When Account Research Costs Dollars, Not Days: The Economics of Engineered ABX

The constraint tiering was built to manage

Account tiering exists for one reason: deep account research used to cost a strategist most of a day per account. Pull the firmographic context, map the buying committee, read six months of hiring and funding activity, and turn all of that into a point of view a rep can use on a call. Done well, that's most of a working day, for one name on a list.

Multiply that by a target account list of any real size and the math breaks fast. A 40-hour week buys maybe eight to ten accounts of thorough research, and that's assuming nothing else is competing for that strategist's time. So programs rationed the work. Tiering became the allocation mechanism: full account-based experience treatment, deep research and multi-threaded outreach, went to a small top slice, and everyone below that line got a lighter pass, or nothing at all. ("Engineered ABX," the phrase in the title above, is my own shorthand for the shift described below, not an established industry term, so read it as a label rather than a citation.)

Here's the part worth being honest about: tiering answered a narrower question than "which accounts matter most." What it answered was "which accounts can we afford to research this deeply, given what research costs." Most programs never separated those two questions, because for years they didn't have to. Once engineering-built account-based programs started treating research as a production problem instead of an analyst's craft, the two questions started to pull apart.

What moves when the per-account cost curve breaks

Something structural changed here, and it's not just that things got faster. The unit of production for a research pass shifted from person-hours to machine time and API calls. A speed improvement would mean the same strategist doing the same work in less time. What happened instead is the work moved to a different production model entirely, one where the cost of researching the next account doesn't track a person's calendar at all.

Here's an illustrative version of the math, with round numbers built to be swapped for your own. Say a thorough strategist-day of account research runs $500 in fully loaded time (illustrative only, not a real or internal figure). A research pass built as an engineered pipeline, pulling from public filings, hiring data, and licensed sources through a defined set of steps rather than an open-ended analyst task, might land somewhere in the range of $10 to $25 per account, all in. Illustrative again, and the real number for any given team depends entirely on which data sources it licenses and how deep the pass goes.

The ratio is the point, not the exact figures. Whether your real numbers are $400 and $15 or $800 and $40, the shape holds: the cost of researching one more account stops tracking a person's available hours and starts tracking a budget line you can dial up or down. A person-hours constraint is hard and physical. There are only so many hours in a strategist's week. A budget constraint is soft. You can spend more of it, reallocate it, or decide it isn't the binding constraint anymore.

And once the binding constraint moves, everything built to manage the old one is worth re-examining, not because it was wrong, but because it was priced for a different set of costs.

Second-order effect one: coverage stops being rationed

Once per-account cost drops by an order of magnitude, target-account-list size stops being a budget question. It becomes a fit question, which is the question it should have been the whole time.

Take an illustrative example. A program runs a quarterly research budget of $30,000 (illustrative, swap in your own number). At an old cost of roughly $500 per thorough pass, that budget covers about 60 accounts a quarter. Every other account on the target account list sits there, technically in scope, functionally invisible, because there was never enough budget to look at it.

Drop the per-account cost to an illustrative $15, and the same $30,000 covers 2,000 accounts. That's a different program, not a marginally bigger version of the old one. Accounts that used to sit below the research line, the ones that fit the ICP fine but never made the cut because the cut was set by capacity rather than fit, are now affordable to research. The question stops being "can we afford to look at this account" and becomes "does this account deserve attention," which is what a tiering system was supposed to be answering all along.

Worth a caveat here. Cheaper research doesn't mean every account deserves the same depth of treatment. It means the depth decision can finally be made on fit and intent instead of on what a strategist's calendar allowed that quarter. Coverage stops being rationed. It doesn't stop being a decision.

There's a sizing question hiding in here too. A bigger TAL isn't automatically a better one. Widen the list past what your ICP supports and you're just paying to research accounts that were never going to convert, at a lower price per mistake. The order-of-magnitude cost drop changes what a program can afford to look at. It doesn't change who's actually a good fit, and treating the two as the same thing is how a coverage expansion turns into noise instead of pipeline.

Second-order effect two: refresh cadence becomes a dial, not a project

Different signal types decay at wildly different speeds. Firmographic facts, headcount band, industry, HQ location, move slowly, and a quarterly refresh barely misses anything there. Buying signals like a new hire in a relevant role, a funding announcement, or a job posting for a function your product touches move fast. Some of that context is stale within weeks.

Most programs handled this with one calendar: refresh everything quarterly, or refresh Tier 1 monthly and everyone else quarterly, applied uniformly regardless of how fast any given signal decays. That uniformity was never a design choice. It was what a strategist's week could support. A full re-research pass cost the same strategist-day whether you were checking for a new funding round or reconfirming an org chart that hadn't moved in a year, so programs picked one cadence and applied it everywhere. Running a different schedule per signal type would have meant more strategist-days than anyone had to spend.

When a refresh costs dollars instead of a strategist's week, cadence can match the decay speed of each signal instead of one number applied to everything. Fast-moving signals, the kind covered by a job-change and funding signals approach, can run weekly or more. Slow-moving firmographic data can run quarterly or less, because refreshing it more often just burns budget on facts that haven't changed. The calendar stops being the constraint. The decay curve becomes the input.

To be clear, refreshing every signal at maximum frequency isn't the right takeaway here. Most teams shouldn't; the marginal value of catching a signal three days earlier rarely justifies the spend. The point is narrower: the cadence decision finally has room to be made on purpose, instead of defaulting to whatever a quarterly calendar happened to say.

Second-order effect three: the tier structure loses its reason to exist

Tiers were a resource-allocation tool for scarce research capacity. Tier 1 got deep, personal treatment because deep, personal treatment was expensive and had to be rationed toward the accounts most likely to justify the spend. Tier 2 and Tier 3 got progressively less, not because they mattered less in any absolute sense, but because the budget for depth ran out before it reached them.

Take away the scarcity tiering was built to manage, and the three-bucket structure stops having a reason to exist in its current form. Two things can happen instead, and they lead to fairly different programs.

The first: tiering collapses into a continuous score. Instead of three discrete buckets with hard cutoffs, an account gets a fit-plus-intent number that updates as new information arrives, and depth of treatment scales smoothly with that number instead of jumping at two arbitrary thresholds. Account scoring that already blends firmographic fit with behavioral intent is most of the way to this model already. Cheap research just removes the capacity reason for forcing that continuous signal back into three buckets.

The second: tiers survive, but they stop gating research depth and start gating something that's still scarce. Sales attention is still finite. Executive time on a call is still finite. Activation budget, the paid spend and event dollars behind a program, is still finite. A tier structure gating those resources still earns its keep. Gating research depth doesn't, not when depth is this cheap.

Being direct about what this argument claims and what it doesn't: tiering isn't disappearing, and no program needs to tear down its bucket structure tomorrow. Plenty of teams will keep three named tiers, because the language is embedded in how sales and marketing already talk to each other, and there's real value in a shared vocabulary. The narrower, more useful claim is this: whatever your tiers gate next, it probably shouldn't be research depth. That constraint is gone, or going. Something else is the real bottleneck now, and your tier structure should be built around that one instead.

What doesn't change

None of this means judgment stops mattering. If anything, it matters more, because the volume of researched accounts a team can look at goes up, and somebody still has to decide what to do with all of it.

Which accounts actually matter to the business, beyond fit and intent scores, stays a human call. Message strategy, what you say to a buying committee once you've researched it, stays a human call. Qualification criteria, the line between "researched" and "worth a rep's time," stays a human call. And the decision of where to spend scarce sales and executive attention, the resource that didn't get cheaper, stays entirely a human call.

Cheap research removes a cost constraint. It doesn't remove the need for someone to decide what the research is for. A team that expands coverage without also deciding what to do with the expanded output just trades one bottleneck (not enough researched accounts) for another (too many researched accounts and no way to act on all of them). For readers newer to the discipline, GTM engineering is as much about building the judgment layer on top of cheap data as it is about the data itself. Cheaper research is the easy part. Deciding what to do with more of it is not.

One honest caveat here, since it's easy to oversell this argument: none of it works if the underlying data sources are bad. An engineered research pipeline pulling from thin, stale, or mismatched sources just produces wrong answers faster and cheaper than a strategist ever could have. Cheap and wrong is worse than expensive and right, because cheap and wrong gets trusted at higher volume. The cost argument above assumes the pipeline behind it is sound.

A worked illustrative example: two accounts, one budget

Everything below is an illustrative walkthrough, built with round numbers to make the argument concrete. It isn't a real case or a client outcome.

Picture a program with a target list of 800 accounts and a quarterly research budget of $25,000. Two accounts on that list: Account A, a mid-market logistics company that scored well on firmographic fit and ranked 12th by intent signal. Account B, a similar-sized company one tier down, ICP-adjacent but without a standout signal, ranked around 140th.

Under the old cost structure, an illustrative $450 per thorough research pass, that $25,000 budget covers roughly 55 accounts. Account A, well inside that range, gets researched. Account B doesn't come close to making the cut, not because anyone decided it wasn't worth researching, but because the budget ran out 85 places before it got there. Nobody looks at Account B's recent hiring activity or notices the operations-director post that would have made a natural opening line. It just never gets looked at.

Under an illustrative new cost of $18 per account, that same $25,000 covers close to 1,390 accounts. Account B is now well inside the budget, along with hundreds of other accounts that used to sit below the line. Researching Account B now costs about what a team lunch does. Account B doesn't suddenly matter more than it did last quarter. The decision of whether to look at it moved from "can we afford this" to "does this account's fit and intent justify a rep's time," which is a far better question to be funding with a research budget.

Swap in your own budget, your own account counts, your own cost figures. The ratio is what matters. When the line separating "researched" from "invisible" is drawn by a budget instead of a person's calendar, that line moves somewhere very different.

What to change first

Three places to start, roughly in order of how cheap they are to test.

First, audit whether your current tier cutoffs are capacity-driven or fit-driven. Pull the account count in each tier and ask the plain question: was that cutoff set because those are the accounts that matter most, or because that's how many a strategist could research in a quarter? Most programs will find it's some of both, which is useful information on its own. Write down what each tier boundary was originally meant to protect against; if nobody remembers, that's itself a sign the line was drawn by capacity, not judgment.

Second, decide refresh cadence per signal type instead of per calendar quarter. Fast-decaying signals get a fast cadence. Slow-moving firmographic data doesn't need to be re-pulled every time you refresh hiring and funding data. Start with the two or three signal types your team already treats as the most time-sensitive, set an explicit cadence for each, and only then work backward through the rest of the list. Trying to redesign every signal's cadence in a single pass is a reliable way to stall the whole exercise before it produces anything usable.

Third, pilot a coverage expansion into your current Tier 2 or Tier 3 before touching your ICP definition. It's the cheapest place to test this whole argument, because it doesn't require changing who you think your buyer is. It only requires researching more of the accounts you already believed fit. Run the pilot on a fixed slice of fifty to a hundred accounts and give it one full sales cycle before drawing conclusions.

The takeaway

Here's the thesis, plainly, one more time: how fast research got is a footnote. What matters is what a program does once it can no longer justify leaving accounts unresearched.

Tiering, rationed coverage, and calendar-driven refresh cycles were reasonable responses to a real constraint.

That constraint is loosening. For some teams, it's already gone.

What replaces it is structure built around the resource that's still scarce: judgment, attention, and the decision of what to do with an account once you know something about it.