Most price increases are decided in a spreadsheet and defended afterwards in a meeting. The arithmetic is sound, the customer reaction is a guess, and the first real evidence arrives when the cancellations do.
This post sets out a four-week alternative: a short trade-off exercise run on your actual customer base, not a panel, that tells you which segments absorb the rise, which switch, and what you could offer to buy back the goodwill.
Why 2026 is a tighter year to get this wrong
Two years ago, most companies could raise prices into an environment where everyone else was doing the same and buyers had largely stopped counting. That cover is thinner now.
Consumer reporting points the same way. US coverage has described older Americans cutting back on groceries as everyday costs squeeze fixed budgets (WTOP News, https://news.google.com/rss/articles/CBMixAFBVV95cUxQamJ2ZXhfbnN5YTJfbFh2ZUt3NktpQWZJQzdiaG94c1U3SmRNMXZqd2Vya3JRWUV0YWpxZS13c3ZkQTBtX252NEd2MWtFTUo2Rlp4X2ZnQXVZZ0hsc3VWZ3NuVE9HYXBBZ2dVbWJhNGhDUmZjOWdvNm0yUk0xQ2dIZEloLWJWcC1PUVc0Q0Y1NklWaHQ1MmIzX0lWakI4dG82dFpiUDRTaUpfVUpwTzdSb2oyWHhXNWI0Vk5SMTk3eTZjbnE4?oc=5). Housing coverage has reported younger buyers showing a greater willingness to cut spending to hit their goals (HousingWire, https://news.google.com/rss/articles/CBMihAFBVV95cUxQSDVoTjRtd3laTmpMRGd0LTEyRURTWGtIMVhwUmdweTl3SzZGNG9nSU9NMExvTWJWc1RCQ0UwZ184cXVQajNDQW9lTVN0ZndEVFZScThXUElMVFpuZVoyLTlOUW5WNGp4a1BRQ0NtcWxFdlF2YUVuRnlTTWpQeF9YRXIwYl8?oc=5).
Those are consumer signals, and they are directional rather than a forecast for your category. But the same behaviour shows up in B2B procurement: renewal reviews that used to be automatic are now itemised, and someone is asked to justify the line. Treat them as a caution, not a number to plan with.
The practical consequence is narrower tolerance. A three per cent rise that would have passed unnoticed in 2023 may now trigger a conversation, and a conversation is where discounts are born.
What the spreadsheet cannot tell you
Margin arithmetic answers one question: what we earn if nobody leaves. It cannot answer the three questions that decide whether the increase works.
- Who absorbs it? Some customers barely notice a rise because the product is embedded, cheap relative to what it replaces, or renewed by someone who does not see the invoice.
- Who switches? Churn risk is never evenly spread. It concentrates in a cohort you can usually name in advance.
- What buys back the goodwill? Almost every base contains a feature, service level or commercial term that customers value more than the money you are asking for. Most companies never find out which one.
A price increase is not a finance decision with a customer consequence; it is a customer decision with a finance consequence.
The four-week sequence
Week 1: segment the base on behaviour, not firmographics
Start in the CRM and billing data, not in the survey. Build segments from things that predict price sensitivity:
- Tenure and renewal history
- Product depth: how many modules, seats or services they actually use
- Usage intensity relative to what they pay
- Discount history and how hard the last negotiation was
- Whether a competitor is already in the account
You are looking for four to six cohorts, each large enough to be worth a separate decision. If the segmentation does not change what you would do, it is decoration.
At the same time, sample the price rise across the base. A flat percentage produces very different absolute increases, and the customer reacts to the absolute number on the invoice.
Week 2: model the trade-off with your own customers
Field a short exercise to the base. Two instruments do most of the work:
MaxDiff when you need to rank what matters. Customers choose the most and least important items from small sets, repeated. It is quick, it tolerates twenty or more attributes, and it exposes which features are genuinely load-bearing rather than merely liked.
Conjoint when you need to price. Customers choose between whole packages that vary on price and features simultaneously. Because they are forced to trade, you get willingness to pay that is anchored in a decision rather than in an opinion.
Run it on your customers. A panel sample tells you what a category-shaped audience thinks; it does not tell you what the accounts on your renewal list will do. Fifteen to twenty minutes, incentivised properly, sent from someone the customer recognises. Response rates from an engaged base are usually far better than people expect.
Keep the framing neutral. You are exploring packaging options, not announcing a decision. If the exercise reads as a warning letter, you have started the negotiation a quarter early.
Week 3: stress-test the highest-risk cohort
The model will identify a cohort where the switching probability climbs sharply. Do not act on the model alone. Interview ten to fifteen customers in that group.
Ask what they would do, who they would call, how long a switch would take, and what would have to be true for them to stay. You are checking three things:
- Whether the alternative they name is real and available
- Whether the switching cost they describe is as low as the model implies
- Whether the buy-back feature the model likes is something they would actually notice
This is where most stated-preference work gets corrected. Survey respondents overstate their willingness to leave; interviews reveal how much friction sits between irritation and action.
Week 4: phase the increase
Now turn the evidence into a plan that differs by cohort. In practice it usually looks like this:
- Absorbers: take the full rise, no sweetener, standard notice.
- Middle ground: full rise paired with a feature or service upgrade the MaxDiff identified as high value and low cost to deliver. The increase becomes a repackaging rather than a demand.
- High risk: a smaller rise, a longer notice period, or a staged increase over two renewals. Brief account managers with the specific objection the interviews surfaced and the specific answer that worked.
- Strategic accounts: hold, and take the decision individually with revenue context.
Then instrument it. Track churn, downgrade and discount concession by cohort against what the model predicted. The second price rise is much cheaper to plan than the first, because you now have a calibrated model instead of an argument.
What this costs you not to do
The usual objection is time. But a price increase is typically announced eight to twelve weeks before it takes effect, and the four weeks described here fit inside that window with room to spare.
The alternative is discovering your price elasticity through cancellations, which is the most expensive research method available and produces the data after you can act on it.