Corporate & Business · NPO & Social Impact · Procurement & Sourcing
Most of the work was deciding where not to change supplier
Food purchasing had grown organically across a distributed care environment: wholesalers, national retailers, specialist meat and dairy suppliers, fresh produce, nutrition providers and local stores, paid for by account, cash and card alike. The task was never to find a cheaper grocer. It was to establish what was actually being bought, where products were genuinely comparable, and which routes should be left alone.
- ~18%
- accepted year-one saving
- 27 → 5
- core procurement channels
- 451
- product lines tested
- 300+
- controlled exceptions retained
Proof context: A distributed multi-site care environment
The situation
Twenty-seven suppliers, and no reliable picture of what was bought
Food purchasing had developed organically across a distributed care environment. Different sites used wholesalers, national retailers, specialist meat and dairy suppliers, fresh-produce outlets, specialist nutrition providers and local stores. Payment methods were just as mixed: supplier accounts sat alongside cash and debit-card purchases.
So the problem was not “find a cheaper grocery supplier”. In a fragmented environment, supplier totals can be perfectly accurate while the underlying product picture remains far too inconsistent to support a like-for-like sourcing decision. The baseline had to be good enough to support a decision, not merely to total invoices.
What KuTh did
Reconstruct, normalise, test — then decide line by line
- Rebuilt twelve months of spend. Roughly 2,400 purchase-line records across supplier invoices, petty cash and debit-card purchases, covering more than 1,200 raw item descriptions before normalisation.
- Established genuine comparability. The same requirement appeared under different descriptions, brands, pack sizes and units. A cheaper pack was not treated as a saving unless size, quantity and commercial use were sufficiently comparable — and where no equivalent could be supported, the model did not invent one.
- Tested 451 product lines forward. Each line compared its historical route against candidate channels with a usable price, using channel-specific escalation inputs rather than one blanket inflation assumption.
- Separated discount from lower cost. The wholesale route created value through competitive base pricing rather than a headline discount; the retail concession was a contractual percentage but carried channel restrictions. Treating both as the same thing would have produced a weak operating recommendation.
- Designed how sites would actually buy. Branch accounts, ordering authority and delivery acceptance were built into the recommendation. A concession attached to in-store purchasing cannot be assumed on a delivered or digital order.
The result
Where the accepted value came from
Scroll table sideways →
| Value source | Year 1 | Year 2 | Movement |
|---|---|---|---|
| Broad wholesale route | 54% | 46% | −8 pp; largest Year 1 contributor, driven by lower base pricing |
| Institutional retail route | 41% | 51% | +10 pp; becomes the largest Year 2 contributor |
| Regional specialist route | 5% | 3% | −2 pp; smaller but valid negotiated regional contribution |
| Top two combined | 95% | 97% | +2 pp; value stays heavily concentrated in the two largest levers |
These percentages describe where the accepted model created value. They are not supplier market shares, and they do not mean every product in a channel moved to that route. The accepted first-year position sits at roughly 82 against a reconstructed historic baseline of 100 — about 18% below it.
The control that matters most
More than 300 places where the answer was “don’t move this”
The amended workbook carried over 300 exception entries: lines where the alternative was more expensive, no longer available, insufficiently comparable, or operationally weaker. Others protected specialist nutrition, dairy and meat requirements where substitution was not evidenced, or preserved donated and preferential value that a spreadsheet would have quietly replaced with a purchased alternative.
The result was not “27 suppliers became 5”. It was 27 historical suppliers becoming five core procurement routes, while the lines that were cheaper, unavailable, donated, specialist or operationally unsuitable elsewhere were deliberately preserved. Five core channels plus documented exceptions — not five exclusive suppliers. Supplier rationalisation should simplify the operating model, but it should not become an objective in its own right.
Validation
The number was allowed to go down
The first report and pricing model produced a higher two-year savings outcome. The model was then amended using refreshed pricing, product availability and additional branch purchasing information. The final accepted two-year result came in approximately 13% lower than that earlier draft.
That matters more than the headline. It shows the project was not managed toward a predetermined figure: when the evidence weakened, the claim weakened with it. The same principle applied line by line — products were removed from the saving where price increases or supply changes meant the alternative was no longer preferable.
Evidence and publication boundary
Published as a client-accepted outcome: the revised procurement structure and its associated savings model were signed off as the agreed result. Figures are expressed as percentages, counts, derived ratios and indexed outcomes.
Client and supplier identities, exact rand values, branch names, invoice references, unit-price schedules, negotiation scripts, scoring thresholds and KuTh’s normalisation and matching logic are withheld. The ~81% routing-complexity reduction is derived from the change in supplier count to future-state core channels; it is not a claim that all former suppliers were eliminated.
Supporting documents