Internal tools: metrics that should trigger action

Most internal tools roadmaps are built from complaints, not evidence. Someone in finance is tired of pasting Tally exports into a spreadsheet, someone in ops is chasing order status on WhatsApp, and the louder of the two gets the next sprint. The backlog grows, nobody can rank it, and money goes to the workflow that shouted hardest rather than the one costing the most. If you own operations, finance or engineering at a 50-250 person company, fixing that usually takes four numbers and a set of agreed trigger lines.

Pick four metrics, not forty

Four metrics are enough to run an internal tools roadmap: cycle time, rework rate, queue ageing and cost per transaction. Cycle time is the elapsed time from request to completion for one core workflow, read off actual timestamps rather than anyone's estimate. Rework rate is the share of items that come back for correction: invoices re-raised, orders re-keyed, tickets reopened. Queue ageing is how long the oldest open item has been sitting. Cost per transaction is fully loaded staff hours divided by volume handled. Everything else — adoption scores, satisfaction surveys, ticket counts — is noise until these four are stable and trusted.

Set the trigger line before you discuss tools

A threshold only works if it is written down and agreed by the process owner before anyone opens a design tool. Sit with the person who runs the workflow, pull 8-12 weeks of real data to set the baseline, then define the trigger as a percentage above that baseline sustained for a fixed period. Two weeks is typical for daily workflows; one full cycle for monthly ones like GST filing or payroll. Keep each line to a single sentence: 'If invoice rework exceeds 5% for two consecutive weeks, we scope a fix.' Written this way, the roadmap ranks itself, and arguments about whose pain is worse turn into arithmetic.

The thresholds that justify real work

In our experience, five conditions justify building or buying a tool; anything softer is usually a process fix. Here is the filter we start with:

• Cycle time sits 20-25% above the agreed baseline for two consecutive weeks.
• Rework rate crosses 5% of items, or 2% for anything touching money or statutory filing.
• The oldest item in a queue passes two full business days with no owner able to explain why.
• One person is the only one who can run a critical workflow — key-person risk is a trigger regardless of volume.
• Fully loaded cost per transaction is high enough that removing 30-40% of it repays a build in under a year.

Below those lines, the honest answer is a checklist, a template or a clearer owner. Saying so out loud makes the roadmap shorter and far more credible.

Measure with the systems you already run

You can compute all four from tools you already pay for, without buying an analytics platform. Tally or Zoho Books gives you invoice timestamps and re-raised entries; your CRM (a self-hosted one is fine — we run our own) gives stage ageing; WhatsApp Business and your email platform show response gaps; a helpdesk gives reopen counts. Extract one CSV a month and keep a single sheet. Under the DPDP Act, measure counts and durations rather than individual customer records, and restrict who can open that sheet. That removes most of the compliance questions before they start.

Worked example: a 1,400-invoice reconciliation queue

Here is how the numbers become a decision, using an illustrative mid-market distributor. Finance processes about 1,400 purchase invoices a month, and three people spend roughly 90 minutes each per day matching invoices to POs in spreadsheets — around 99 person-hours a month. At a fully loaded ₹450 an hour, that is about ₹44,500 a month, or ₹31 per invoice. Rework sits at 7%, well past the 2% line for money-touching work, and the oldest item is four days old. A small matching tool that reads Tally exports and flags mismatches is roughly 5-6 weeks of work; if it removes 60% of manual matching, it pays back inside nine months. The decision is build, in the same way these projects were scoped. Had rework been 1.5% and cycle time flat, the right answer would have been a checklist and a weekly review.

Review the thresholds every quarter

Thresholds decay, so review them quarterly rather than treating them as permanent. A line that was correct in January is often wrong in July: volumes shift with the festive quarter and the March year-end, a shipped tool removes the bottleneck it was measuring, and headcount changes move cost per transaction. Retire any metric that has not triggered in two quarters; it is consuming attention and returning nothing. Keep the review to 45 minutes with the process owner and one person from finance. If a threshold has been crossed three quarters running and nothing was built, that is a prioritisation problem rather than a measurement problem, and it belongs in front of whoever owns the budget.

Where to start

Pick one workflow — invoice matching and order entry are the usual candidates — and instrument it this month with the four metrics. Agree one trigger line with the owner, then watch and wait instead of building. Most teams find that two of their top five roadmap items never clear the threshold, and the one that does is smaller and cheaper than expected. If you want help instrumenting that first workflow or scoping the build that follows, it is the kind of work Maysan Labs does — start here.