Choosing Between SaaS ERP, an In-House Rewrite, and Doing Nothing
Golden Harvest Foods is a 120-person food manufacturer with revenue of PHP 180 million. Its order-entry and invoicing system was written in Visual Basic 6 by an employee who retired three years ago; there is no documentation, and it has failed three times this quarter. Each stockout caused by the system costs about PHP 1.2 million in lost margin, and the firm averages two such stockouts a year. The owner is weighing three options. Option A: a SaaS ERP at PHP 2.4 million per year plus a one-time implementation of PHP 6 million over 9 months; the vendor has reference sites in the same industry. Option B: rewrite in-house, hiring two developers at PHP 900,000 each per year for 18 months plus PHP 1.5 million in tools and hardware, then keeping both for support at the same salary. Option C: keep the legacy system, spend PHP 400,000 per year on maintenance and accept the failures. The firm has PHP 8 million cash, no CIO, and the owner's son, a business graduate, would sponsor the project. The board wants a five-year view.
1. Build a five-year cost comparison for the three options, stating your assumptions, and give a recommendation with the reasoning that the numbers alone do not settle.
Read the worked answer
Assumptions. Five-year horizon; stockout exposure continues until a replacement is stable; Option A reaches stability after month 9, Option B after month 18, Option C never; no inflation or discounting, which is stated as a simplification.
Option A: implementation 6.0M + subscription 2.4M x 5 = 6.0 + 12.0 = 18.0M, with near-zero stockout cost after go-live.
Option B: developers 2 x 0.9M x 5 = 9.0M; tools and hardware 1.5M; stockouts during the 18-month build at roughly 2 per year x 1.2M x 1.5 = 3.6M. Total 14.1M.
Option C: maintenance 0.4M x 5 = 2.0M; stockouts 2 x 1.2M x 5 = 12.0M. Total 14.0M.
The arithmetic is uncomfortable: doing nothing ties with the rewrite and beats the SaaS option. That is exactly why cost alone cannot decide. Option C's 14.0M is almost entirely risk exposure - it is contingent, but it is also unbounded, since a failure during a peak season or a tax deadline is worse than the average. Option B is only cheaper if two developers can be hired, retained, and managed to deliver a system in a domain they do not know, with a sponsor who has no IT delivery experience and no CIO to arbitrate. A missed 18-month estimate pushes B toward A's cost while the failures continue.
Recommendation. Option A, conditional on negotiated terms, because it is the only option that removes the single point of failure within the cash position, carries vendor references in the same industry, and does not depend on hiring scarce skills. The premium over B is roughly 3.9M, which is about three stockout events - an insurance premium the firm can quantify. If the board rejects A, B is defensible only with a named delivery partner and a hard 12-month checkpoint.
Whichever is chosen, a legacy-exit plan and documented data export are non-negotiable, or the firm simply changes which supplier it is locked into.
2. The vendor offers 15% off if Golden Harvest prepays all five years of subscription on signing. Evaluate the offer against the firm's cash position and risk profile, and state the contract terms you would require before signing either version.
Read the worked answer
The offer in numbers. Five years at 2.4M is 12.0M; a 15% discount saves 1.8M, taking the five-year figure from 18.0M to 16.2M. That is real money, roughly one and a half stockout events, so it should not be dismissed.
Why it still fails today. Prepayment of 12.0M plus the 6.0M implementation is 18.0M against 8.0M of cash. The firm would have to borrow or draw down working capital to take a discount, converting an operating expense into a financing decision plus a five-year bet on one vendor. Prepaying also removes the firm's only leverage: after the cheque clears, the vendor has no commercial reason to fix implementation problems, and if the vendor is acquired or exits the market, the prepaid amount is an unsecured claim. For a firm with no CIO and no ability to run a second implementation, lock-in is the dominant risk, not the 1.8M.
Counter-proposal. Take a shorter committed term (two years) with an optional extension at a capped price index, or ask for the discount against annual prepayment instead of a five-year upfront. Either keeps most of the price benefit while preserving an exit point.
Terms to require before signing. Data ownership and a contractual right to a complete export in a documented, machine-readable format at any time and at exit; an escrow or continuity clause covering vendor failure; a service level for availability and support response with service credits, not just goodwill; a defined acceptance test and payment milestones tied to it (never pay implementation in advance); a change-control clause with rates; a clear statement of where data is stored and who can access it, to satisfy data privacy obligations for customer and employee records; and a termination-for-convenience clause with a defined transition-assistance period. The discount is worth having, but not at the cost of these.