Your impact numbers

Impact is the one screen that tries to answer “is this actually helping?” rather than “what happened?”. Five figures, a chart, and one estimate.

The estimate is the reason this page needs a help article. Four of the five figures are counted from records and cannot really be argued with. The fifth — Estimated support cost saved — is a multiplication involving a number you choose, and it is the one somebody will eventually put in a slide. Most of this page is about making sure that when they do, it holds up.

Where to find it: Impact, under Monitor, next to Reports. Admin access only. Depending on your plan you may see it locked or not at all.

The subtitle sets the scope: “What your assistant handled, and how much of it needed no one from your team.”

The window

Impact always reports the last 30 days, ending today. It has no date pills of its own, unlike Overview and Reports. It also compares that window against the 30 days immediately before it — always the same length, so like compares with like.

If nothing at all happened in the window, the page says “Nothing to report for this period” and shows no figures. It never fills an empty period with zeroes.

Conversations handled

The headline count: phone calls plus chat threads in the window.

• Calls are counted for the day they happened.
• Chats are counted for the day the thread started. A conversation that began three weeks ago and is still running counts once, on its first day.
• Everything is bucketed on the Indian calendar day (IST), so a chat at 1am belongs to that morning’s date, not the previous one.

The line beneath breaks it down: 312 calls · 44 chats.

⚠️ Every call counts, including your own testing. Calls placed from the Bots tester, from a website widget and from the public voice demo are all in this figure — it is not filtered by origin, and this screen has no origin filter. In your first weeks, when your team is trying things out, this number reads high. Overview can separate them with its Origin filter — see Building a report.

Below the figures, one chip compares the total against the previous 30 days: ↑ 12.4%, No change when the movement is under a twentieth of a percent, or New in this period when the previous window was empty. There is no percentage against zero, because there is no honest one.

The Conversations per day chart underneath plots the same two things by day.

Handled without a human

The number people quote. It deserves a paragraph of exactness, because it is narrower than it sounds.

It is a share of chat replies. Of every reply sent in this window that has a known author, the percentage written by your assistant rather than by a person on your team:

replies written by your assistant
─────────────────────────────────────────────────
replies written by your assistant + replies typed by your team

Read carefully, that means:

• Phone calls are not in it at all. They contribute nothing to the top and nothing to the bottom. A month of calls and one chat thread gives you an autonomy rate calculated from that one thread. The figure sits beside a call count, which makes this easy to misread — it is the single most important thing on this page to get right.
• A “human reply” means a reply typed from the Inbox, by anybody on your team. Kooday does not yet record which teammate, so they all count the same.
• Customers’ own messages are excluded. Only replies count, either way.
• Automatic sends with no author are excluded from both halves — a WhatsApp template going out on its own is neither a win nor a loss for this figure.
• The window is the reply’s own day, not the thread’s. A thread from last month that your team answered today puts a human reply into today’s figure.

The line beneath the percentage shows both halves — 1,204 by assistant · 98 by your team — so anyone can check the arithmetic. When no replies of either kind were sent, the figure is —, never 0% and never 100%.

Typical reply time

The median gap between a customer’s message and your assistant’s answer, in chat. Shown as seconds under a minute, then minutes, then hours; the line below says how many replies it was measured over.

The precise definition:

• Only your assistant’s replies count. A reply typed by your team measures how busy your team was, not how fast the product is, so it is left out.
• Median, not average — one thread that stalled overnight would drag an average and misrepresent the ordinary experience.
• Where a customer sent several messages before getting one answer, the clock starts at their last message. The reply is being timed against the finished question rather than against how long they spent typing. This is a choice, and it makes the figure slightly kinder than the alternative.
• Gaps over 24 hours are thrown out entirely. A thread picked up three days later is not a reply time in any useful sense, and capping it instead of dropping it would flatter the median.
• A reply just after midnight is still paired with the message that came before it on the previous evening.

Phone calls have no reply time, so they are not here either. With no chat replies at all, the figure is — and the line reads “no chat replies in this period”.

Talk time

Total minutes spent on calls in the window, rounded to whole minutes, with the number of completed calls beneath it.

This comes from the same daily totals your usage figures and your invoice are built from, so it matches Talk minutes on Overview — as long as Overview has no origin or bot filter applied, since Impact never filters by either. Note the two figures on this card are not the same measure: the minutes are for all calls, the count beneath is of completed ones.

How people reached you

Three chips, and each counts something slightly different — worth knowing before comparing them to each other:

Chip	What it counts
Phone calls	Calls in the window — the same figure as above
Chat messages	Text messages, both directions — yours and theirs
Voice notes	Voice notes sent as chat messages, again both directions

So this is a mix by channel, not a like-for-like comparison of volumes. One call and one message are not comparable units.

Estimated support cost saved

The one estimate on the page, and it is fenced off on purpose: its own card, an Estimate tag beside the heading, the formula printed under the figure, and the assumption behind it sitting on screen in a box you can type into.

The formula

It is a multiplication, and the card shows it in full:

1,120 conversations handled end-to-end by the assistant
  × ₹25.00 assumed cost per human interaction
  = ₹28,000.00

The left-hand number is not the autonomy rate, and it is stricter than anything else on this page. It counts conversations where your assistant replied and no person ever did — a single message from your team, at any point in the thread’s life, disqualifies the whole conversation. Not “mostly automatic”: no human touch at all.

Two further limits on that count:

• Chat only. Calls do not appear in it, so nothing you saved on the phone is in this figure.
• Counted by the day the conversation started. A thread opened just before this window does not count, even if your assistant handled the whole of it inside the window.

Both limits push the figure down. Whatever else is wrong with this estimate, it is not inflated by its arithmetic.

Setting the assumption honestly

The box reads Cost of one human interaction (₹). Until you change it, it shows ₹25 and the card labels it (platform default — set your own). That default is a placeholder, chosen to be unremarkable rather than impressive. It is not a claim about your business, and it should not survive your first look at this page.

To set your own, work out what one handled conversation costs you today:

1. Take the monthly cost of the people who answer customers — salary and everything that goes with it.
2. Take the share of their time spent on this kind of enquiry.
3. Divide by how many enquiries they get through in a month.

For a team member on ₹30,000 a month spending half their time on enquiries and handling around 500 of them, that is ₹15,000 ÷ 500 = ₹30. Type 30. The figure and the printed formula update as you type.

Three rules for keeping the number defensible:

• Use the cost of the people who would otherwise do this work, not a blended company-wide rate and not what an outsourced agency quotes.
• Do not include anything you still pay. If your team is the same size as it was, this is time released, not money returned — say so when you present it.
• Write down where the number came from. It is not stored anywhere and it is not shared with your colleagues: the box is per person, per visit, and it resets to the default when the page reloads. Two people can open the same screen on the same day and read different totals if they typed different assumptions.

How to talk about it

The card’s own footnote is the safest wording anyone can use, and it is on the screen for exactly that reason: it is an estimate, not a measurement, it counts only conversations no human ever touched, it depends on an assumption you supplied, and it does not represent billed or invoiced amounts.

The line that survives being questioned is “1,120 conversations went end to end without anyone from our team — at ₹30 each, that is ₹33,600 of handling we did not do”. The one that does not is “Kooday saved us ₹33,600”.

⚠️ Never put this figure next to your invoice as if the two were comparable. One is measured money that moved; the other is your own assumption multiplied by a count. Anyone can change the second by typing in a box.

Where revenue came from

If any order was paid in the window, one extra card appears listing revenue by channel and, where the customer had been targeted by one, by campaign.

This one is measured, not estimated — every rupee ties back to a recorded payment. Attribution is last-touch: an order whose origin cannot be traced is shown as unattributed rather than being shared out among the others. With no paid orders, the card is hidden rather than shown empty.

What is deliberately not here

You will not find a resolution rate or a delivery-failure rate on this screen, or anywhere else in Kooday’s reporting.

Nothing currently marks a conversation as resolved, and nothing currently marks a message as having failed to deliver. Both figures would therefore be the same fixed number on every account forever — 0% and 0 — presented as though something had been measured. They are left out until there is something real behind them, which is the same standard the rest of this page is held to.

Troubleshooting

”Handled without a human” shows — but we handled plenty of calls

The figure is a share of chat replies. With no chat replies in the window there is nothing to take a share of, and a dash is the honest answer. Your calls are still counted in Conversations handled and Talk time.

The percentage looks too good

Check the two counts beneath it. If it reads 12 by assistant · 0 by your team, it is 100% of twelve replies — accurate, and far too small a sample to quote. The same caution applies to a reply time measured over a handful of replies.

Conversations handled is much higher than the calls we know about

Testing counts. Calls from the Bots tester, from a website voice widget and from the public demo are all included, and this screen cannot separate them. Use Overview, press Phone, and compare.

The cost saved figure changed and nobody edited anything

The assumption is not saved. Reload the page, or open it on another device, and it returns to the ₹25 default. If two people quote different totals, this is why — compare the assumption printed on the card before comparing the totals.

The numbers here don’t match my report

They are answering different questions over different sets of records. Impact counts chats by the day the thread started; Reports counts calls only, and by the day of the call. Talk time is the one figure that reconciles exactly with your usage numbers — provided Overview is showing all origins and all bots.

I can’t see Impact in the menu

It is part of a paid capability and admin-only. If it is missing entirely or shown with a lock, check Billing and invoices or ask whoever administers your account.

Where to go next

• Building a report — the detail behind the totals, call by call.
• Billing and invoices — the measured money.