Your SAP Change Impact Business Case Needs Five Numbers, and None of Them Come From a Vendor
If a vendor is telling you what your ROI is going to be, probably just stop right there. We sell tools, from various vendors. We pass through partner discounts from our bulk buying discounts and we supply services to make sure you can adopt the tools to get to the best possible ROI number, but the ROI estimates we often see have sparked this article.
Most pitches include a percentage or good looking metric somewhere in the deck. Somebody in finance asks where it came from, and the honest answer is a product page, and the meeting moves on. The box has been checked but it's always struck me as odd when it bypasses scrutiny. And the times it does get scrutinized, it generally ends poorly so we're going to talk about how you can skip that entirely. All you need is five numbers, all of them yours, and you can probably produce four of them in an afternoon.
We apply this same approach in our SAP Change Impact Analysis article why nobody can prove what your transport touched which talks about the need for the tool, and this one prices it.
Why the published figures don't belong in your model
Vendors, in this case Tricentis and its partners (us included), publish numbers on test scope reduction, release acceleration, production error rates, and hypercare cost. Some of it is marketing copy and some comes from commissioned research, which is better sourced but still commissioned.
Both produce useful metrics but they are telling you the same thing, "which levers are worth measuring". Put either one in your model and you've handed the skeptic in the room a free win, because if they ask where the number came from, and you give them any of the answers above...you can't answer what they are really asking, which is "how does this number apply to our business".
A word of caution, your own numbers will be way less impressive but they will be so much more difficult to argue with. I'm a firm believer well supported facts rarely result in long arguments, which gives us more time to focus on impact.
The five numbers
- Cost of one regression cycle. Tester days times loaded rate, plus environment time, plus the coordination overhead nobody books to the project. Include the functional analysts who get pulled in for validation, because they're the expensive ones. Most SAP teams have never produced this figure, which is itself worth mentioning to your steering committee.
- Cycles per year. Support packs, enhancement packs, functional releases, and the emergency transports that trigger a partial regression anyway. Count the partial ones honestly. Multiply by number one and you have your annual regression spend, probably for the first time.
- Your real change rate. Take the last three transports and work out by hand which of your test cases exercise an object that actually changed. This is slow, irritating work. The irritation is the datapoint, and it's the number that makes the rest of the case obvious.
- A reduction you would defend. Ignore the published percentage. Pick a figure you'd say out loud to your CFO and stand behind six months later, then check whether the case still closes there. If it closes at half the vendor's claim, you never have to argue about the vendor's claim.
- The price of one escaped defect. Covered below, because it's the one nobody models and usually the largest line item in the equation.
Pricing an escaped defect
Take a single realistic failure, perhaps it's a pricing condition that posts wrong, a delivery that won't create, an authorization that quietly blocks a posting path at month-end. Then add up what it actually costs:
- Business interruption: hours that process is degraded, times what an hour of it is worth
- Rework: developer and functional analyst time to diagnose, fix, retest, and re-transport under pressure
- Hypercare: the extra staffing you carry for weeks after a bad release, which is a real budget line even though it never appears in a test plan
- Schedule: for anyone with delivery commitments and penalty clauses, the cost of missing a date because the release slipped or got rolled back
- Audit exposure: what it costs to be unable to show what changed, what validated it, and who accepted the residual risk
Run that once, at a conservative interruption estimate, and compare it to your annual regression spend from number two. In regulated and delivery-committed environments, the escaped-defect line routinely lands an order of magnitude above the testing savings. A lot of business cases lead with the savings and this gets buried, which is backwards.
What the answer usually looks like
Three outcomes and all are probably useful.
The case closes comfortably on numbers one through four alone. Buy it, and use number five to decide how fast you roll it out.
The case only closes when you plug in a vendor's best-case percentage. Don't buy it. Your release cadence probably isn't high enough yet, and change impact analysis pays back on programs running transports monthly or faster.
The case closes on number five by itself. I've seen it, and it's great when the realization happens but it's slightly awkward. It means you've been carrying that exposure for years and never put a figure against it or never acknowledged it, and the tooling decision is now the smaller conversation.
Before your next steering committee
Bring one slide with five numbers on it and no logos. If you can't fill in number one, that's the finding, and it's worth reporting on its own. A team that can't cost its own regression cycle cannot tell you whether any testing investment is working.
Merito helps SAP teams build this model and then act on it, including licensing and implementing Tricentis LiveCompare across all three packaging options. This article applies to SAP programs but it's the same approach for any of the vendors we support. We'll also tell you when the numbers say wait, which happens more often than you'd expect from someone who sells licenses.
We're happy to chat about this further. You can bring your release cadence and your last regression invoice and we'll do the arithmetic with you.
