The invoice arrives in the first week of August. It is for July, the busiest month of the year, when the website was flooded with people asking about availability, delivery dates and prices. The assistant did its job well. It answered hundreds of visitors. And because the plan charged per conversation, the bill for that month is roughly ten times what it was in February.
Nothing went wrong. That is simply how per conversation pricing works. The better your season, the more you pay, and the busiest months are also when you have the least time to notice.
This post follows one invented business through a full year to show the pattern clearly, then compares the same year on a flat monthly plan.
Meet the example business
Every figure here is hypothetical. Picture a hot tub hire and servicing company. Winter is quiet. Spring bookings start in March, the garden party and staycation months run from May to August, and there is a smaller December bump for Christmas and New Year hire.
Most of its website enquiries are the same handful of questions: is this weekend available, how big is the tub, how long does it take to heat, what does delivery cost to my postcode, do I need a garden tap nearby. A website assistant answers these well and captures hire requests for the team to confirm.
For the usage-based option, we will assume a rate of $0.90 per conversation. For the flat option, we will use a $59 monthly plan with a 9,000 message allowance, which is SpideyChat's Pro plan. You should replace the usage rate with the real quote in front of you.
The year, month by month
| Month | Conversations | Per-conversation bill at $0.90 | Flat plan bill |
|---|---|---|---|
| January | 60 | $54 | $59 |
| February | 70 | $63 | $59 |
| March | 120 | $108 | $59 |
| April | 260 | $234 | $59 |
| May | 480 | $432 | $59 |
| June | 620 | $558 | $59 |
| July | 700 | $630 | $59 |
| August | 540 | $486 | $59 |
| September | 220 | $198 | $59 |
| October | 110 | $99 | $59 |
| November | 150 | $135 | $59 |
| December | 310 | $279 | $59 |
| Year | 3,640 | $3,276 | $708 |
In January the usage-based plan is marginally cheaper. In July it costs more than ten times the flat plan. Across the year, on these numbers, the difference is $2,568, and $1,870 of that gap opens up between May and August.
If the business pays the flat plan annually, at $49 a month, the year costs $588.
Why the timing hurts more than the total
The annual figure is striking, but the shape of the bill causes subtler damage. Four effects show up in seasonal businesses on usage-based pricing:
- A bill that punishes success: a warm weekend, a good advert or a mention in a local group all raise the cost of answering the people they bring in.
- A reason to turn things off: owners watching the meter start disabling proactive greetings or limiting the assistant to certain pages in peak season, which is exactly when those features catch the most enquiries.
- Budgets that cannot be set: you do not know next July's bill until July is over, which makes cash planning guesswork.
- Growth that feels like a penalty: if next season is 30 percent busier, the assistant bill is 30 percent higher before you have earned a penny from it.
None of this makes usage-based pricing wrong for everyone. It simply means the risk of a busy period sits with you rather than with the provider.
The honest catch with flat pricing
Flat plans are not unlimited, and pretending otherwise would be misleading. They come with an allowance, and the allowance has to cover your peak month, not your average.
SpideyChat's allowance is counted in messages, and a conversation contains several of them. In our example, suppose a typical conversation runs to around 6 messages. July's 700 conversations would use about 4,200 messages. That is well beyond the Starter plan's 2,000, which is why the example uses Pro with 9,000. Sizing on the average month would have produced the wrong answer.
To size your own plan:
- Find your busiest month's website conversations from last year, or your best estimate.
- Open ten real transcripts and count the messages in each to get a typical figure.
- Multiply the two, and add a margin for a better season than last year.
- Pick the plan whose allowance covers that total. Current allowances are on the pricing page.
Peak months are when the assistant earns most
There is an irony in all of this. The months that produce the largest usage-based bills are also the months when an assistant is most useful. In July, the team behind our example business is out delivering and collecting tubs, the phone rings constantly and nobody has time to answer website questions promptly. Every enquiry the assistant captures in that month is one the business would otherwise have been too busy to catch.
A pricing model that makes those months expensive quietly nudges owners to use less of the tool when it is worth the most.
When usage-based pricing is the better deal
Be fair to the model. A business with a small, steady trickle of website conversations, say 30 or 40 a month with no seasonal peak, could pay less on a per-conversation plan than on a flat one. A business that only wants an assistant for a short campaign might too.
The test is not which model sounds cheaper. It is which one produces the lower and more predictable bill across your actual year. AI chatbot pricing models explained lays out each model's trade-offs, and what an AI receptionist really costs covers the other pricing structures you will meet.
Rebuild the table with your year
Open your website analytics or enquiry records and write down conversations or enquiries for each of the last twelve months. If you do not have monthly figures, use your quietest, typical and busiest months as a rough stand-in.
Then fill in a copy of the table above with the real per-conversation rate you have been quoted and the flat plan that covers your peak. If your business is seasonal, running an AI receptionist in a seasonal business and handling the holiday season enquiry surge are useful next reads before your next peak arrives.