retype count, and almost certainly a high exception rate, since pricing calculated by hand varies by whoever calculates it. What was built covered automated price calculation, centralized order management, and performance tracking, which maps to those specific problems rather than to automation generally.
The point is not the outcome. It is that the diagnosis was legible before the build started.
Which One to Fix First
Once you have all five, the sequencing is fairly consistent:
- Number three, if it is below 60. Nothing else means anything until you know what the actual process is.
- Number two, if it is two or more. Data movement before task automation. Most automation failures are integration failures under another name.
- Number five, if it is above 15 percent. Redesign before enforcement.
- Number one, if the ratio is high. Coordination, routing, and visibility work.
- Number four. Usually improves as a byproduct of the four above, which is why chasing it directly is often wasted effort.
Notionmind’s stated delivery order follows a similar logic, running workflow assessment and process redesign ahead of automation and integration, with most assessments taking two to four weeks.
Bringing These to a Vendor Conversation
Two things change immediately. Proposals become comparable, because you can ask each firm which of your five numbers they expect to move and by how much. And you get a defensible baseline, which is the difference between proving a result later and arguing about impressions.
Note the date and method beside each figure. Context shifts and an undated baseline is arguable. Ten seconds of work now removes a whole category of disagreement in six months.
