Home / Insights

Innovation in a Small Business: Six Practices, Cheap Tests and an Idea Log

Contents
  1. Six practices that build innovation
  2. 1. Start from the customer's problem
  3. 2. Make curiosity part of leadership
  4. 3. Mix perspectives
  5. 4. Run small, cheap experiments
  6. 5. Make ideas visible
  7. 6. Work in short cycles
  8. Worked example: a new service from existing skills
  9. Protecting time and keeping an idea log
  10. Measuring whether it is working
  11. Common traps
  12. Sources

Innovation in a small business is rarely a laboratory breakthrough. It is a new service built on skills the team already has, a product variant customers keep asking for, or a change to a process that removes a daily frustration. What separates firms that keep finding these from firms that don't is not talent. It is a routine: a clear problem, ideas from more than one head, cheap tests, and a decision on every idea.

This article sets out six practices that make that routine work, a two-hour idea session, a test card, a worked example with numbers, and a calculator that shows why many small tests beat one big bet. It was first published in November 2025 under a title that credited the practices to McKinsey & Company. We found no McKinsey source for the list, so the practices are described here in our own words, and the article has moved to a shorter address.

Six practices that build innovation

Table 1. Six practices, what they mean and how to start

PracticeWhat it meansHow to start
1. Start from the customer's problemUnderstand the problem before proposing solutionsFive customer interviews and one problem statement in a sentence
2. Make curiosity part of leadershipLeaders ask questions and change course on evidenceIn each review, ask which assumption the plan depends on most
3. Mix perspectivesDifferent roles see different optionsPut a front-line person and a customer-facing person in every idea session
4. Run small, cheap experimentsLearn before investingA one-page offer, a mock-up or a two-week trial with a pass mark
5. Make ideas visibleSketches and models get better feedback than descriptionsWhiteboard, storyboard, paper prototype
6. Work in short cyclesPlan, try, review and adjust every few weeksTwo- to four-week cycles with a short review at the end
1 Define thecustomer's problem2 Generate ideas,alone, then together3 Run a cheap testwith a pass mark4 Decide: stop,change or scaleChange: improve the idea and test again. Stop: record what was learned.Short cycles of two to four weeks
Figure 1. The innovation loop. Each practice in Table 1 supports one or more of the four steps.

1. Start from the customer's problem

Most weak ideas are answers to a question nobody asked. Write the problem as one sentence about a real customer: "Contractors cannot get a firm delivery date when they order" leads to better ideas than "improve the customer experience". Design thinking, a widely used method, follows the same order: understand the people you serve, define the problem, generate ideas, build a rough prototype and test it with them.

Five short customer interviews are usually enough to sharpen a problem statement. Ask about the last time the problem happened, what it cost them and what they did about it. Avoid asking whether they would like your idea; people are polite.

2. Make curiosity part of leadership

Teams copy what leaders do. A manager who asks why, listens and changes course when the evidence says so gives everyone permission to do the same. A useful habit is to ask, in every plan review, "What would have to be true for this to work, and how could we check that cheaply?" The first principles thinking article covers this way of questioning in more depth.

3. Mix perspectives

People who do the same job tend to see the same options. Bring sales, operations, finance and someone who talks to customers daily into the same session. In a small firm, inviting a trusted customer or supplier can be the most productive change of all.

How the session runs matters as much as who attends. Research on brainstorming has found for decades that people generate more ideas working alone first than talking in a group, mainly because in a group only one person can speak at a time while the others wait and forget. So have people write ideas silently first, then share and build on them.

Table 2. A two-hour idea session

TimeStepHow
10 minThe problemRead the one-sentence problem and what the customer interviews found
15 minSilent ideasEveryone writes at least five ideas on separate notes, without discussion
30 minShare and buildEach person reads theirs out; similar ideas are grouped; others add to them
15 minVoteThree dots each, placed silently, so the loudest voice does not decide
40 minDesign testsOne test card (Table 3) for each of the top two or three ideas
10 minOwners and datesEach test gets an owner, a budget and a review date

4. Run small, cheap experiments

Most ideas fail, so the aim is to fail cheaply and learn quickly. Before a test starts, write down what you expect, how you will measure it and the result that counts as a pass. Setting the pass mark first stops a disappointing result being explained away afterwards.

Table 3. A test card

FieldExample
We believe thatExisting clients will pay for help preparing for a regulatory audit
To test it we willOffer a one-page service to five clients at a pilot price of $4,500
We will measureHow many accept, and the hours each engagement takes
It passes ifAt least three accept, and the work can be done in under 35 hours once standardised
Budget and timeNo marketing spend; eight weeks
Decision byThe leadership review at the end of week eight

Why many small tests beat one big bet

If one idea in five passes its test, a business that runs 12 cheap tests a year can expect two or three successes. Each success might be worth far more than all the tests together cost. The calculator shows the arithmetic with your own figures.

Try it: the value of a portfolio of small tests

The starting values are illustrative: 12 tests a year at $3,000 each, one in five passing, and each success worth $40,000 a year in extra profit.

Expected successesa year
Cost of testinga year
Expected valuea year from the successes
Return on testing

At the starting values, $36,000 of testing is expected to produce 2.4 successes worth about $96,000 a year, 2.7 times the cost. The result is sensitive to the pass rate and the value of a success, both of which are uncertain. The point is the shape: keep each test cheap, and the few that succeed pay for all of them.

5. Make ideas visible

A rough sketch, a storyboard of the customer's experience or a cardboard model draws better feedback than a written description. People find it easier to criticise a picture constructively, and the picture improves faster. A whiteboard and sticky notes are enough.

6. Work in short cycles

Agree a short goal, work on it for two to four weeks, review what was learned and decide the next step. Short cycles keep momentum and limit the cost of a wrong turn. They also fit naturally with the plan-do-check-act cycle many firms already use; see OODA vs PDCA for how the two compare.

Worked example: a new service from existing skills

A 15-person engineering firm hears the same request from several clients: help preparing documents for a regulatory audit. The tempting move is to launch a new service line with a web page, a brochure and an advertising budget. Instead, the team writes the test card in Table 3 and offers the service to five existing clients at a pilot price of $4,500, against an intended price of $7,500.

Table 4. The three pilots and what changed

PilotHoursWhat was learned
142A third of the time went on chasing the client for documents; a request list is written for the next pilot
238The request list helps; a standard checklist of 40 items replaces notes made from scratch
334Checklist and templates in place; the remaining time is the review itself

Three of the five clients accepted, so the test passed. The pilots earned $13,500 against $10,830 of staff time at $95 an hour, roughly breaking even while the method was worked out. With the checklist and templates, the team expects 30 hours per engagement, a cost of $2,850 against the $7,500 price, or $4,650 contribution each. Ten engagements in the first year would add $75,000 of revenue and $46,500 of contribution, and the firm launches with a proven method, a realistic price and two client testimonials. The figures are illustrative.

Protecting time and keeping an idea log

Innovation competes with daily work and usually loses. Protect a small, regular slot: an hour a month for the leadership team to review ideas and tests, and a quarterly idea session with a mixed group. Keep every idea in one log and decide on each within a month.

Table 5. An idea log

IdeaRaised byProblem it solvesCheapest testDecisionOwner and date
Online booking for service visitsService coordinatorPhone tag to schedule visitsBooking link for one technician for a monthTestService lead, end of month
Reusable crates for local deliveriesDriverPackaging cost and wasteTen crates with two customersTestOperations, six weeks
New product colour rangeSalesCustomer requestsShow samples to five customersPark until springSales manager
Audit preparation serviceProject engineerClients asking for audit helpPilot with five clients (Table 3)ScalePrincipal, next quarter

Measuring whether it is working

Ideas raised60Ideas decided within a month52Tested cheaply12Passed the test3Launched2
Figure 2. The idea funnel at the example firm over one year. Most ideas never reach a test; that is normal.

Table 6. Measures for an innovation routine

MeasureWhy it helpsExample
Ideas raised per quarter, and by whomShows whether people across the business feel able to contribute15
Share of ideas decided within a monthShows whether ideas are taken seriously87%
Tests run per quarterMeasures learning, not talk3
Average cost per testKeeps tests cheap$3,000
Revenue or savings from ideas launched in the last two yearsLinks the routine to results over a realistic periodTrack yearly

Do not judge the routine on a single big win. Most of the value comes from many small, tested improvements, and tests that fail cheaply are a sign the process is working.

Common traps

  • Ideas without a problem. A session with no clear question produces a list that goes nowhere.
  • The loudest voice wins. Write ideas silently first and vote silently.
  • Testing too late. Months spent polishing an idea before any customer sees it.
  • No pass mark. Without one, every result looks "promising" and nothing is ever stopped.
  • No follow-through. People stop suggesting ideas once they see nothing happens to them. Tell each person what was decided about theirs.

Changing how people work also needs care; the change management toolkit covers the main models.

Takeaway. Start from one customer problem, collect ideas silently before discussing them, and test the best with a pass mark set in advance. Keep tests cheap and decide on every idea within a month. In the example, three pilots costing almost nothing turned a client request into a priced, proven service.

The engineering firm, its pilots and the calculator's starting values are illustrative examples with round-number inputs. They do not describe any real company.

Sources

Related reading

Keep reading

Related articles

All 160 articles

Ready to find your next improvement?

Book a strategy call. We will talk through your goals, pain points and where AI, systems or process work can make the biggest difference.