When you de-chunk a process, the risks stop hiding.

Why a real process, not a textbook case

Last week I ran two connected workshops: one on project management fundamentals, and one on using SIPOC for process mapping — the backbone of good supplier, customer and stakeholder management.

Theory only sticks when you make it real. So instead of a textbook case, we mapped a process my team actually lives with: the pre-submission of an imported, off-nomenclature pharmaceutical product to the ANPP (Agence Nationale des Produits Pharmaceutiques). We zoomed into one concrete sub-process and mapped it end to end — paying the 25% pre-submission fee.

It sounds administrative. It isn't. It's where three institutions, an internal department, a cheque and a company stamp all have to line up. And the real lesson wasn't about speed — it was that the moment we broke the process into its smallest pieces, task by task and output by output, the risks we'd never named started appearing at every single step.

How we ran it

I displayed the source material on screen, then split the room into specialist teams — each owning one lens of the map and defending it to the others: Suppliers + Inputs, the Process crew, Outputs + Customers, Risk Spotters, Time Guardians, and a WBS group breaking the work into phases and chunks.

Splitting the lenses is what made it click. The Process crew wanted to move fast; the Time Guardians and Risk Spotters kept pulling the map back to reality. That friction is the learning.

The map we built

• 1

  Submit the BV on E-TASDJIL

  Inputs: ANPP access and the RA product owner's information. Supplier: ANPP. Output: a submitted BV screenshot as evidence, which goes to the ANPP.

  InstantRisk: platform bug or internet issue — which is why the screenshot matters
• 2

  Collect the ANPP-signed BV

  When the platform status shows "BV signed," we take the company stamp to the ANPP, collect the signed and stamped BV, and acknowledge their register. Output: the signed original physical BV — its customer is the tax office, Impôts El Madania.

  ~5 daysRisk: signature not ready, travel, register or cachet delays
• ∥

  In parallel — the cheque

  Simultaneously, internal Finance issues the cheque from the fees we provide; its customer is also Impôts El Madania. Crucially, this runs alongside Task 2 — a simultaneous task we only saw once we de-chunked.

  72h – 1 weekRisk: cheque late or wrong amount; Finance agent on annual leave
• 3

  Pay at Impôts El Madania

  We take the signed BV and the cheque to the tax office; they hand back the quittance cachetée and the BV cachetée. Output: the stamped proof, whose customer is the ANPP dossier.

  Same dayRisk: counter queue; cheque rejected; missing or illegible cachet

The real lesson: de-chunk, then the risks appear

At first, the Process crew mapped in big blocks — they didn't split the work task by task, output by output. Nothing looked wrong on the surface. But the hand-offs, the simultaneous tasks and the step-level risks all stayed invisible.

The moment they forced themselves down to one task → one output per line, the map changed. Two things surfaced on their own: the cheque could run alongside the ANPP retrieval instead of after it, and each little step revealed its own failure point — early, before it could bite.

Definition of Done: write the output so no one can miss a step

Here's the discipline that ties it together. In SIPOC, the Output column is where a process quietly succeeds or fails. An output written without its Definition of Done — the explicit, shared statement of what "done" actually means — leaves room for interpretation, and interpretation is where steps get skipped.

This matters most in a written SOP, where the reader may be a newcomer with no tacit knowledge of the process. They will do exactly what is written — no more. So the output description itself must carry the full definition of done.

Rule of thumb: for every Output, ask "how would a brand-new colleague know this is truly done, and how would they verify it?" Put that answer into the output description. That's the difference between a map that looks complete and an SOP that actually is.

From spotting risks to a plan

SIPOC didn't just draw a picture. It fed a structured plan: spot a failure point at every step, categorize it (platform, external institution, internal Finance, logistics, compliance evidence), score it by likelihood × impact, then give each risk a mitigation and an owner.

Appendix — the full SIPOC, one line per task

The takeaway

A process you "already know" almost always hides single points of failure and undocumented hand-offs — but only at the block level. De-chunk it, define each output to its Definition of Done, and they have nowhere left to hide.

Next: mapping the full pre-submission flow the same way. If you were in the room — which process should we dissect next?

Source: ANPP — Direction d'Enregistrement (anpp.dz). Fee amounts vary by product type (imported non-essential: 250,000 DA at 25%; imported essential: 150,000 DA at 25%). Timelines are field estimates and should be validated per product.