How to Hold Price in Late-Stage B2B Negotiation Without Losing the Champion

Five moves, in this order: diagnose the ask, write the cost line, price the trades, test any rival, then brief the champion. The judgment in each is yours. None of them adds a value figure, and none can tell you what to concede.

Step 1. Diagnose the ask before you answer it

About this example. Bolna is a real company. Bolna has not reviewed this guide. Facts about Bolna come from public sources, linked alongside with the date I checked them. Every number marked illustrative is an input I chose to show the method; it is not Bolna's data, and the deal is an exercise, not Bolna's real strategy.

The "I" here is an exercise role, not a seller at Bolna. Bolna is a Bengaluru voice AI company that raised a $6.3 million seed round in January 2026 and, Inc42 reported, charges "per minute of orchestration". Its pricing page says "Telephony and the platform fee are added per minute", with no amount, and offers "volume discounts based on your committed usage".

The deal, all illustrative, in dollars: a mid-sized consumer lender, an NBFC, moving payment reminder and verification calls to voice agents. The commit is 2.4 million minutes a year at 4.5 cents, telephony and platform fee included: $108,000 a year, with 60 calls at once at peak. The Head of Customer Operations is the champion, the CFO signs, and procurement asks for 20% off before the seller's 31 December quarter end. The buyer bought the $500 pilot in September, so procurement holds an invoice at 4.2 cents. About 10% of pilot minutes passed to a person. The buyer estimates its outsourced call centre at 8 cents a minute, $192,000 a year for these minutes.

I set both rates against the public record, alongside. Inc42 reported on 15 July 2026 (the page may ask for a free login) that enterprises on "significantly larger, long-term volumes" can negotiate Bolna rates "as low as ₹1.75-₹2 per minute", and that human calling costs "around ₹7-₹8 per minute" before hiring, training and attrition. It does not say whether the rate includes telephony, or what volume earns it.

Before any number, I ask two questions: who set the 20%, and what happens if it is not met? The answer decides the rest, as alongside. Here I assume a scorecard target, a pilot invoice, and a rival's rate card nobody has raised.

Two rules touch this deal. India's telecom regulator asked large NBFCs to move service and transactional calls to 1600-series numbers by 1 February 2026 and the rest by 1 March (Outlook Business, 19 Nov 2025, checked 31 Oct 2026). RBI's recovery directions for NBFCs (checked 1 Nov 2026; retention wording rechecked 3 Nov) start on 1 January 2027, the month of this rollout, and set call hours, recording and six-month retention for dues from borrowers in default, with records kept until disposal for cases before a court. My reading, not legal advice: reminders before the due date sit mostly outside, overdue calls sit inside, and neither text says how a voice agent is treated. Settle who supplies the numbers and who holds the recordings before calling a rate all-in.

Step 3. Price the trades and set the floor

My answer to the pilot invoice, on the illustrative deal: "The 4.2 cents is a one-time pilot: 12,000 minutes, up to 25 calls at once, about two days of your volume. Your rate covers 2.4 million minutes, 60 calls at peak, telephony and platform fee included. That is the committed-volume price. A lower rate comes from more committed minutes or a longer term." If telephony alone cost what the rival publishes, about 0.43 cents, the pilot would be about 4.6 cents and the top tier about 4.9, both above 4.5. If your real figure does not put your commit below the pilot, drop that argument.

For a first list of trades, input to the pricing negotiation guide, all illustrative: scenario discount request; deal value $108,000; discount 20%; business_model "transactions"; value_delivered, the step 2 CFO line.

What came back: "Revenue at Risk" of $21,600, a "Post-Discount Value" of $86,400 and four trades. $86,400 is 3.6 cents, under each rate on Bolna's page, about level with the rival's published low end, and above Inc42's "as low as". So the ask is not absurd. Skip the printed reframe: it is a template, so write your own.

The four trades follow business_model, not this deal: "Committed monthly volume", "A longer term", "Exclusivity for a product line or region" and "A case study after go-live". I keep the first two. Exclusivity earns nothing against a call centre, and a case study comes only with another trade. My ladder is alongside.

Term is a give, so price it. Assume, for illustration, a written proposal with a 7% annual uplift: $115,560 in year 2 and $123,649 in year 3. Holding year 2 at 4.5 cents gives up $7,560. That is cost avoidance, which many procurement teams report apart from hard savings, so ask what their scorecard counts. If your proposal has no uplift, this saving does not exist; never add one in order to waive it. Your ladder then starts at the added-minute rate and the credit.

Procurement will see that year 1 moves by $0, $3,000 or $5,400 against a $21,600 ask. So offer 2 follows the public logic that committed usage earns discounts: an illustrative 600,000 added minutes at 4.0 cents, $24,000 a year, for a blended 4.4 cents. Offer it only if the buyer's forecast reaches 3.0 million. At the 2.64 million in step 5, it costs the buyer $13,200 more than offer 1, and unused commit lands on the champion.

The floor's credit is one-off, so the order form still says 4.5 cents. If all three years run, it costs me an illustrative $28,609 against the 7% proposal ($5,400 credit, $7,560 in year 2 and $15,649 in year 3), and the next price starts from $108,000, not $123,649. The 20% would cost about $69,400, 20% of the $347,209 three-year proposal. The usage cap is a give as well: I assume 240,000 minutes a year above the commit at 4.5 cents, $10,800, which I forgo if the champion approves no overage.

Term also costs the buyer. Inc42 reports that models and telephony "continue to become cheaper", and this vendor is seed-stage. So the floor carries exit terms. Each is a give, so each is priced, all illustrative. Step-out: from month 24, if collections on matched accounts run 2 points below a group left at the call centre, on a baseline both sides sign before go-live. The buyer repays the credit. My worst case bills $216,000, not $318,600: the same as offer 1, so year 3 is only as firm as collections. Rate review at month 18: added minutes only, never below 4.0 cents, at most $1,800. Return of data: an assumed $2,000 of engineering time. Offer 1 has none: exit terms are what the third year buys.

If procurement says no at the floor, I stop trading, and the champion takes the decision to the CFO with the step 5 brief. If the CFO holds at 20%, I do not lower the rate. I offer a smaller order at the same rate, say reminders only, so the January rollout still ships. If that fails, I let the deal go.

Losing a $347,209 proposal costs more revenue than the 20% does. Walking is right only if the lower rate would spread to other accounts or break your margin floor.

This is the point where I would want a second pair of eyes: a quarter-end concession is about to cross your floor. Bring me your priced trades, approval authority, walk-away rule and CFO brief, and we will work it through in 45 minutes.

Step 4. If a rival is real, count it the same way

The call centre sum cannot answer a rival, because it flatters every voice vendor equally. Sarvam, a large Indian rival, publishes ₹3.00 a minute on its Business plan, with telephony on top at ₹0.20 a 30-second pulse, before GST. Enterprise is custom from "2 lakh minutes a month", which is this deal's volume. That is ₹3.40, about 3.7 cents, or 0.8 cents under my illustrative commit: $19,200 a year, close to the 20%. A signed quote could be lower. If my price is fixed in dollars, the rupee moves this: Inc42's feature puts it past ₹95 to the dollar, and at ₹95 the gap is about $22,100, above the $21,600 ask, so the rival alone would cover the 20%.

Confirm phone-number rental, forwarded calls and test-run charges in the rival's quote before treating that per-minute gap as a saving.

Compare per completed call, not per minute. That page rounds each call up to a 60-second pulse, while Bolna's pilot bills in 30-second pulses, so the same call can count differently.

For a draft checklist, run the guide again with scenario competitor pricing. It does no sum with the rival's price, so the sum is yours.

My sum, all illustrative: each point of minutes passed to a person costs the buyer $1,920 a year, 24,000 minutes at 8 cents. So the rival's $19,200 is worth ten points, about 11.5 at ₹95. Delay matters less than you hope: the year 1 saving is $54,000, $4,500 a month, so a three-month rerun costs $13,500 against $19,200 a year.

So test, on the buyer's own calls, side by side, scored by language, task and promises to pay. If procurement quotes Inc42's "as low as" instead, that sentence is tied to larger, long-term volumes, so ask what volume and term they can sign. If your real floor is near that figure, rebuild this example on your own numbers.

Step 5. Write the champion's brief to the CFO

Only now does the champion enablement kit help, with the illustrative inputs alongside. Its memo is mostly my inputs handed back in a layout. I keep the layout and the value lines, and rewrite the rest.

It opens "I request validation of the business case for Bolna Enterprise", and it calls the first value point "the point that decides it", so the decision and the order are yours. Start with the fixes table. Add the trade once by hand: "Vendor offer: two-year term, year 2 held at the year 1 rate, usage capped (all illustrative)."

My arithmetic for the CFO is in the sum table. Offer 1 is a two-year term, so I assume a renewal in year 3 at the 7% uplift on the held base, $115,560. The line I add: "Over three years, with year 3 at the 7% uplift, buying costs about $422,000 on our own estimate, against about $576,000 at the call centre (all illustrative)." Quote about $409,000 only once the floor is agreed.

The buying side keeps usage above the commit, so it covers 2.64 million minutes a year against 2.4 million. Year 1 there is an illustrative $138,000 ($108,000, $10,800 and $19,200), or 5.75 cents across 2.4 million minutes. So the case stops working if the call centre really costs about that, or 5.5 cents if hand-offs reprice with it. Neither side counts integration or quality review; get those figures before the CFO asks.

1. Treating the case for buying as a defence of the price.
2. Deciding the trade on the call instead of pricing it beforehand.
3. Discounting without a commitment in return.
4. Leaving a public rate unexplained, your own or a rival's.
5. Sending a tool draft unedited.

• One B2B Deal From MEDDPICC Discovery to a CFO-Ready ROI Case and a Dated Close Plan
• From Discovery Call to a Demo the Buying Group Remembers
• All 30 guides