Everybody filters on traffic. Almost nobody filters on the derivative of traffic, which is where the money is. A traffic threshold is a size filter; it tells you how big the store is today and nothing about whether today is a good day to talk to them. The slope tells you the second thing, and the second thing is what determines whether your email lands in a month when the merchant has budget, urgency and permission.
ELDB carries a trend of roughly 170 monthly buckets per store, each with organic traffic and organic positions, on top of Semrush worldwide traffic for 118,324 stores. That is fourteen years of monthly history per domain, available through the trend endpoint. This piece is about what to do with it.
The level tells you size; the slope tells you timing
Two stores at 40,000 monthly organic visits look identical in any filter. One did 12,000 a year ago and one did 90,000. They are opposite businesses. The first has momentum, cash appearing faster than process, and a founder who knows the current setup will not hold. The second has a boardroom argument, a scapegoat, and probably a hiring freeze. Same row in the spreadsheet.
Timing is the variable that nobody controls and almost everybody underweights. Outbound is a bet that the prospect has the problem right now, this month, and not in some abstract future. A store that has tripled its organic traffic in nine months has, mechanically, acquired a set of problems it did not have before: infrastructure strain, support volume, inventory planning, conversion on a traffic mix that has changed underneath it, and a person internally who suddenly has budget and permission to fix things.
That is the closest thing cold outreach has to a warm introduction. You are not guessing that they might care. You are arriving in the specific window where the problem exists and the money to solve it also exists.
What fourteen years of monthly buckets actually buys you
Long history lets you separate things that are otherwise entangled. Seasonality from trend. A platform migration scar from a genuine collapse. A recovery from a flat line that happens to be at a low level. With eighteen months of data you are guessing; with a hundred and seventy buckets you can see the shape of the business across multiple cycles.
Three derived numbers do most of the work. Year-over-year growth, comparing this month against the same month last year, which cancels seasonality for free. Three-month momentum against the prior three months, which catches inflections early. And drawdown from the all-time high, which tells you whether the current number is a peak or the wreckage of one. Compute those three for every store in your segment and you have a targeting layer no static filter can reproduce.
One discipline worth stealing from people who watch these curves professionally: judge the trend by the troughs, not the peaks. Peaks are noisy and often campaign-driven. The floor — the level of the worst months of the year — moves slowly and tells the truth. A rising floor is a business genuinely growing. A rising peak with a flat floor is a business that ran a good campaign.
Four shapes and what each one is worth
The ramp: consecutive year-over-year gains with the floor rising too. This is the best target and it is not close. Nothing is broken, everything feels possible, budget follows momentum, and the merchant has not yet standardised on vendors. Buying is easy in this state because the decision is about upside rather than damage.
The plateau: eighteen months or more of flat after a growth period. Emotionally the hardest and commercially the most interesting if you have a specific unlock, because this merchant already knows something is stuck and has already tried the obvious things. Enthusiasm will not work on them. A named mechanism will.
The cliff: a sharp drop with no recovery. Superficially attractive because the pain is loud. Genuinely dangerous, because the budget often left with the traffic, because roughly half of these are self-inflicted migration disasters, and because every competitor you have is running the same query. The plateau is a better business than the cliff more often than people expect.
The sawtooth: strong annual seasonality on a structurally flat trend. Perfectly viable target, but only inside the run-up window. A gifting store contacted in January is being contacted at the precise moment it has the least appetite, the least cash and the most exhaustion. Sell them in August.
organicPositions is the leading indicator, not the lagging one
The trend carries ranked positions alongside traffic, and the divergence between the two series is where the interesting signal lives. Positions climbing while traffic stays flat usually means the store is accumulating rankings that do not yet produce sessions — long-tail expansion, or a lot of keywords sitting between positions eleven and thirty. Something is being built and it has not surfaced yet.
Traffic falling while positions hold means demand shifted or the results page changed shape and absorbed the clicks. That is not a ranking failure, and telling the merchant that it is will cost you the credibility you spent the email earning. Diagnose the divergence before you name a cause.
Positions falling before traffic falls is the early-warning pattern and the most valuable one in the dataset. Catch a store in that window and you arrive before the crisis is visible internally, before the panic, and before the twelve competitors who are watching the traffic number instead.
The growth window is a real window and it closes
Reason it through from first principles rather than trusting a rule of thumb. A fast-growing store has a period — call it three to nine months — where three conditions hold simultaneously: the founder can see that the current setup will not survive the next doubling, the growth has produced cash, and no vendor relationships have hardened yet. Before that period they cannot pay you. After it they have a stack, a contract and an incumbent.
The trend series lets you detect entry into that window instead of guessing at it. A workable rule: year-over-year growth above your chosen threshold, sustained across at least four consecutive months, with the trailing floor also rising. The multi-month requirement is what keeps a single anomalous spike from polluting the list.
Treat the thresholds as tuning knobs, not facts. Start deliberately tight — a few dozen stores you can genuinely work — and loosen until the reply rate stops justifying the extra volume. Tightening later is much harder than loosening, because by then you have a habit and a spreadsheet full of half-worked rows.
The decline play, and why most people run it badly
The obvious move is to find the losers and sell the rescue. It works less often than it should, for a structural reason: adverse selection. Every consultant in the market can build the same query, so the stores in visible free-fall receive a disproportionate share of all outbound in the category, and their owners have learned to ignore it.
The version that works has three properties. It catches decline early, using positions rather than traffic. It names a specific mechanism rather than offering general help. And it is willing to say the honest thing, including that some of the decline may not be recoverable. Merchants in decline have been told three times already that everything is fixable, and the fourth person who says it sounds exactly like the first three.
Always check whether the decline is theirs or the market’s. If every store in the segment dropped in the same month, you are looking at an algorithm update or a demand shift, and the correct pitch is completely different — and much better, because you can tell them something about their category that they cannot see from inside it.
Operationalise it as a monthly diff, not a one-off query
The play is a scheduled job, not a search. Snapshot the derived metrics per domain every month, diff against the previous snapshot, and alert on transitions: entered ramp, exited ramp, positions inflection, new drawdown, recovery from drawdown. The transitions are the trigger, not the states.
This matters because the list of stores that crossed a threshold this month is small, fresh, and effectively unworked. The list of stores that are currently big is enormous and everybody has it. Working transitions instead of states is the difference between arriving first and arriving fourth.
Keep the derived values in your own store rather than in the CRM. The CRM should receive the label — ramp, plateau, cliff, sawtooth — and the date it last changed. Nobody has ever usefully queried a hundred and seventy monthly buckets from inside a sales tool.
What the trend cannot tell you
It is organic, and it is modelled worldwide. A store putting eighty percent of its revenue through paid social will look permanently flat while tripling. Never diagnose the business from the organic curve alone. Diagnose the channel, and hold the rest as an open question you can ask about.
It is an estimate, directionally strong and absolutely imprecise. Which means you should never quote a number back to a merchant as though it were their analytics. Say that their organic visibility looks like it roughly doubled since the spring, and let them correct you. A correction is a reply, and a reply is the whole objective.
And it is completely silent on margin. A store can triple traffic and be quietly dying. The trend gets you the timing. The catalog and the local profile get you the shape of the business. You need all three before you decide what to sell.