Pricing Pages Lie: Working Out What a Live-Chat Conversation Really Costs Your SME
£35 a seat. £79 for the plan with routing. £0 for the free tier that gets you started. Every live-chat pricing page leads with a number, and almost none of those numbers tell you what a conversation actually costs to run.
That matters more than usual right now. Black Friday lands on 27 November, the Christmas cut-off conversation starts in earnest within a fortnight, and every extra pair of hands you put on chat in December has to be paid for in January. If you do not know your cost per conversation, you cannot work out whether the extra cover earns its keep.
So let us do it properly — the way an accountant would, not the way a pricing page would.
Seat price is the smallest line on the invoice
For a typical two-person sales and support setup, subscription fees come to well under 5% of total chat cost. Everything else is people, process and the gaps between them. Yet seat price is the only figure most teams compare before signing.
That is a bit like choosing a van on the price of the tax disc. The real number is the mileage.
The eight lines that make up a conversation
- Seats. Per agent, per month, sometimes per additional channel.
- Bot and AI usage. Billed per message, per resolution, per session or bundled into a credit pool — vendors are endlessly creative here.
- Agent handling time. Wages, employer National Insurance, pension, holiday cover, the lot.
- CRM admin. Tagging, logging, writing the follow-up note after the tab is closed.
- Handoff dead air. The minute and a half where the bot went quiet and nobody noticed.
- Quality and rota oversight. Someone has to read transcripts and build the shifts.
- Add-ons. Extra routing rules, a second number, the integration that should have been standard.
- Training and tuning. Onboarding a new agent or re-teaching the bot after a product change.
Most pricing pages quote line one and stay quiet about the rest.
A worked example: 480 conversations a month
Two agents covering chat alongside other work, a fully loaded cost of £19 an hour (salary, NI, pension, holiday cover), 11 minutes of average handling per conversation, and seats at £35 each.
| Line | Monthly cost |
|---|---|
| Seats (2 × £35) | £70 |
| Bot / AI usage | £27 |
| Agent handling time (480 × 11 min at £19/hr) | £1,672 |
| CRM tagging and follow-up admin | £228 |
| Dead air and post-handoff abandonment | £42 |
| QA and rota oversight (5 hrs) | £120 |
| Platform add-ons | £55 |
| Training and bot tuning | £90 |
| Total | £2,304 |
Divide by 480 and you get £4.80 per conversation.
Notice what happened. The number everyone argues about — £70 of seats — accounts for 3% of the cost. The argument worth having is about 11 minutes of handling time, because shaving two minutes off that saves more than switching vendors ever will.
What the number is actually for
Cost per conversation on its own is trivia. It becomes useful the moment you sit it next to conversion.
At £4.80 a chat and one sale in twelve conversations, your cost per chat-sourced sale is £57.60. On a £600 sofa that is a rounding error. On a £25 candle it is a business model problem, and the honest answer may be that chat should be answering delivery questions rather than chasing orders.
Run the same sum on your December numbers, when volume doubles and handling time stretches because everyone is asking about cut-off dates, and you will see why peak-season cover needs its own budget line rather than being absorbed on the quiet.
The handoff tax nobody puts on the spreadsheet
Here is the leak that hides between lines. A customer opens chat on a product page, the bot offers three helpful articles, they reply with a real question, and then — nothing. Forty seconds. A minute and a half. They open another tab, then another retailer.
You paid for the bot session, the queue time and the agent who eventually picked up a dead conversation. You also paid the far larger cost of a visitor who will not come back. In the model above that waste is £42 a month, which looks small until you notice it represents 41 conversations that had already raised a hand.
Dead air is cheap to measure and expensive to ignore. Timestamp every handoff request and every first human reply. The gap between those two figures is your handoff tax, and it is the single fastest thing most UK SMEs can fix.
A worksheet you can fill in over a coffee
Ask each vendor these, and ask yourself the ones they dodge.
- How are AI resolutions billed — per message, per resolution, per session? What counts as a resolution?
- Do we pay for chats that bounce in ten seconds with no agent involvement?
- Can we export conversation-level billing to check the invoice against reality?
- What is the average handoff wait across their UK customer base, not the best case?
- Which features cost extra — routing rules, post-chat surveys, extra channels, data retention beyond twelve months?
- How long does onboarding take, and who pays for those hours?
- What happens to our data if we leave, and in what format?
- What does a quiet February look like on this contract compared with a busy November?
Fill the same eight lines of the table for two or three platforms, using your own numbers rather than a vendor's illustration. The gap between headline price and true unit economics usually runs between 30% and 80% — and it rarely favours the platform with the cheapest seat.
Three things to do before signing
First, model both a quiet month and December. A contract that works at 300 conversations can look very different at 900. Second, ask for a billing export at conversation level as a condition of the trial, not a favour after it. Third, set a review date in the diary for 90 days out with the actual numbers in front of you.
A £4.80 hello is not a scandal. It is a price — and a price you can reason about. Know it, compare it against what a conversation earns you, and the decision stops being about seats and starts being about sales.
