“Our quotations, inventory, approvals, customer data, production tracking and reconciliation are all a mess. I want to fix all of it. Which one should I start with?”
That line opens almost every first consultation. Owners usually say it a little apologetically, as if listing this many problems were their fault. It isn't. It means you know your company well. You can name six specific pain points, which puts you far ahead of the many people who can only say “I want digital transformation.”
But first, some cold water: If you plan to change all six at once, you won't end up with six fixed. You'll end up with six stuck halfway and everyone back on Excel. Then you'll conclude that “this system isn't right for us.”
The real question was never “which one matters most.” You already know which one hurts most. The real question is “which one, done first, makes the other five easier.” Those two questions often have different answers.
I'm Johnny Yang, Brightstream Technology, Taiwan. I've led my team through implementation projects for seven years: 252+ projects across 32+ industries. In those seven years, the failure I've seen most isn't tech that couldn't be built, and it isn't a budget that ran short. It's making the first cut in the wrong place. Pick the wrong first process and you hit a three-way standoff before the project reaches the midpoint: staff complaining, the owner doubting, the vendor explaining. After that, no amount of patching really saves it.
Conclusion first.
To rank processes, scoring four dimensions is enough: frequency, time per run, cost of an error, and resistance to change. The first three tell you whether it's worth changing. The fourth tells you whether you can change it. Score every process you want to change from 1 to 5 on each. Add the first three, subtract resistance, and you have a list you ranked yourself. The first one to tackle is not the highest scorer. It's the one that hurts clearly enough but has the least resistance, because the job of the first project is to build trust and momentum, not to solve the biggest problem. Save the biggest problem for the third or fourth. By then your team already believes this will work.
Why “fix everything” always means nothing gets finished
First, why you can't do them together. It's not a resource allocation problem. There are three structural reasons.
First: process change consumes the attention of the same few people
When owners count resources, they usually count money and headcount. But what process change actually uses up is neither. It's the attention of key employees.
Think it through. Who has to work with you on quotations? Sales support staff and the sales manager. Inventory? The warehouse keeper and purchasing. Approvals? Every manager. Customer data? Sales and customer service. Production tracking? Floor team leads and production control. Reconciliation? Accounting.
That looks like six different groups. In a small business, they overlap heavily, especially that senior woman who knows everything and whom everyone asks. She alone may be the key person on quotations, reconciliation and customer data. Open six workstreams at once, her calendar explodes, and all six stall together.
And these people already have full-time jobs. To them, process change is overtime work, not the actual job. The more workstreams you open, the faster they conclude “this company is just flailing.”
Second: doing it all together means no one owns the outcome
With six workstreams running at once, each one's progress depends on the other five. The warehouse says “I'm waiting on the data cleanup.” The data person says “I'm waiting on sales to confirm customer codes.” Sales says “I'm busy with month-end.” When you finally ask how the project is going, the answer is always “everything's in progress.”
“Everything's in progress” is the most dangerous phrase in a project. It's neither done nor failed, so you can't make any decision about it.
Third: you lose the most valuable thing, one clean success
This is the key point. After the first process is done, one sentence either shows up around the office or it doesn't: “Hey, this really is faster.” That sentence decides how hard the next five will be.
With that sentence, everything after it goes with the current. Without it, you have to re-convince people every single time. Open six workstreams at once and the likeliest outcome is all six at 70%, all six still “a bit clunky,” none cleanly finished. You spent six times the effort and didn't get a single success out of it.
If you're still unsure whether to start at all, read what the first step of digital transformation should be first. Keep “should we do this” and “which do we do first” as separate questions. Don't let them blur into one big anxiety.
Scoring four dimensions: a process priority table you can build yourself
This section is the core of the article. I'm not going to give you a lazy answer like “generally, start with inventory.” Every company's situation is too different. I'm giving you a method you can run yourself.
Grab a sheet of paper or open a spreadsheet. In the left column, list every process you want to change, the five or six you can name. Then add four columns to the right, each scored 1 to 5.
Dimension 1: Frequency. How many times a day does this happen?
Simple question: how many times a day does this process run? A week? A month?
A quotation that runs twenty times a day and a month-end close that runs once a month pay back on completely different scales once improved. For high-frequency processes, even a small efficiency gain gets multiplied by a large number.
Scoring guide: multiple times a day, 5. Once a day, 4. Several times a week, 3. Once a week, 2. Once a month, 1.
A common misjudgment by owners: confusing “important” with “frequent.” Month-end reconciliation is important, of course, but it only runs once a month. The quotations sales put together every day don't feel as important, yet that's where a lot of labor actually goes.
Dimension 2: Time per run. How much staff time does one run take?
The question here: each time this process runs, how many people in the company spend how much time on it? Note: the total, not one person's time.
One quotation: sales spends thirty minutes filling it in, the manager ten minutes reviewing, the assistant fifteen minutes entering it into the system. Total time for that one quote is nearly an hour, not thirty minutes. What small businesses underestimate most is exactly this kind of time spread across many people. Nobody feels they spend much, but it adds up.
Scoring guide: rank by total staff time per run. The most time-consuming gets 5, the least gets 1. You don't need precision. You just need to rank high to low.
Dimension 3: Cost of an error. What do you pay when this goes wrong once?
Many owners skip this one, yet it's often the real reason something should go first.
Errors are not all the same. A wrong phone number in customer data: retype it. A quotation sent out missing one material cost: that whole order may lose money. Inventory numbers wrong: you may have already promised a customer a delivery date you can't meet. Reconciliation wrong: you may have to renegotiate with a customer, and what takes the hit is trust.
Ask yourself three questions. Does this error turn directly into lost money? Does it damage customer relationships? Does the owner personally have to step in to fix it? Yes to all three: 5.
A process with a high cost of error deserves to move up even if it doesn't run often. It's not an efficiency problem. It's a risk problem.
Dimension 4: Resistance to change. How many people's habits will this step on?
The first three measure whether it's worth changing. This one measures whether you can actually change it. It counts as a deduction.
Resistance comes from a few places. How many departments need to cooperate? Will some senior employee's way of doing things be overturned? Does the current process involve outside parties (customers, suppliers, the accountant)? Does it touch sensitive questions like who gets to see what?
Scoring guide: within one department, affecting two or three people, no outside cooperation needed: 1. Across three or more departments, changing habits a senior employee has had for ten years, and requiring customers to change too: 5.
How to calculate your ranking
The formula is simple:
Priority score = Frequency + Time per run + Cost of error − Resistance to change
Scores will fall between −2 and 14. When you're done, you'll have a table like this (the scores below are illustrative; fill in your own):
| Process | Frequency | Time per run | Cost of error | Resistance | Priority score |
|---|---|---|---|---|---|
| Quotations | 5 | 4 | 5 | 3 | 11 |
| Inventory | 5 | 3 | 5 | 4 | 9 |
| Approvals | 4 | 2 | 2 | 2 | 6 |
| Customer data | 3 | 2 | 3 | 2 | 6 |
| Production tracking | 4 | 4 | 4 | 5 | 7 |
| Reconciliation | 1 | 5 | 5 | 3 | 8 |
The biggest value of this table isn't the number. It's the act of filling it in. When you're working on the resistance column and start seriously asking “if we change this, will Mr. Zhang in the warehouse push back,” you're already doing real project planning.
I strongly recommend not filling this in alone. Get two or three department heads to each fill in a copy, then lay the three side by side. The cell with the biggest disagreement is where perceptions inside your company diverge most. That cell is valuable intelligence on its own.
Why the first one shouldn't be the most painful: pick the least resistant of the painful ones
This is the most counterintuitive section of the article, and the one I most want you to remember.
Once the scores are in, the owner's gut reaction is always: “So I start with the highest score.”
My advice: Don't.
The first project's job isn't to solve the biggest problem
The first process change has three real jobs, and none of them is “fix the biggest pain point”:
First, prove this can be done. Prove it to your staff, and prove it to yourself.
Second, teach the team how to work on this kind of project. How to state requirements, how to test, how to explain clearly “here's how I do it now.” These are skills. They take practice.
Third, build momentum. Turn “changing a process” from a special, all-hands event into something that just normally happens at the company.
All three can be done with a mid-level pain point and a low-resistance process, quickly and cleanly. Use the biggest pain point and you face two mountains at once: technical complexity and people's resistance. Climb the hardest one on your first try and the odds of falling are high.
The biggest pain point is usually the most complex one. That's not a coincidence
Think about why a process becomes the company's biggest pain point. Usually because it involves the most departments, it has piled up the most exception rules, it's most tightly tied to outside parties, or nobody has dared touch it.
The reasons it hurts are exactly the reasons it's hard to change. Here, pain and difficulty are highly correlated, not two separate things.
So “solve the biggest pain point first” sounds bold. In practice, it puts the hardest level at the moment your experience is lowest, your team's trust is lowest, and you know the system least. In project management terms, that's a terrible order.
So how do you pick
Here's how I pick:
Rank by priority score and take the top three or four. Then among those, pick the one with the lowest resistance score.
If the top few all have high resistance, keep going down the list. Find a process that runs often and takes noticeable time, but can be changed entirely within one department. It doesn't have to be the most painful. It just has to hurt enough that people notice.
The test is one sentence: once this process is changed, will a group of people say on their own, “this really is faster”? If yes, pick it.
This is also why many companies have an easier time starting with quotations or inventory. Not because they're necessarily the most important, but because they happen every day, people feel the effect directly, and the scope of change can often be fenced into one or two departments at first. For which one suits your industry, see what's actually the difference between inventory systems and ERP. Figure out first whether you're changing a single module or the whole backbone.
Three of the most common wrong starting points. I see them almost every year
That's the ranking method. Next, three opening moves you can tell at a glance will go wrong.
Mistake 1: Starting with what the owner feels most and staff don't feel at all
This is the most common. What does the owner feel most? War-room dashboards, real-time revenue figures, charts on the phone.
I completely understand the need. Not being able to see the numbers makes owners anxious. That's real.
The problem is: where does a dashboard's data come from? From extra data entered by frontline people. So for staff, the project feels like this: I do a pile of extra work, and the boss gets another screen to watch me on.
Can that project move forward? On the surface, yes, because no one dares say no to what the owner wants. But you'll get the most half-hearted data entry in the world: fields picked at random, notes left blank, every date filled in as today. Three months later your dashboard looks great, and not one number on it can be trusted.
The test is simple: what do frontline people get out of this change? If the answer is “nothing, but the company as a whole will be better off,” it shouldn't go first. It can go third or fourth, once the team has a basis of trust.
Mistake 2: Starting with the most complex core
The second common mistake is deciding “if we're doing it, do it right in one go” and going straight for the company's most central, most complex process. Usually that's production scheduling, or the entire costing system.
The reasoning sounds legitimate: “Everything else is a side issue. Fix the core and the rest falls into place.”
What actually happens: core processes have countless exception rules. You'll find the same work order gets calculated one way for Customer A, another way for Customer B because of a verbal promise made ten years ago, and a third way that only the plant manager knows and nobody ever wrote down. Just mapping the current state takes months, and during those months nobody in the company sees any progress.
The longer the stretch with no progress, the more likely the project gets questioned. This is how I've seen the most projects die. Not by failing, but by taking so long that people stop believing it will ever finish.
For how projects walk step by step toward failure, why ERP implementations fail has a fuller breakdown. A wrong starting point is just one way to die.
Mistake 3: Starting with something that needs major cross-department change
Third: opening with a process that requires three or four departments to change how they work at the same time. The classic case is overhauling the approval system.
Why is cross-department work hard? Because it's not just a technical problem. It's a power problem. Who should have approval authority, whose authority gets narrowed, whose signature moves from the second step to the third. In any organization, these are sensitive.
Step into that on your first project and something that technically takes two weeks is still undecided after two months of meetings. And what people argue about in those meetings isn't the system at all. It's friction between departments that was already there, using the system as the excuse this time.
Should cross-department processes be changed? Yes. But only after you've had one success and already proven that this thing isn't here to punish anyone.
Comparison: wrong starting points vs. right starting points
| Aspect | Wrong first process | Right first process |
|---|---|---|
| Who feels it | Only the owner; the frontline sees extra work | The frontline says on their own “this is faster” |
| Scope | Three or more departments; responsibilities need negotiating | Can be finished within one or two departments |
| Exception rules | Too many to map; the current state alone takes months | Main flow is clear; exceptions can wait |
| Time to visible results | Six months or more, no milestones in between | Within weeks, someone is using it daily |
| Cost of failure | Core process stalls; company operations take a hit | Can revert to the old way; limited loss |
| Effect on what follows | Team concludes “this kind of thing can't be done” | Team is willing to take on the next one; resistance drops on its own |
The first process is done. How do you know you can move to the next?
This section matters just as much, and many people judge it by feel. “It feels like people are using it okay” is not a signal. It's self-reassurance.
What you need are verifiable signals. Here are four. I'd want to see at least three before moving to the next process.
Signal 1: The old way actually stopped. It's not “still running in parallel”
The hardest signal. Has that old Excel file, that handwritten notebook, those verbal updates in the group chat actually stopped?
Check the file's last-modified date. If someone touched that old Excel file last week, the new process hasn't replaced the old one. It's just layered on top. Open a second workstream in this state and the first one will fall apart.
Running in parallel for the first week or two after launch is normal and necessary. Still running in parallel after a month isn't a transition period. It means the new way hasn't been accepted. For why Excel is so hard to let go of, where Excel management hits its limit goes into more detail.
Signal 2: Someone starts asking “Could this also...”
This is the best signal, and it comes up on its own. You don't fish for it.
When staff start saying unprompted, “There's another thing on my end that's a real hassle. Could we fix that too?”, it means two things. They believe this can be done. And they feel it's helping them, not policing them.
The day that sentence shows up is the day you can open the second workstream. Conversely, if a month passes and nobody raises a single follow-on request, check again even if everyone appears to be using it. They may just be using it because they're forced to.
Signal 3: It runs without you watching
Give yourself a test: don't ask about this process for two straight weeks, then check whether it's still running normally.
Many process changes look successful but are actually held up by the owner asking every day, “Did you enter it today?” The owner's attention is the most expensive resource there is. If a process has to keep consuming it to survive, it hasn't truly landed, and you shouldn't split your attention to open a second one.
Signal 4: Did the specific pain specifically go away?
Go back to why you picked this process. Was it “sales spends forever digging up past quotations” or “month-end stock counts never match”?
Ask the people involved the same question now. The answer should be concrete, like “we used to dig through it, now we just search,” not “hmm, it's a bit better.” A vague answer usually means nothing improved and they're too polite to say so.
On the other hand, if any of these three show up, don't move on
First, people are still using the old way and nobody's stopping them. That means tacit acceptance has set in. Opening more workstreams will just copy that state.
Second, data quality is dropping. At first everyone fills things in carefully. Two months later, fields start going blank and notes get sloppy. That's a sign momentum is draining. Deal with this first.
Third, only one person is holding it up. If the whole process runs smoothly only because of one especially cooperative employee, and it falls apart when she takes a week off, it isn't a process yet. It's one person's habit.
Honestly: some processes shouldn't be digitized at all. Cut them
In this last section, I'll say some things that are bad for my own business.
Among the five or six processes you listed, there's likely at least one that isn't worth paying to digitize, because it shouldn't exist in the first place.
Three kinds of processes to cut, not systematize
First: patch processes that grew up to fix some other problem.
The classic example is the various double checks. Someone made a mistake once, so a check was added. Then another mistake, another check. Eventually one form needs three people to confirm it. What you should do is not move those three checks into the system as three approval steps. Ask why it goes wrong, fix the source, and cut three checks down to one.
Move a bad process into a system and all you get is a bad process that runs fast.
Second: processes that run only a few times a year.
Some processes really are a hassle, but happen three times a year. You pay to build it into the system, train everyone, then use it three times a year. After those three times nobody remembers how it works, so they still call someone to ask.
For these, a better approach is a clear written procedure in a shared folder, or just keep doing it by hand. Not every hassle is worth a system. Some are solved by one clearly written form.
Third: processes that exist only because “that's how it's always been done.”
The most common are weekly, daily and monthly reports. Ask “who reads this report?” and you'll often hear “I don't know. It was already there when I took over.”
Before digitizing anything, ask one brutal question: If this stopped for a month, would anyone notice? If nobody would, stop it. That saves more money than any system.
Why I'm willing to say this
Because cutting processes that shouldn't exist is the highest-return move in the entire project, and it costs nothing.
My position has always been this: if your problem can be solved by stopping something, you shouldn't spend money on a system. If you're at a stage where cash flow is tight and orders are unstable, all the more reason to read survive first, then talk about transformation. Put your energy where it pays back right away, instead of getting pushed along by the anxiety that everyone else is doing digital transformation.
An honest consultant should help you spend less, not buy more. You can use this as a test when talking to vendors too: When you ask “could we just not do this process at all,” does the other side seriously discuss it with you, or immediately tell you “our system can handle that”? That answer is information in itself.
FAQ
I want to change five processes. Can I really only pick one first? Could I do two?
You can do two, on conditions: the two must involve completely different people, and both must be low-resistance. For example, the warehouse's inventory process paired with accounting's reconciliation process. The two groups don't overlap, so they can run in parallel. But if both need the sales support staff, that's one, not two.
My advice is to do just one the first time. Not because you lack resources, but because the first time you need to learn how to run this kind of project, and learning goes fastest when you practice one thing at a time. Once the first is done and the team has experience, opening two or three at once in the second round is completely doable, and each will go much faster than the first.
The owner thinks we should do A first, the staff think B. What now?
This disagreement almost always happens, and it's usually a good thing. It means you're seeing different sides of the same company.
Here's how to handle it. Bring out the four-dimension table, have both sides fill one in separately, then compare only the resistance and frequency columns. You'll find the disagreement sits almost entirely in those two. The owner underestimates resistance, because he isn't the one doing the work every day. Staff underestimate the cost of error for certain processes, because they don't see the money side.
Lay out the gaps and discuss them. It usually converges within fifteen minutes. If it doesn't, I lean toward the staff, because they're the ones who use it every day. If they don't accept it, building it is pointless.
We already opened three workstreams and they're half done. Should we stop two now?
It depends on whether any of them is close to done. If one is at 70–80%, concentrate resources and close it out, and freeze the other two. Freeze, not cancel. Tell the people involved clearly, “we'll come back to this one,” or they'll feel their effort was thrown away.
If all three are at 30–40%, re-rank them with the four-dimension table and go all in on the one with the least resistance. At this point what matters most isn't saving time. It's getting one complete success inside the company, fast. Let three workstreams sit stuck halfway any longer and the project burns through its trust.
The vendor says they can do it all at once. Isn't that a better deal?
Not necessarily. Vendors usually say “do it all at once” for two reasons. One, a bundled quotation looks better and the unit price is lower. Two, their product is a full suite anyway, and selling it in pieces is more trouble for them.
The question isn't price. Ask three things. Is go-live phased, or a single cutover? If I decide not to continue after phase one, how is the contract handled? How much more does a phased rollout cost in total than doing it all at once?
The answer to the third usually shows a price gap. That gap is the insurance premium you pay. Whether it's worth it depends on your confidence in this vendor and in your own team. To understand how to read the cost structure, see how to calculate ERP implementation costs. Figure out what the money is made of before deciding whether to split it up.
We're a small company, just a dozen or so people. Does this ranking method still apply?
It applies, and it matters even more. The fewer people you have, the harder the one-thing-at-a-time limit. A company of a dozen or so may have only two or three key employees. You have no room for parallel work.
But one thing needs adjusting: at that size, weight the resistance column more heavily. In large companies, resistance to process change comes from rules and departmental interests. In small companies it comes almost entirely from individual people. If one employee who's done it the same way for fifteen years doesn't like the change, that workstream will be very hard to push. So when a small company ranks, “is this person willing” can almost serve directly as the test of whether it's doable. It's more accurate than the score.
Finally, if you're stuck here right now
If you happen to have a list of processes you want to change, do one very small thing today: open the list, add four columns, and fill them in yourself. No vendor, no meeting. Twenty minutes.
Once it's filled in, you'll find the table itself answers your original question. You'll also see things you hadn't thought of. A process you thought was important may have a resistance score so high you shouldn't touch it now. A small process you never paid attention to may turn out to be leaking time every day.
There's one answer the table can't calculate: this shouldn't be automated at all. Some processes change too fast, run at too little volume, or have so many exceptions the rules can't be written down. Forcing a system on them just adds another layer of burden. Rule those out first and your list gets much shorter. I've listed the criteria in Some things you shouldn't automate: a checklist before you rush into a system.
If you already know which one to tackle, these two are the ones I'm asked about most and the easiest to underestimate: Purchase orders sent over LINE? Where things break between requisition and receiving, and Who's holding up approvals? How to find the process bottleneck. With the second one, people often score all four dimensions and only then discover the real bottleneck isn't the process at all. It's waiting for one person to say yes.
If you're stuck here, filled in the table and still can't rank, or want to know where other companies in your industry usually start, listen to our podcast, or browse our case library to see the order similar-sized companies went in. I won't give you a standard answer, because there isn't one. But I can help you set up the framework for deciding, and you'll do the ranking yourself.
After all, the person who knows best where your company hurts has never been me. It's you.




