What a savings baseline is and why it matters

Ask a vendor how much they saved you, and you will get a number. Ask them "compared to what," and the conversation gets quiet. That second question is the entire discipline of baselining, and it is the single most protective step a buyer can take before entering any results-based engagement.

A savings baseline is a written record of what a process costs today, captured before anyone touches it. It is not a forecast, not a benchmark from another company, and not an estimate produced after the fact. It is a measurement, taken from your own systems, frozen in a document both sides sign. Every savings claim that follows is checked against it. When the baseline is honest, the savings number is honest. When the baseline is soft, every downstream number is negotiable, and negotiable numbers favor whoever wrote them.

A baseline is the only document in a results-based deal that protects the buyer and the provider equally, because it removes the argument before it starts.

What gets baselined

Four families of metrics cover most operational work, and all of them must be expressed in absolute numbers rather than ratios.

Unit costs. What one instance of the work costs: dollars per invoice processed, per order picked, per claim handled, per ticket resolved. Fully loaded, including labor, software, and rework.

Cycle times. How long one instance takes from start to finish, in minutes, hours, or days. Not the average the team remembers, but the figure the timestamps show.

Volumes. How many instances flow through per week or per month. Volume is the multiplier that turns a per-unit improvement into a dollar figure, so it must be pinned down as carefully as the cost itself.

Error rates. How many instances need rework, in counts, not ratios: forty-one invoices bounced back for correction last month, not "a small share." Absolute counts survive audits. Ratios invite argument about the denominator.

Here is what that looks like in practice. A mid-market distributor baselines its invoicing process: each invoice takes 45 minutes of staff time and costs $14 fully loaded, across 900 invoices a month. Written down, that is 675 staff hours and $12,600 a month, or roughly $151,000 a year, on one back-office process. Suppose automation later brings the cost to $6 and the time to 15 minutes per invoice. The saving is $8 per invoice, $7,200 a month, about $86,000 a year, and 450 staff hours returned to the team monthly. Every one of those figures traces back to two numbers written down before the project started. Without the baseline, the same project produces a debate instead of a result.

When the baseline must be locked

Before any change ships. This is not a preference; it is a matter of physics. Once an improvement goes live, the old process no longer exists to be measured. Anyone reconstructing it afterward is working from memory, partial exports, and incentive. A baseline built after go-live is an estimate wearing a measurement's clothes.

Locking means the document is complete, signed by both parties, and dated, with the underlying data archived where the client can reach it. From that point forward the numbers do not move. New information does not revise the baseline; it gets logged as context for the verification stage instead.

1. 1Agree the metrics: unit cost, cycle time, volume, error rate
2. 2Pull twelve months of raw data from systems the client controls
3. 3Compute the numbers and note seasonality
4. 4List exclusions: savings already planned, budgeted, or contracted
5. 5Both parties sign and date the document
6. 6Only after signature does any change ship

The three baseline games

Buyers who have been burned by "savings" engagements were almost never cheated at the measurement stage. They were cheated at the baseline stage, months earlier, usually by one of three moves.

The moving target. The provider keeps the baseline informal, then firms it up after results arrive, choosing the version of "before" that makes "after" look best. A written baseline defeats this completely: the before-number was signed and dated when nobody knew what the after-number would be.

The cherry-picked season. The baseline is drawn from the most expensive stretch of the year, so ordinary months later register as improvement. A retailer baselined in December will show "savings" every February without anyone changing anything. The defense is a twelve-month data window with seasonality noted in the document itself. One month of data is a snapshot; a year is a baseline.

The savings that were coming anyway. A contracted vendor discount kicks in, a planned retirement reduces headcount, a system migration approved last year lands, and the provider counts all of it. The defense is an exclusions list written into the baseline: every saving already planned, budgeted, or contracted is named up front and carved out. If it was going to happen without the engagement, it does not count as a result of the engagement.

Notice the pattern. All three games require ambiguity about the starting point. A baseline that is written, dated, seasonal-adjusted, and explicit about exclusions removes the ambiguity, and with it, the games.

What to insist on seeing

Before signing any results-based agreement, a buyer should be able to hold one document that contains: the metrics in absolute numbers; the data sources and the method used to compute each figure; a twelve-month window with seasonal patterns noted; the exclusions list; and two signatures with a date that precedes any change going live. The client should also keep an untouched copy of the raw data.

One more test separates serious providers from the rest: ask who pays if the baseline shows the opportunity is smaller than the sales pitch suggested. At Qonnex, the analysis and strategic review are free; beyond that, you don't pay unless verified savings exist. The mechanics of how a baseline flows into measurement and verification are laid out on how it works.

A provider who resists writing the baseline down is telling you how the engagement will end. A provider who insists on it before you do is telling you something too. Choose accordingly.