Know what the comparison can actually establish

Quotation analysis requires more than a wins column and a losses column. I begin with when each offer entered the cohort, what qualifies as a resolved decision and how long the customer has had to decide. An industrial buyer comparing a repeat component can respond sooner than one evaluating a new application, so a snapshot of completed decisions may overrepresent the easier purchases. I show the pending population beside the resolved one and ask what further observation would change the business decision. This does not prevent management from acting before the cohort matures. It prevents our report from presenting that action as though every customer had already answered. The commercial owner still interprets the opportunity, and a capacity commitment still needs its own economics, technical assessment and approval. My contribution is a defensible description of what the quotations have, and have not, established.

A distributor-offer experiment can look large because it contains many transactions while having only a few independently assigned commercial groups. I ask where the offer is assigned and whether customers can move between routes before discussing a percentage improvement. The design needs an explicit outcome, observation period and account of cross-route behaviour. Where assignment was by catchment, treating each transaction as an independent experiment exaggerates how much separate evidence we have. Excluding inconvenient buyers after seeing their behaviour can introduce a different bias. I preserve the original assignment account, document departures and make the practical limits intelligible to the commercial team. Random allocation is useful, but it does not manufacture a large independent sample from a small network. A promising result may justify another bounded trial rather than a broad launch; the next step should match the strength of the evidence.

An average purchase cost changes when item prices change, when the company buys different quantities and when the composition of what it buys changes. Those mechanisms can coincide. I compare a fixed basket where suitable item matches exist, then explain what the comparison leaves outside. The method requires defined units, comparable terms and clarity about freight, taxes and specification changes; it cannot be rescued by a polished index title. Procurement colleagues help determine whether the supposedly same item is actually comparable. Finance determines the accounting treatment and the place of any estimate in the business forecast. I resist turning the remaining difference into a single named cause without further work. A decomposition helps executives investigate; it does not by itself prove supplier conduct, procurement failure or a cash saving available to be recovered next quarter.

A service report should distinguish meeting the customer's original commitment from meeting a subsequently revised date. Both can be useful. The original date describes what was first promised; the later one may reflect an agreed change in specification, quantity or delivery arrangement. I retain the sequence and the reason rather than allowing the latest field to overwrite the earlier account. When classification changes, I ask whether the report explains that change to its reader. The objective is not to label every revision a failure or to rank individual employees from incomplete order records. Operations and commercial colleagues must interpret why dates moved. My responsibility is to ensure that an improved percentage cannot arise invisibly from changing the reference point, and that the people using the report know which promise their decision is actually being measured against.