The Funded Competitor Taxes the Company That Stays Disciplined

A competitor’s round is a price change in your market. You did not set it, you cannot bill for it, and inside a quarter it shows up in what a customer costs, what an engineer costs and what a distributor asks for. Work out which fronts you concede on purpose. Then put what is left on the two you can hold.

Three numbers settle that. What it would cost them to take each front. What you get if you contest every front at once. And how many months their aggressive phase can run before the cash behind it is gone.

Why these three, and what got folded in

This piece runs the Wire Model: score the decision’s features, route to a small ensemble, then force it to produce dated actions. The scores that drove the routing:

• Exposure topology (0.9). Their money cannot reach your customers directly. It travels through a countable set of intermediaries, and the count differs on every front.
• Strategic actors, finite resources, many fronts (0.9). Two players, one contested market, budgets that are not close. That is a solved class of game, and the solution is unkind.
• Contagion and spread (0.8). A bid does not stay with the bidder. It moves into auctions, pay bands, and what a customer believes a fair price is.
• Historical-analog density (0.7). This market has run the experiment in ride-hailing, in B2B distribution, and now in fixed broadband.
• Cognitive over-reaction (0.7). Real, and it earns no card. Nobody decides to match. They approve one exception at a time.

That routes to network centrality (what a front costs to take), Colonel Blotto with its General Lotto relaxation (how to spend across fronts), and SIR contagion (how the pressure travels and when it stops). Complex, equilibrium and cycle-regime. Three different failure directions, which is the entire reason to run three.

Two members were folded rather than shipped. The behavioral layer lives in the blind spot, because the matching reflex changes no allocation, it only explains why a good allocation gets abandoned. Signaling sits inside the investor portfolio, because for the founder the signal is a by-product of the concession list and not a lever you pull separately.

The framework: name the fronts, price them, then spend on two

1. The map: a front is priced by its gatekeepers, not by its customers

Begin here, because the next two models need this output.

Revenue does not sit in a market. It sits behind intermediaries. A distributor’s book. A procurement officer. A telco’s aggregator shortlist. An estate WhatsApp group. A SACCO chair. An agent with three years of tenure. A funded rival is not buying four thousand shops. They are buying the eleven people those shops call.

So price each front by node count, not by revenue. Kenya shows the full range inside one regulatory filing. Fixed data subscriptions are concentrated: Safaricom holds 35.6% with 815,037 subscriptions, Jamii Telecommunications 20.4%, Wananchi 11.8%. Mobile money runs the opposite way, spread across 480,216 registered agents serving 48,630,797 subscriptions.¹ One of those structures can be reshaped by a single cheque. The other cannot be bought at any round size an early-stage company will see.

Twiga Foods read its own graph and acted. After years of owning farms, fleets and supply chains, it took controlling stakes in three Kenyan FMCG distributors in April 2025, picked up eight distribution centres, and kept their management in place.² That buys nodes rather than customers, which is also what a well-funded rival does to you when your front has few enough of them.

Count the people who would have to say yes for half your revenue to move. Under ten, that front is cheap.

Network centrality, the map lens

Assumes: demand reaches you through identifiable intermediaries whose reach you can count.

Fits because: exposure topology scored 0.9, and node counts differ by orders of magnitude across one company’s own channels.

Breaks when: the graph rewires faster than you map it. A new aggregator, a platform rule or a licence regime can create a decisive node inside a month.

Counteracts: valuing a front by the revenue sitting on it.

May reinforce: loyalty to people who are warm and no longer central.

2. The allocation: half their budget does not buy half the market

Now the unkind part. Blotto games model this exact situation: two players spread finite resources across contested fronts, and the larger allocation takes each front. Optimal play requires mixed strategies, because any deterministic allocation is exploitable by an opponent who can anticipate it.³

The General Lotto relaxation gives a closed form. With your budget X against their Y, and X below Y, your equilibrium share of the total contested value is X divided by 2Y.³ Sit with that. A rival with four times your money does not hold you to a fifth of the market. You get an eighth. The result turns only on the budget ratio, not on how many fronts exist or what each is worth. Contesting everything at once is a game whose answer has already been published.

Two exits appear in the same literature, and a disciplined company can build both.

The first is favouritism, which is a pre-allocation. Formally, a front carries a parameter added to your spend before either side bids. In practice it is a signed exclusive, a licence printed on the invoice, an integration already in production, three years of repayment history on a lending book. Where that parameter is large their money buys very little, and the literature is explicit that a player can engineer it deliberately ahead of the engagement rather than wait to inherit it.⁴

The second is concession, and here the theory is sharper than the folklore. Removing resources from the contest to demonstrate restraint has been shown worthless: no game instance exists in which a player improves its position by unilaterally cutting its own budget. Conceding the contested value of a front does help, across a measurable set of cases, because it redirects the opponent’s budget toward a different competitor.⁴

That result carries a condition worth reading closely. The concession pays by sending their spend somewhere else, so in a two-company category there is nowhere for it to go. Most African categories hold three or four funded players plus a large informal incumbent, so there usually is.

Only one kind of restraint earns anything. Give up a front in public, and their money goes looking for someone else to beat.

Colonel Blotto, the allocation lens

Assumes: both sides hold budgets they will spend, fronts are contested simultaneously, and the larger allocation wins each one.

Fits because: strategic actors with finite resources across many fronts scored 0.9.

Breaks when: your rival is not buying share at all. If they are buying an input you both need, the fronts stop being independent and conceding one removes your ability to contest the rest.

Counteracts: the belief that effort substitutes for capital.

May reinforce: premature surrender of a front whose pre-allocation you never measured.

3. The clock: their aggressive phase has a length you can compute

The pressure is not weather, and it is not permanent. Model it as transmission with a recovery rate.

Transmission is structural in at least one channel. The generalised second-price auction selling your paid traffic has no equilibrium in dominant strategies, and truth-telling is not an equilibrium of it.⁵ Every bid there is a best response to the other bids. A rival with a different cost of capital does not merely take slots. They move the number everyone left should rationally submit. Pay bands travel the same route through recruiters, and a discount seen once resets what a customer thinks normal costs.

Recovery is the part founders skip. The infectious period is their runway: round size, less whatever reserve their board will insist on, divided by monthly burn. Savanna Fibre is running the play in Kenya now, listing 100 Mbps at KES 2,000, about $15.40, undercutting the market leader’s equivalent tier by roughly 80%, while Wananchi charges five times Savanna’s rate for the same speed.⁶ The question worth putting on a page is the date it stops.

Sometimes the answer is days. Taxify entered Lagos against Uber with a 40% rider discount and up to 15% higher driver earnings, then cut it to 25% inside ten days.⁷

The wave also ends when the pool of companies willing to match is used up. Immunity here is a written pre-commitment, made before the pressure arrives rather than during it.

You are not losing to a better company. You are losing to a longer clock.

SIR contagion, the timing lens

Assumes: the spend spreads through shared channels and stops at a rate set by the funder’s cash.

Fits because: contagion scored 0.8. Auctions, pay bands and reference prices are all transmission surfaces.

Breaks when: the funding is not a wave. A strategic parent or a listed incumbent carries no burnout date, and every duration estimate here turns into fiction.

Counteracts: treating a rival’s price as the new market price.

May reinforce: waiting out a competitor who is genuinely building something durable.

GEER: the levers, ranked by what their money cannot buy

Four channels carry the exposure: acquisition cost, pay, distribution access, price expectation. Pull the cheap reversible ones first.

1. Write their clock on one page. Announced round size, visible headcount, a burn estimate, the month it ends. An afternoon.
2. Rank every revenue cluster, team and channel by cost to flip. The cheque they would write to move it, not the revenue on it. Two days.
3. Draft the concession list. Name the fronts you will not defend. Circulate it internally so nobody spends on them by reflex. One meeting.
4. Move the conceded budget, do not delete it. Reallocate onto the two fronts with the largest pre-allocation, in the same week.
5. Convert a relationship into a document. Sign the exclusive, ship the integration, get the licence onto the invoice. Weeks, and it raises the parameter that discounts their spend.
6. Raise price on the segment that never asks about price. Their discount is doing your segmentation work for free.
7. Lock the three people whose exit closes a front. Equity with a real path, a named scope, a decision right. Things a larger balance sheet cannot copy quickly.
8. Randomise what remains. Vary spend across defended fronts week to week. A predictable allocation is the one they can plan against.
9. Match, on one front only, and only where you already hold the pre-allocation. This is the expensive lever and it goes last.

No-lever flag. If every front you have is purchasable and none carries a pre-allocation you can name in a sentence, capital beats you and no reallocation prevents it. Say that to your board in your words before they read it in someone else’s diligence memo.

RADAR: what to settle before their spend lands

DO NOW, by T+3 days. Reversible, and correct whatever they do next.

1. Their clock, on one page, dated and shared with your leadership.
2. Every cluster ranked by cost to flip, with the named gatekeepers listed beside it.
3. The concession list, signed. Two fronts defended, the rest released.

HEDGE, by T+14. Bounded cost against the tail where they last longer than you modelled.

1. One exclusive or one integration signed on your highest-value defended front.
2. Retention conversations with the three named people, before a counteroffer exists to respond to.
3. A tested price rise on your least price-sensitive segment.
4. One channel opened that they cannot bid into: an owned list, an agent cohort, a distributor relationship with a term.

DEFER AND TRIGGER. Irreversible, so wait, and commit the observable trigger today.

1. Defer: matching across the board, a new market, a rebrand, any headcount that only pays back at their volume.
2. Trigger to match on one front: you hold a nameable pre-allocation there, their clock has under two quarters left, and contribution margin on that front stays positive at the matched price.
3. Counter-trigger at T+28: if cost to flip has fallen on two defended fronts, you defended the wrong ones. Redraw the list rather than raise the budget.

If you are underwriting the category. DO NOW: ask for the concession list in writing. A founder who cannot name the fronts they are not defending is defending all of them, and the General Lotto share tells you how that ends. HEDGE: put the funded entrant’s round size and burn estimate in your memo, because their clock, not the incumbent’s plan, sets the next four quarters. DEFER: the markdown. Acquisition cost that moved with a rival’s round is a price change on a schedule, and it reverses. Capital efficiency only reads as a signal when the founder can name what it bought. Efficiency with no defended front reads as an inability to grow, and that reading is often correct.

CHAIN: how funded pushes have actually ended in this market

Group by mechanism rather than by sector label. The class is every category where the price of a customer got set by one company’s cost of capital instead of by what a customer earns. Ride-hailing in Lagos belongs. B2B distribution in Nairobi belongs. Fixed broadband in Kenya belongs today.

The base rate for that class is documented and recent. African tech raised $4.65 billion across 941 deals in 2022, then $2.92 billion across 554 deals in 2023, a 37.2% fall, then $2.24 billion in 2024. Between January 2023 and March 2026 the ecosystem recorded at least 56 layoff events and 4,948 disclosed job losses. The five largest events accounted for 71% of those losses, and Copia Global and KOKO Networks alone were 43% of them. Both have ceased operations.⁸

The usual ending is not one winner standing over a taken market. Wasoko, valued at $625 million after a $125 million raise, merged with MaxAB. Between them the two had raised close to $245 million.⁹ Two funded rivals in one category more often become one company than one survivor.

Present conditions push in a specific direction. In the first four months of 2026 African startups raised $887 million across 84 disclosed deals, against $803 million across 173 deals in the same window of 2025, with no round above $100 million recorded.¹⁰ More money in fewer hands, which cuts both ways. Your funded rival is likelier to be the only large cheque in the category, so the asymmetry is sharper than it would have been in 2022. The pool of companies with cash to match them is also smaller, so the escalation burns out faster.

Strip out what was going to happen anyway before you size a defence. Some of the share that left this quarter was leaving regardless, priced out by a currency move, a levy, a bad release, a founder who stopped visiting distributors. Budget only against the movement their money caused.

Matrix-break flag. All three models assume you and your rival are contesting the same customers across separable fronts. That assumption dies if they are not buying share but buying a shared input: exclusive supply, the only two data engineers in your city, a licence category, a distributor’s entire book. Then the fronts are not independent, conceding one can remove your ability to contest the others, and the concession list becomes the most dangerous document in the building. The second break runs through the rails. If the funded company owns the layer you sell on top of, this is not a contest across fronts, it is a landlord decision, and you need a different piece.

The parts of this the models never see

Four, and each of them is load-bearing.

Their mandate. You are computing a clock from a press release and a headcount page. One board meeting extends it. One down round ends it. Treat the date as a planning assumption with error bars.

Whether your pre-allocation is real. The parameter that makes a front defensible is easy to assert and hard to test. You believe the distributor is locked. The distributor has a phone, and somebody with more money is going to use it.

The match reflex. Concession lists die through exceptions. One key account, one keyword, one counteroffer, and by the fourth exception you are contesting every front again at X over 2Y, having never made the decision to.

What a subsidised customer is worth after the subsidy. Nobody has that number for your market, including the company writing the cheque. It is the whole argument for conceding price-led fronts rather than buying them back, and it remains an argument rather than a measurement.

Do the thing that survives all four. By T+3, produce two pages. One carries their clock. One carries the fronts you are not defending, with your signature on it. Every other move in this piece is downstream of a list you have not yet been willing to write.

Sources and notes

1. Communications Authority of Kenya, “First Quarter Sector Statistics Report for the Financial Year 2025/2026 (1st July to 30th September 2025).” Full report. Fixed data subscription market shares of 35.6%, 20.4% and 11.8% for Safaricom, Jamii Telecommunications and Wananchi Group are stated in the fixed data section, with Safaricom’s 815,037 subscriptions in the accompanying table. Registered mobile money agents (480,216) and mobile money subscriptions (48,630,797) are in the report’s key statistics summary.
2. “Twiga’s pivot to asset-light model begins with acquisition of Kenyan distributors.” TechCabal, 20 May 2025. Report. Controlling stakes in Jumra, Sojpar and Raisons acquired in April 2025; eight distribution centres across Central, Coast and Western Kenya; original management retained.
3. Paarporn, K., and Marden, J. R. “Move Over, Prisoner’s Dilemma: Colonel Blotto has arrived.” arXiv:2603.25979v3, 9 June 2026. Full text. The requirement for mixed strategies, and the statement that a deterministic allocation is exploitable by an adversary who can anticipate it, are in the equilibrium analysis section. Theorem 3.1 gives the General Lotto equilibrium payoff to player X as the total contested value multiplied by X/2Y when X is below Y. The survey attributes the underlying Colonel Blotto solution to B. Roberson, “The Colonel Blotto Game,” Economic Theory 29(1), 2006, 1-24.
4. Same survey, arXiv:2603.25979v3, sections 6.5 and 7. The favouritism formulation and the observation that a player can deliberately engineer the pre-allocation before the engagement are in the pre-allocation discussion. Theorem 7.1 states that no game instance exists in which a budget concession benefits the conceding player; Theorem 7.2 states that a positive measure set of instances exists in which a value concession does, by redirecting the opponent’s budget toward another player. Both are adapted from Chandan, Paarporn, Kovenock, Alizadeh and Marden, “The art of concession in General Lotto games,” Games and Economic Behavior 155, 2026, 89-106.
5. Edelman, B., Ostrovsky, M., and Schwarz, M. “Internet Advertising and the Generalized Second-Price Auction: Selling Billions of Dollars Worth of Keywords.” American Economic Review 97(1), 2007. Full text. The finding cited here is the strategic one rather than the charging rule: “Unlike the VCG mechanism, GSP generally does not have an equilibrium in dominant strategies, and truth-telling is not an equilibrium of GSP.” It is in the abstract and developed in Section II.
6. “Savanna Fibre sparks price war in Kenya’s stagnant broadband market.” TechCabal, 31 March 2026. Report. The KES 2,000 (about $15.40) 100 Mbps plan, the roughly 80% undercut of the market leader’s equivalent offering, and Wananchi (Zuku) charging KES 10,000 for the same 100 Mbps are quoted from the piece.
7. “In just 10 days after its Lagos Launch, Taxify might be backing away from the price war with Uber.” TechCabal, 28 November 2016. Report. Launch terms of a 40% rider discount and up to 15% increased driver earnings, reduced to a 25% discount within ten days.
8. “Inside Africa’s tech layoffs: What the data says (2023-2026).” TechCabal, 29 June 2026. Report. Funding of $4.65 billion across 941 deals in 2022, $2.92 billion across 554 deals in 2023 (a 37.2% fall) and $2.24 billion in 2024; at least 56 layoff events and 4,948 disclosed job losses between January 2023 and March 2026; the five largest events at 71% of disclosed losses; Copia Global and KOKO Networks at 43%, both since ceased.
9. “Wasoko and MaxAB say merger will create a clear e-commerce leader with tens of millions of runway.” TechCabal, 23 December 2023. Report. Wasoko last valued at $625 million after raising $125 million; the two companies had raised almost $245 million between them. Completion: TechCabal, 27 August 2024.
10. “Will H1 2026 cross the $1B mark? Funding hits $887M despite deal slump.” TechCabal Insights, 7 May 2026. Report. $887 million across 84 disclosed deals in January to April 2026 against $803 million across 173 deals in the same window of 2025, with no deal above $100 million tracked so far in 2026.