Operations Modernization

Consolidate

Migrate the remaining business onto one system of record, with operational history brought forward rather than stranded.

Seven workstreams get planned and scoped from day one: master data, inventory, open transactions, financial, history, customizations and integrations, and reporting. The first four are what everyone plans for. The last three are what run migrations late, because they are typically discovered in month three rather than scoped in month one.

Guiding principles we hold to:

• One system of record, with a hard date. Indefinite parallel operation is the most expensive available outcome and the most common one.
• Migrate the business, not the database. Where the destination does something differently and adequately, adopt it. Recreating legacy behaviour requires justification, not default acceptance.
• Balance before beauty. Inventory quantity and value, receivables, payables and open backlog reconcile exactly before anything else is judged.
• Every extract is repeatable. Scripted, versioned, re-runnable. No hand-edited spreadsheets in the load path — these get run many times before the run that counts.
• Retain the cross-reference permanently. The old-key-to-new-key map is the only way to answer questions about pre-migration history for years afterwards.

History strategy is a business decision, not a technical one, and it is the single largest lever on cost and timeline. We generally recommend migrating open items plus item-level usage and sales summary — preserving what the analytical work needs at a fraction of full-detail cost — while treating pick, lot and bin history as in-scope regardless, because that is the input to Track 2.

Electronic trading partner integration is almost always the critical path. Map rebuilds and partner-side test cycles are measured in weeks and cannot be compressed by adding staff. We enumerate partners in week one and open those conversations during build, not after it.

Optimize

Once the data is in one place, the same analytical capacity that accelerated the migration turns to the operational questions.

Slotting. Assign every item to a bin whose travel cost matches its pick frequency, subject to cube, weight, zone and handling constraints. Re-evaluated on a cadence rather than set once a decade.

Conventional ABC classification ranks by dollar volume, which is the wrong primary axis for slotting — a high-value item picked twice a month should not occupy prime real estate. The model uses three inputs instead: velocity derived from pick-line frequency weighted toward recent periods, handling class from cube, weight and pack configuration, and affinity — items that repeatedly appear on the same order, slotted near each other regardless of individual velocity. That third one is where pattern detection genuinely beats judgement, because nobody holds pairwise co-occurrence across tens of thousands of items in their head.

Multi-order consolidation. Three orders shipping today each containing the same item from the same lot means someone travels to that pallet three times and breaks it down three times. The pull should happen once, split downstream. We measure the duplicate-touch baseline first — that number is the size of the prize, and most operations do not know it — then build wave grouping that maximises shared touches while respecting ship date, carrier cutoff and staging capacity.

Slotting and consolidation compound: consolidation reduces the number of trips, slotting reduces the cost of each trip. Done in isolation, either is diluted by the other.

Purchasing. Where analytical capacity converts to cash fastest, because the levers are inventory dollars, freight dollars and unit price — all recorded in detail and rarely analysed in aggregate. In recommended sequence: lead-time reality check against promised dates with variance rather than averages; order point tuning where policy and actual usage have drifted apart; purchase order consolidation across sites to reach price breaks and freight thresholds; container fill analysis; forecast variance by programme; a forward expedite-risk list; and a vendor scorecard that is a standing report rather than an argument.

Network replenishment. Most multi-site operations maintain sourcing item by item and never analyse it as a network. Items that should be centrally stocked and transferred are almost certainly being bought at several sites, and items needing local stock are almost certainly being transferred at a cost nobody has priced. Neither condition announces itself. This one waits for Track 1 to complete, for the reason in the table above.

Enable

The failure mode of analytical work is that it lives in one analyst's spreadsheet and leaves when they do.

The model gets built once and parameterised per site. Each location has its own velocity profile, zone topology and constraint set, so recommendations differ — but the logic, the queries and the review workflow are shared. That is the difference between seven separate projects and one project run seven times.

Alongside it: documented playbooks held somewhere the whole company can reach, and site leads brought into the analysis months before their site is affected. People who have followed the work adopt the result faster than people handed a finished workbook on go-live day.