# Starting DX with small improvements: comparing 12 measures to consider at ¥100,000–¥300,000

> Where to start DX on a small budget: SaaS, partial automation and custom builds compared, 12 measures scored, and a payback formula. Amounts are not guaranteed.

- Published: 2026-11-09
- Author: 友田 陽大
- Tags: DX, 業務改善, 中小企業, コスト最適化, 技術選定
- URL: https://tomodahinata.com/en/blog/small-business-dx-10-30man-improvements
- Category: Small process improvements

## Key points

- What is realistic at a small budget is configuring an off-the-shelf tool for one task, or partial automation. Building a new business system with logins and permissions rarely fits at this scale.
- Comparing 12 measures by return, running burden and prerequisites narrows the first candidates to work that repeats every week, follows a fixed procedure and has consistent input data.
- Payback is the initial cost divided by (hours saved × hourly staff cost) minus running costs. In the hypothetical example, halving the expected hours saved more than tripled the payback period.
- For 2026 the subsidy is called the 'Digitalization & AI Introduction Subsidy 2026' and works by introducing registered IT tools through a registered support vendor. Decide on the measure first, then check eligibility.

---

If you want to start DX on a small budget, the first step is not choosing a tool. It is **picking one task you repeat every week and measuring how long it takes.** If the measured time is small, no tool will pay for itself. If it is large, configuring an off-the-shelf tool may be all you need.

**The amounts in the worked examples are neither market rates nor my price, and ¥100,000–¥300,000 is no guarantee that a measure can be done for that sum.** Cost varies widely with the volume of work, how scattered the data is, the tools you already use and the time needed for checking.

This article gives a comparison table and a diagnostic table for choosing a measure; the detailed steps live in these articles:

- Automating one manual task: [automating manual Excel work](/blog/excel-manual-work-automation-small-business)
- A system for handling quote requests: [bringing quote requests out of Excel and email into one place](/blog/quotation-request-management-excel-email)

If you run the business on your own and want the thinking behind moving the whole operation onto SaaS, [SaaS for the solo company owner](/blog/solo-president-saas-digitalization-guide) is the better fit. This article is for companies with staff deciding what to try first on a small budget.

## What a small budget covers, and what it does not

Turning the budget into hours gives a sense of scale. My own base rate is ¥7,000 an hour (excluding tax), with scope and price fixed in writing before work starts. At that rate, ¥100,000–¥300,000 is about 14–43 hours of work.

Think of those hours in terms of the kinds of work involved.

| What you pay for | What tends to happen in small improvements |
|---|---|
| Initial setup of an off-the-shelf tool | Designing fields, permissions and notifications |
| Moving data across | Cleaning up inconsistent entries in old Excel files takes time |
| Building the automation | Copying, totalling and generating documents by a fixed procedure |
| Checking and testing | Confirming results match on real data |
| Explaining it in-house | Procedures and practice |
| Monthly costs | SaaS subscriptions and maintenance time |

The items people overlook are data migration and checking. Each person builds tables differently, dates are typed as text, the same customer is registered under three names. Sorting that out can take longer than configuring the tool.

On the other hand, assume that building a new business system with logins and permissions, migrating existing data into it and rolling it out across several departments rarely fits a small improvement. Design for each screen, permission testing and migration rehearsals each take time.

## Off-the-shelf SaaS, partial automation and custom builds compared

| Aspect | Off-the-shelf SaaS | Partial automation | Custom build |
|---|---|---|---|
| Suited to | Work you can do the same way as other companies (invoicing, attendance, bookings) | Work where you keep the current Excel and procedures and cut only copying or totalling | Work whose flow or permissions are specific to you and cannot be expressed by a product |
| Initial effort | Configuration and data migration | Tidying the procedure and building | Requirements, design, development, testing |
| Monthly cost | Subscription (often grows per user) | Often small, but maintenance time is needed | Server costs and maintenance |
| Ease of change | Within the vendor's features | Easy if you know who built it | Unlimited, but each change costs money |
| Main risks | Price or feature changes, getting your data out | Only the builder can fix it; stops when the input format changes | Cost and schedule overruns, nobody to maintain it |
| How it ends | Export the data and switch | You can go back to manual work | Needs someone to take it over |

My order of decision is to check first whether off-the-shelf SaaS is enough, then fill only the missing part with automation or development. I would start with a custom build only when the workflow itself is a competitive advantage and bending it to fit a product would cost too much.

Partial automation has the strength of being reversible to manual work. Keep the original procedure so that if the automation stops, the business does not. Put the other way, if you leave it in a state only its builder can fix, in a few years nobody will be able to touch it. If you commission one, get a written procedure and an explanation of how it fails when the input format changes.

## A diagnostic table of 12 measures

The table compares 12 measures that small businesses commonly consider, by return, running burden and prerequisites. Return is a qualitative guide combining the expected cut in weekly working time and in errors and missed items; it is not a numerical forecast. Re-score it for your own situation.

Key: ◎ large / ○ medium / △ small. Running burden is the in-house effort needed after introduction (lower is better).

| # | Measure | Main form | Return | Running burden | Prerequisites |
|---|---|---|---|---|---|
| 1 | Standardise where files live and how they are named, using cloud storage | Off-the-shelf SaaS | ○ | Low | Someone owns the rules |
| 2 | Move quotes and invoices into an invoicing tool | Off-the-shelf SaaS | ◎ | Low | You can tidy the customer and item lists |
| 3 | Gather enquiries into one form and a shared inbox | Off-the-shelf SaaS, configuration | ○ | Low | Agreement on who replies |
| 4 | Hand scheduling to a booking tool | Off-the-shelf SaaS | ○ | Low | Calendars are shared |
| 5 | Turn paper daily reports and application forms into form input | Off-the-shelf SaaS, no-code | ○ | Medium | Devices and connectivity on site |
| 6 | Move attendance and shift totals into a tool | Off-the-shelf SaaS | ◎ | Low | Work rules can be expressed as settings |
| 7 | Turn the job or inventory ledger into a shared no-code database | No-code | ◎ | Medium | Someone decides the fields |
| 8 | Automate copying data out of Excel | Partial automation | ◎ | Medium | Source and destination formats are fixed |
| 9 | Automate CSV import and monthly totals | Partial automation | ○ | Medium | The exported CSV columns are stable |
| 10 | Generate standard documents (delivery notes, work reports) as PDFs automatically | Partial automation | ○ | Low | The document layout is fixed |
| 11 | Record quote requests in a ledger and notify the assigned person | Automation, off-the-shelf SaaS | ◎ | Medium | Requests can come through one channel |
| 12 | Connect existing tools through APIs (interfaces that let systems talk to each other) | Small custom build | ○ | High | Both tools expose an API |

Three ways to read the table:

- Even with ◎ return, defer a measure whose prerequisites you cannot meet. Measure 7, for instance, only adds an unused ledger if nobody decides the database fields
- For measures with high running burden, decide who will maintain it before you start. Measure 12 can stop when the connected service changes its specification, and you also need a way to notice that it has stopped
- Measures 1–4 and 6 can often be done by configuring off-the-shelf SaaS, so you can start without outsourcing

The first candidates are tasks that repeat every week, follow a fixed procedure and have consistent input data. Measures 8, 10 and 11 tend to meet these conditions; their steps are covered in the two articles listed at the start.

### Score it for your company

Turning the symbols into numbers makes candidates easier to compare.

```text
Priority = return (1–3) × prerequisites met (0 or 1) ÷ running burden (1–3)

Return: large 3 / medium 2 / small 1
Prerequisites: can be met now 1 / cannot be met 0 (= out of scope for now)
Running burden: low 1 / medium 2 / high 3
```

Prerequisites are multiplied so that a measure you cannot meet scores zero and drops out. On a tie, do the measure you can revert to manual work first.

## The payback formula

Whether automation or a new tool pays for itself can be checked with this formula.

```text
Monthly benefit = hours saved (hours/month) × hourly staff cost
Monthly net benefit = monthly benefit − monthly running cost (subscription + maintenance time)
Months to payback = initial cost ÷ monthly net benefit
```

Fewer errors and fewer missed items are benefits too, but they are hard to price, so calculate on time alone first and note the rest as supporting factors.

Hypothetical example: every figure below is chosen for the arithmetic and represents no real company or market rate. Copying work takes 10 hours a month; after automation, 2 hours of checking remain and 8 hours are saved. Assume an hourly staff cost of ¥2,500, monthly running costs of ¥3,000 for the tool plus one hour of maintenance (¥2,500), and an initial cost of ¥200,000.

| Item | 8 hours saved | 4 hours saved |
|---|---|---|
| Monthly benefit | ¥20,000 | ¥10,000 |
| Monthly running cost | ¥5,500 | ¥5,500 |
| Monthly net benefit | ¥14,500 | ¥4,500 |
| Months to payback | About 14 months | About 44 months |

Halve the expected hours saved and the payback period more than triples, because running costs are fixed. So enter hours saved from measurement, by logging the current work for one to two weeks, not from hope.

```text
Work log (keep for 1–2 weeks)
Date / task / person / start and end time / number of items / any rework and why
```

When people log their work, they often find most of the time goes on searching, checking and redoing. In that case, file rules (measure 1) or a standard input format may help more than automation.

## Before asking for quotes: log the work and use free trials

There are things you can do before outsourcing:

- Log the work to identify the tasks that take the most time
- Try an off-the-shelf SaaS on real work for a week during its free trial
- Consult a public service. Japan's [Yorozu Support Centers](https://yorozu.smrj.go.jp/) are established by the government and say they handle management consultations for small businesses free of charge, as many times as needed. Since April 2026, productivity improvement support centers have also opened inside Yorozu Support Centers nationwide

Outsourcing is a poor fit when:

| Situation | Why |
|---|---|
| The target task has not been chosen | Quotes without a clear target get you whatever is easiest to build |
| Each person follows a different procedure | The procedure needs standardising before automation |
| Nobody has been named to use it afterwards | Only the running costs remain |
| You have not tried off-the-shelf SaaS | Configuration alone might be enough |

Adding tools also adds accounts to manage. IPA's [Information Security Guidelines for SMEs](https://www.ipa.go.jp/security/guide/sme/about.html) describe cloud service security as maintained by the provider and the user sharing roles and responsibilities, and set out what users should do in the [Guide to Using Cloud Services Safely](https://www.ipa.go.jp/security/guide/sme/ug65p90000019cbk-att/sme_guideline_v4.0_app_cloudservice.pdf). Decide at introduction who the administrator is and how leavers' accounts will be shut off.

## Decide on the measure first, then check subsidies

For fiscal 2026 the so-called IT Introduction Subsidy, a Japanese government grant for small businesses adopting IT tools, is offered under the name **"Digitalization & AI Introduction Subsidy 2026" (デジタル化・AI導入補助金2026; formerly the IT Introduction Subsidy)**. According to the [official programme overview](https://it-shien.smrj.go.jp/about/), eligible IT tools are those registered after prior review by the secretariat, and applicants apply in partnership with a registered "IT introduction support vendor" (except in the multi-company collaboration category).

When I checked the official [schedule](https://it-shien.smrj.go.jp/schedule/) on 9 November 2026, the standard category listed a deadline of 17:00 on Tuesday 1 December 2026. The same page says only confirmed application rounds are published, so if you are considering applying, always check the latest dates on the official site.

Because of this structure, custom builds made only for your company rarely qualify, and the programme suits measures that introduce off-the-shelf tools (such as measures 2 and 6 in the diagnostic table). Categories, subsidy rates and cases that do not qualify are covered in [the buyer's guide to the IT Introduction Subsidy 2026](/blog/it-introduction-subsidy-2026-buyer-guide). Do not pick a measure because a subsidy exists; decide on the measure, then check whether the subsidy applies.

## Three steps to choose the first measure

1. Write down three tasks you repeat every week, and log the work for one to two weeks
2. Score them in the diagnostic table and drop measures whose prerequisites you cannot meet
3. For the one that remains, try an off-the-shelf SaaS on a free trial, or get a quote for automation

When you ask for a quote, write down the target task, the input and output formats, how results will be checked and the expected monthly running cost; the quotes you get back will be easier to compare. If you work with me, I likewise recommend an off-the-shelf tool when one is enough, and fix the scope and price in writing before work starts. Details are on the [routine work automation](/services/business-automation) page.